Ekagra Partners, LLC, et al. v. United States

In the United States Court of Federal Claims

Jon D. Levin, Maynard Nexsen PC, with whom were W. Brad English, Emily J. Chancey, Joshua B. Duvall, and Nicholas P. Greer, all of Huntsville, AL, for plaintiff Ekagra Partners, LLC.

* This opinion was originally filed under seal on 28 February 2024 pursuant to the protective order in this case. The Court provided the parties an opportunity to review this opinion for any proprietary, confidential, or other protected information and submit proposed redactions by 6 March 2024 at 5:00 p.m. (ET). The parties proposed redactions by the deadline. The Court accepts the parties’ proposed redactions and reissues the order, with redacted language removed as follows: “[XXXXX].”

Jeffery M. Chiow, of Greenberg Traurig, LLP, with whom were Eleanor M. Ross, Cassidy Kim, and Jordan N. Malone, all of Washington, DC, for plaintiff Unissant, Inc. Daniel J. Strouse, of Cordatis LLP, of Arlington, VA, for plaintiff AttainX, Inc.

John R. Tolle, of Barton, Baker, Thomas & Tolle, LLP, of McLean, VA, for plaintiff Logistics Systems Incorporated.

Mark F. Rosenow, with whom were Matthew E. Feinberg, Lauren R. Brier, Mansitan Sow, and Annie B. Hudgins, of PilieroMazza PLLC, of Washington, DC, for plaintiff Arch Systems, LLC.

Noah B. Bleicher, with whom were Moshe B. Broder, Andrew L. Balland, and Lauren A.

Watson, of Jenner & Block, LLP, of Washington, DC, for plaintiff Alpha Omega Integration, LLC.

John E. Jenson, with whom were Robert C. Starling, Whitney N. Alston, of Pillsbury Winthrop Shaw Pittman, LLP, of McLean, VA, for plaintiff Garud Technology Services, Inc. William E. Weisburg, of Law Offices of William Weisberg PLLC, of McLean, VA, for plaintiff Constellation, Inc.

Beth V. McMahon, of ReavesGovCon Group, of Chesapeake, VA, for plaintiff Chakrabarti Management Consultancy, Inc.

Igor Helman, Senior Trial Counsel, with whom were Andrew Hunter, Trial Attorney, Gisela A. Westwater, Trial Attorney, Steven J. Gillingham, Assistant Director, Patricia M. McCarthy, Director, and Brian M. Boynton, Principal Deputy Assistant Attorney General, Civil Division, Department of Justice, all of Washington, DC, for defendant.

William A. Shook, of The Law Offices of William A. Shook PLLC, of Washington, DC, and Steven Barentzen, of The Law Office of Steven Barentzen, of Washington, DC, for defendant-intervenor Novilo Technology Solutions, LLC.

James C. Fontana, of Fontana Law Group, PLLC, of Tysons, VA, and David R. Warner, with whom was Heather Mims, of Warner, PLLC, of Reston, VA, for defendant-intervenor CAN Softtech, Inc.

James C. Fontana, of Fontana Law Group, PLLC, of Tysons, VA, and David R. Warner, with whom was Heather Mims, of Warner, PLLC, of Reston, VA, for defendant-intervenor NiyamIT, Inc.

Nicholas T. Solosky, with whom was Reginald M. Jones, of Fox Rothschild LLP, of Washington, DC, for defendant-intervenor Catalina Solutions.

Matthew T. Schoonover, with whom was Matthew P. Moriarty, John M. Mattox, Ian P.

Patterson, and Timothy J. Laughlin, of Schoonover & Moriarty LLC, of Olathe, KS, for defendant-intervenor Chevo Consulting, LLC.

HOLTE, Judge.

Plaintiffs Ekagra Partners, LLC (Ekagra), Unissant, Inc. (Unissant), AttainX, Inc.

(AttainX), Logistics Systems, Inc. (LSI), Arch Systems, LLC (Arch), Alpha Omega Integration, LLC (AOI), Garud Technology Services, Inc. (GTS), Constellation, Inc. (Constellation), and Chakrabarti Management Consultancy, Inc. (CMCI) filed separate post-award bid protests challenging the award of two Blanket Purchase Agreements (BPAs) with U.S. Customs and Border Protection (CBP), each worth approximately $450 million over five years. The Solicitation was a General Services Administration (GSA) Multiple Award Schedule (MAS). The Solicitation called for the award of two distinct multiple award BPAs, divided into Professional Services (“Track 1”) and Emerging Technology, Data Transformation, Development and Maintenance Support (“Track 2”). The Solicitation covered both Tracks, and bidders who wished to be evaluated for both requirements were to submit two quotations. CBP received over 100 quotes in total. CBP awarded seven multiple award BPAs for Professional Services and six for Emerging Technology, Data Transformation, Development and Maintenance Support.

For Track 1, the Solicitation stated: “CBP has a requirement for consulting and providing technical support to advise and assist the Government as the Subject Matter Expert (SME) for strategic planning, risk assessment and mitigation, cost analysis, data management and analysis strategies, and senior program management.” AR at 346–47 (TAB 12a⁴) (Solicitation – Statement of Work). The Solicitation also noted “CBP requires vendor support to recommend enhancing and supporting CBP’s emerging technology, data management, reporting and analytical capabilities.” Id. For Track 2, the Solicitation stated: “CBP requires a wide range of services and business disciplines supporting innovation and digital transformation. CBP’s overarching objective is to sustainably improve the total experience and to achieve business agility by integrating people, processes, data, and technology.” AR at 374–75 (TAB 12b⁴) (Solicitation – Statement of Work). When asked at oral argument to provide more concrete detail regarding the Solicitation, the government responded the broad purpose of the contract is to support CBP’s data-related endeavors, including processing customs information, scanning license plates, patrolling the border, and developing language interpretation tools. Oral Arg. Tr. (Tr.) at 104:16–19 (“[GOVERNMENT]: So CBP has a whole broad host of responsibilities that generate[] a ton of data, and they are trying to leverage AI and machine learning to better perform their duties.”), ECF Nos. 143 and 145; Tr. at 104:3–24 (“[GOVERNMENT]: They do license plate scanning. They patrol the borders, . . . [and] when they stop somebody from crossing the border, . . . [they] collect information. . . . [They also collect] information about Customs duties [such as from] cargo ships coming in with lots of both legal goods and illegal goods, and then the legal goods may or may not be properly admitted into the United States because of patent disputes and things like that, or they have to be subject to certain tariffs. . . . [They also frequently have to work] with a person who doesn’t speak the language that the officer speaks at the border.”). CBP intends to incorporate novel applications of AI into these tasks. Tr. at 74:13–15 (“[GOVERNMENT:] [T]his is the type of work that the agency is looking to procure in the future as emerging technology, so novel applications of AI.”). CBP’s team of technical experts reviewed all quotes with this broad purpose in mind, given the highly technical nature of the contract.

For the reasons discussed at length infra, the Court does find CBP erred in its analysis of some Track 2 plaintiff’s RREPs, and the government largely concedes such mistakes, see infra Sections V and VI. Per the terms of the Solicitation for Track 2, however, CBP finding only one RREP failed was acceptable grounds for CBP to eliminate an offeror from award, see infra Sections VI.A. & VI.D. These errors were therefore not prejudicial. In sum, the Court finds no prejudicial error and denies all plaintiffs’ Motions for Judgment on the Administrative Record.

For the foregoing reasons, the Court: (1) denies plaintiff Ekagra’s Motion for Judgment on the Administrative Record; (2) finds as moot plaintiff Ekagra’s Motion for Preliminary Injunction; (3) denies plaintiff Unissant’s Motion for Judgment on the Administrative Record; (4) denies plaintiff AttainX’s Motion for Judgment on the Administrative Record; (5) denies plaintiff LSI’s Motion for Judgment on the Administrative Record; (6) denies plaintiff Arch’s Motion for Judgment on the Administrative Record; (7) denies plaintiff AOI’s Motion for Judgment on the Administrative Record; (8) finds as moot plaintiff AOI’s Motion to Strike; (9) denies plaintiff GTS’ Motion for Judgment on the Administrative Record; (10) denies plaintiff Constellation’s Motion for Judgment on the Administrative Record; (11) denies plaintiff CMCI’s Motion for Judgment on the Administrative Record; (12) grants the government’s Cross-Motions for Judgment on the Administrative Record; (13) grants defendant-intervenor Novilo Technology Solutions, LLC (Novilo)’s Cross-Motions for Judgment on the Administrative Record; (14) grants defendant-intervenor CAN Softtech, Inc. (CAN Softtech)’s Cross-Motion for Judgment on the Administrative Record; (15) grants defendant-intervenor NiyamIT, Inc. (NiyamIT)’s Cross-Motions for Judgment on the Administrative Record; (16) grants defendant-intervenor Catalina Solutions, Inc. (Catalina)’s Cross-Motion for Judgment on the Administrative Record; (17) grants defendant-intervenor Chevo Consulting (Chevo)’s Cross-Motion for Judgment on the Administrative Record; (18) finds as moot defendant-intervenor Chevo’s Motion to Dismiss; and (19) denies defendant-intervenor Catalina’s Motion for Leave to File Supplemental Briefing.

A. Contract Details

U.S. Customs and Border Protection (CBP) issued Solicitation No. 70B04C23QOITESB2 for Professional Services (“Track 1”) and Emerging Technology, Data Transformation, Development and Maintenance Support (“Track 2”). AR at 110 (TAB 10) (Solicitation). The Solicitation “establish[ed] two (2) multiple award Blanket Purchase Agreements (BPAs) for Information Technology (IT) support services to meet the mission and business needs of CBP utilizing the General Service Administration (GSA) Multiple Award Schedule (MAS) Information Technology Category, Small Business Community.” AR at 627 (TAB 18.2) (Solicitation). The Enterprise Small Business (ESB) Request for Quote (RFQ) called for two BPAs: (1) “Professional Services” and (2) “Emerging Technology, Data Transformation, Development and Maintenance Support.” Id. CBP stated it would “issue up to eight (8) awards” per Track. AR at 688 (TAB 18.2) (Solicitation).

For Track 1, the Solicitation stated: “CBP has a requirement for consulting and providing technical support to advise and assist the Government as the Subject Matter Expert (SME) for strategic planning, risk assessment and mitigation, cost analysis, data management and analysis strategies, and senior program management.” AR at 346–47 (TAB 12a⁴) (Solicitation – Statement of Work). The Solicitation also noted, “CBP requires vendor support to recommend enhancing and supporting CBP’s emerging technology, data management, reporting and analytical capabilities.” Id. Relevant to Track 1 plaintiffs’ claims, the Solicitation set the contract aside for small businesses certified under North American Industry Classification System (NAICS) 541511. AR at 110 (TAB 10) (Solicitation). It likewise provided preferential scoring for the use of small business subcontractors. See AR at 696 (TAB 18a¹) (Self-Scoring and Document Verification Worksheet). Relatedly, in instructing offerors to self-score their past performance examples, the Solicitation advised past contracts could not have a value below $1,000,000, AR at 687 (TAB 18.2) (RFQ Amend. 8), and provided preferential scoring for contracts between one and ten million dollars. AR at 223–25 (TAB 10b¹) (Self-Score Worksheet). The Solicitation likewise cautioned offerors offers could be deemed “unacceptable if the . . . labor categories” proposed to fulfill the relevant tasks “[we]re not appropriate.” AR at 688 (TAB 18.2) (Evaluation Criteria).

For Track 2, the Solicitation stated: “CBP requires a wide range of services and business disciplines supporting innovation and digital transformation. CBP’s overarching objective is to sustainably improve the total experience and to achieve business agility by integrating people, processes, data, and technology.” AR at 374–75 (TAB 12b⁴) (Solicitation – Statement of Work). The Solicitation contained multiple “Tasks,” five of which are relevant in this protest: (1) Digital Transformation and Software Development Support; (2) Operational Maintenance Support Services; (3) Artificial Intelligence/Machine Learning Support Services; (4) Robotic Process Automation Support Services; and (5) Security and Privacy Support. AR at 376–380 (TAB 12b⁴) (Solicitation – Statement of Work). Those Tasks stated:

Task 1 – Digital Transformation and Software Development Support . . .

The Contractors shall have proven expertise in moving customer organizations from a waterfall environment to an agile software development and maintenance environment. . . . The Contractor shall support new applications and existing applications as required[, including] . . . [s]oftware development practices. . . .

Task 2 – Operational Maintenance Support Service . . .

The Contractor shall perform operational maintenance support activities for the newly developed and enhanced applications developed by the Contractor. The Contractor shall perform defect resolution and corrective maintenance categorized as break/fix. . . . The Government and the Contractor shall collaborate and identify what corrective action is required. The Contractor shall provide 24x7x365, on-call remote support to corrective issues on applications identified by CBP as requiring 24x7x365 service. The Contractor shall develop mechanisms to automate notification of issues requiring correction; any issue that impacts the ability of a user to interact with applications must be resolved immediately. The Contractor shall notify the Government when Corrective issues are corrected.

. . .

Task 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services . . .

There is a . . . need for artificial intelligence (AI) and machine learning (ML) expertise. . . . The Contractor support shall include, but is not limited to the following: . . .

•Developing an approach to labeling, annotation, and ontology development to support application of AI at CBP.

•Applying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.

•Establishing and managing a centralized library of annotated data for use across the organization for AI model training.

. . .

•Managing and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.

. . .

Task 4 – Robotic Process Automation (RPA) Support Services . . .

The Contractor shall provide the full range of development, maintenance, architecture, technical, management, planning, communications and engagement support services with the associated skills and qualifications required to support CBP Robotic Process Automation (RPA) Community of Practice (COP) Program.

. . .

Task 5 – Security and Privacy Support . . .

The Contractor shall provide security and privacy support to CBP’s portfolio of

projects to attain the appropriate authorities and support security / privacy activities. The Contractor shall support the development of documentation including Privacy Threshold Assessments (PTAs), Authorizations to Test (ATTs), and Authorizations to Operate (ATOs) for . . . CBP’s portfolio of projects.

•Coordinate with stakeholders to support . . . CBP security / privacy process and work to reconcile issues and blockers impeding approval of PTA, ATT, or ATO.

•Communicate with project owners on status of PTA, ATT, or ATO and escalate issues to appropriate Government POCs. . . .

AR at 376–380 (TAB 12b⁴) (Solicitation – Statement of Work).

CBP stated the following as to its evaluation method and basis for award in the Solicitation:

The self-scored scorecard methodology approach will be utilized for this procurement. Only submissions from a valid GSA MAS Schedule Information Technology Category, IT Services, IT professional services (SIN 54151S) who are Small Business Schedule Holders will be accepted. The Government will compile a ranking of quotes by reviewing the quotations for compliance, accuracy, and completion of all required areas to select the technically highest rated submissions.

During this process, Quoters’ self-scores can either remain the same or decrease.

The highest ranked quoters will be evaluated for relevancy on the two (2) required RREPs submitted for each task areas. The RREPS will be evaluated for similarity in scope and complexity to the task areas identified in the ESB BPA scope of work.

If the Government determines any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated. Any additional RREP submissions will not be substituted by the Government for any RREP submissions that are deemed not relevant. Once the Government has evaluated a Quoter’s Attachment 1 Section 2 for relevancy, the Government will then continue to review and validate the remaining sections of Attachment 1. If the Government deems irrelevant and subsequently eliminates any of the additional RREP submissions from Section 2, the Government will also eliminate its corresponding points associated with that RREP from all other self-scoring sections.

The Government will review the Quoter’s proposed labor categories and corresponding labor rates for each task area in accordance with the Quoter’s GSA MAS Schedule Contract to determine reasonableness. The Government will review the Quoter’s proposed labor categories for each task to determine appropriateness.

AR at 689 (TAB 18.2) (Solicitation). CBP awarded seven Professional Services BPAs to the following contractors: Zantech IT Services Inc.; Savvee Consulting Inc.; Flatter, Inc.; Dignari, LLC; Cybermedia Technologies Inc. DBA CTEC; Chevo; and Catalina. AR at 10165 (TAB 74) (SSA Award Memo). CBP awarded BPAs to the following six awardees for Track 2: Tarkik Solutions Inc.; NiyamIT; Novilo; INDEV, LLC; Centrifuge, LLC; and CAN Softtech. Id.

B. Procedural History

On 19 September 2023, Ekagra filed its Complaint in this bid protest, ECF No. 1. Also on 19 September 2023, Ekagra filed its Motion for Preliminary Injunction, ECF No. 2.1 Novilo, ECF No. 11, and CAN Softtech, ECF No. 12, filed unopposed motions to intervene on 20 and 21 September 2023, respectively. The Court granted these Motions to Intervene on 24 September 2023, ECF No. 14. On 27 September 2023, Unissant filed its Complaint in Unissant, Inc. v. United States, No. 23-1667 (Fed. Cl. Sept. 27, 2023), along with a notice the protest is directly related to case number 23-1610, captioned Ekagra Partners, LLC v. United States. See Compl., Unissant, Inc. v. United States, No. 23-1667 (Fed. Cl. Sept. 27, 2023), ECF No. 1; Notice of Directly Related Cases, Unissant, Inc. v. United States, No. 23-1667 (Fed. Cl. Sept. 27, 2023), ECF No. 5. On 28 September 2023, AttainX filed its Complaint in AttainX, Inc. v. United States, No. 23-1668 (Fed. Cl. Sept. 28, 2023), along with a notice the protest is directly related to case number 23-1610. See Compl., AttainX, Inc. v. United States, No. 23-1668 (Fed. Cl. Sept. 28, 2023), ECF No. 1; Notice of Directly Related Cases, AttainX, Inc. v. United States, No. 23-1668 (Fed. Cl. Sept. 28, 2023), ECF No. 5. On 28 September 2023, LSI filed its Complaint in Logistics Systems Incorporated v. United States, No. 23-1675 (Fed. Cl. Sept. 28, 2023), along with a notice the protest is directly related to case number 23-1610. See Compl., Logistics Sys. Inc. v. United States, No. 23-1668 (Fed. Cl. Sept. 27, 2023), ECF No. 1; Notice of Directly Related Cases, Logistics Sys. Inc. v. United States, No. 23-1668 (Fed. Cl. Sept. 28, 2023), ECF No. 6. On 28 September 2023, Arch filed its Complaint in Arch Systems, LLC v. United States, No. 23-1676 (Fed. Cl. Sept. 28, 2023), along with a notice the protest is directly related to case number 23-1610. See Compl., Arch Sys., LLC v. United States, No. 23-1676 (Fed. Cl. Sept. 27, 2023), ECF No. 1; Notice of Directly Related Cases, Arch Sys., LLC v. United States, No. 23-1676 (Fed. Cl. Sept. 28, 2023), ECF No. 2. In light of the number of related protests and in anticipation of additional protests being filed in this court, on 29 September 2023, the Court rescheduled the initial status conference for 5 October 2023 and instructed the parties to submit an updated joint status report (JSR) by 3 October 2023, ECF No. 28.

1Plaintiff Ekagra agreed at oral argument its motion for preliminary injunction is moot. Oral Arg. Tr. (Tr.) at 15:23–16:1 (“[THE COURT:] Ekagra initially filed a motion for preliminary injunction, ECF number 2, early in the case. Do you agree that this motion is now moot? [EKAGRA]: Yes, Your Honor.”), ECF Nos. 143 and 145.

On 3 October 2023, AOI filed its complaint in Alpha Omega Integration, LLC v. United States, No. 23-1721 (Fed. Cl. Oct. 3, 2023), along with a notice the protest is directly related to case number 23-1610. See Compl., Alpha Omega Integration, LLC v. United States, No. 23-1721 (Fed. Cl. Oct. 3, 2023), ECF No. 1; Notice of Directly Related Cases, Alpha Omega Integration, LLC v. United States, No. 23-1721 (Fed. Cl. Oct. 4, 2023), ECF No. 4. During the 5 October 2023 status conference, the parties agreed case numbers 23-1610, 23-1667, 23-1668, 23-1675, 23-1676, and 23-1721 should be consolidated. On 6 October 2023, the Court consolidated case Nos. 23-1667, 23-1668, 23-1675, 23-1676, and 23-1721 with lead case No. 23-1610, ECF No. 31. On 12 October 2023, GTS filed its complaint along with a notice its protest is directly related to case No. 23-1610. See Compl., Garud Technology Services, Inc. v. United States, No. 23-1792 (Fed. Cl. Oct. 12, 2023), ECF No. 1; Notice of Directly Related Cases, Garud Technology Services, Inc. v. United States, No. 23-1792 (Fed. Cl. Oct. 12, 2023), ECF No. 5. On 16 October 2023, Constellation filed a complaint along with a notice its protest is directly related to case No. 23-1610. See Compl., Constellation, Inc. v. United States, No. 23-1814 (Fed. Cl. Oct. 16, 2023), ECF No. 1; Notice of Directly Related Cases, Constellation, Inc. v. United States, No. 23-1814 (Fed. Cl. Oct. 16, 2023), ECF No. 3.

On 17 October 2023, the Court consolidated case Nos. 23-1792 and 23-1814 with lead case No. 23-1610, ECF No. 42. On 17 October 2023, NiyamIT filed a motion to intervene as intervenor-defendant, ECF No. 45. On 19 October 2023, Catalina filed a motion to intervene as intervenor-defendant, ECF No. 52. Also on 19 October 2023, CMCI filed its Complaint along with a notice its protest is directly related to case No. 23-1610. See Compl., Chakrabarti Management Consultancy, Inc. v. United States, No. 23-1844 (Fed. Cl. Oct. 19, 2023), ECF No. 1; Notice of Directly Related Cases, Chakrabarti Management Consultancy, Inc. v. United States, No. 23-1844 (Fed. Cl. Oct. 19, 2023), ECF No. 5. On 19 October 2023, the Court consolidated case No. 23-1844 with 23-1610 and granted NiyamIT and Catalina’s Motions to Intervene, ECF No. 56.

On 26 October 2023, the government filed (1) a status report notifying the Court it was taking corrective action with Arch’s quote as CBP had not evaluated Arch’s offer due to a “system error,” ECF No. 61, and (2) its corresponding Unopposed Motion to Remand Arch’s claims, ECF No. 62, which the Court granted the next day, ECF No. 63. On 1 November 2023, GTS filed its Amended Complaint (“GTS Am. Compl.”), ECF No. 64. On 6 November 2023, Chevo filed its Motion to Intervene, ECF No. 65, which the Court granted the next day, ECF No. 67.

On 8 November 2023, four plaintiffs filed their respective Motions for Judgment on the Administrative Record. See ECF No. 68 (“AttainX MJAR”); ECF No. 74 (“Ekagra MJAR”); ECF No. 75 (“CMCI MJAR”); ECF No. 77 (“Unissant MJAR”). Also on 8 November 2023, Arch filed its Amended Complaint in light of corrective action (“Arch Am. Compl.”), ECF No 76. On 13 November 2023, the government filed a JSR notifying the Court it completed corrective action as to Arch’s quote and did not award a portion of the contract to Arch, ECF No. 78. On 15 November 2023, five plaintiffs filed their respective MJARs. See ECF No. 80 (“LSI MJAR”); ECF No. 81 (“AOI MJAR”); ECF No. 82 (“GTS MJAR”); ECF No. 83 (“Constellation MJAR”); ECF No. 84 (“Arch MJAR”).

On 28 November 2023, the government filed its first Cross-MJAR and Response (“Gov’t’s Cross-MJAR and Resp. to Group 1”), ECF No. 93, with the Tran Declaration and Gartner Article attached (“Tran Decl. and Gartner Article”), ECF No. 93-1. On 28 November 2023, three intervenors also filed their Cross-MJARs and Responses. See ECF No. 90 (“NiyamIt Cross-MJAR and Resp.”); ECF No. 91 (“Can Softtech Cross-MJAR and Resp.”); ECF No. 92 (“Novilo Cross-MJAR and Resp. to Group 1”). On 8 December 2023, four additional intervenors filed their Cross-MJARs and Responses. See ECF No. 101 (“Niyamit Cross-MJAR and Resp.”); ECF No. 103 (“Chevo Cross-MJAR and Resp. and Mot. to Dismiss”); ECF No. 104 (“Novilo Cross-MJAR and Resp. to Group 2”); ECF No. 105 (“Catalina Cross-MJAR and Resp.”). With its Cross-MJAR, Chevo filed a Motion for Partial Dismissal for Lack of Subject Matter Jurisdiction. See ECF No. 103 (“Chevo Mot. to Dismiss”). On 8 December 2023, the government filed its second Cross-MJAR and Response (“Gov’t’s Cross-MJAR and Resp. to Group 2”), ECF No. 106, and attached the Duneja Declaration (“Duneja Decl.”), ECF No. 106-1. Also on 8 December 2023, Attainx (ECF No. 99) and CMCI (ECF No. 100) filed their Responses and Replies. On 11 December 2023, Ekagra (ECF No. 109) and Unissant (ECF No. 110) filed their respective Responses and Replies. On 17 December 2023, LSI filed its Response and Reply (“LSI Resp.”), ECF No. 111. On 18 December 2023, four plaintiffs filed their Responses and Replies. See ECF No. 113 (“AOI Resp.”); ECF No. 114 (“Constellation Resp.”); ECF No. 115 (“Arch Reply”); ECF No. 116 (“GTS Resp.”). Also on 18 December 2023, AOI filed its Motion to Strike the Duneja Declaration (“AOI Mot. to Strike”), ECF No. 112, and its Cross-MJAR and Response (“AOI Cross-MJAR and Resp.”), ECF No. 113. On 21 December 2023, the government filed its first Response (“Gov’t’s Group 1 Reply”), ECF No. 124, and Novilo filed its first Response (“Novilo Group 1 Resp.”), ECF No. 122. On 28 December 2023, Chevo, Novilo, and Catalina filed their Responses and Replies. See ECF No. 126 (“Chevo Reply.”); ECF No. 129 (“Novilo Group 2 Resp.”); ECF No. 130 (“Catalina Reply”). On 28 December 2023, the government filed its second Reply (“Gov’t’s Group 2 Reply”), ECF No. 131. The next day, the government filed its Response to AOI’s Motion to Strike, ECF No. 132. On 5 January 2024, AOI filed its Reply in support of its Motion to Strike (“AOI Mot. to Strike Reply”), ECF No. 136.

On 17 and 18 January 2024, the Court held oral argument on all pending motions, ECF No. 89. On 16 February 2024, Catalina filed a motion for leave to file a supplemental brief, ECF No. 165. On 19 February 2024, Arch filed a response to Catalina’s Motion, ECF No. 168, and on 21 February 2024, Catalina filed its Reply, ECF No. 174.

Plaintiffs LSI, Arch, GTS, and Constellation challenge CBP’s award decision related to Track 1. The four Track 1 plaintiffs’ arguments can generally be grouped into three distinct categories: (1) challenges related to the size of subcontractors; (2) challenges related to RREPs; and (3) related miscellaneous challenges. The Court describes each, as well as the specifics of the parties’ arguments, in turn.

A. Challenges Related to Subcontractor Size

Track 1 plaintiffs LSI, Arch, and Constellation argue CBP wrongfully found subcontractors used by each plaintiff were “other than small under [North American Industry Classification System (NAICS) code] 541511 and [therefore] made a[n improper] point adjustment.” LSI Reply at 2. Under the Solicitation, bidders were entitled to “preferential scores favoring lower percentages of large business utilization.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 61 n.10; see also AR at 696 (TAB 18a¹) (Self Scoring and Document Verification Worksheet). LSI, Arch, and Constellation all assert CBP’s “point adjustment[s] [related to their subcontractors] should be reversed,” LSI MJAR at 4, because the agency improperly “determined that . . . [their] subcontractors . . . [were] not [] small business[es], resulting in losing points on the small business participation score.” Constellation MJAR at 3; see also Arch MJAR at 1, 20–25 (“CBP improperly deducted 375 points based on Arch Systems’ use of a subcontractor that DBP deemed (incorrectly and without authority) to be a large business.”). At oral argument, plaintiffs clarified their primary issue is the FAR makes clear “it’s the size standard associated with the NAICS codes [that matters, detailing number of employees and/or threshold dollar amount.] Not the NAICS code per se.” Tr. at 381:6–8 (Constellation). Thus, plaintiffs argue CBP should not have decreased their scores because, although their subcontractors were not self-certified under NAICS code 541511, they were beneath the relevant size threshold. See, e.g., Tr. at 357:14–358:4. In response, the government contends “CBP did not act arbitrarily or capriciously when it disallowed small business preferential points” for plaintiffs’ subcontractors. Gov’t’s Group 2 Reply at 7. Multiple defendant intervenors, although making distinct legal arguments, echo the government’s sentiment the Court should not disturb CBP’s determinations related to plaintiffs’ subcontractors. See Catalina Cross-MJAR and Resp. at 2–3 (“One offeror (Arch Systems) included RREPs from a large business subcontractor in its proposal and the other (Catalina Solutions) did not.”); Chevo Cross-MJAR and Resp. at 5–6 (“[T]he Group 2 protesters challenge the agency’s decisions related to subcontractor size . . . [t]his Court does not have jurisdiction to adjudicate these claims, so they must be dismissed.”).

B. Challenges Related to RREP Point Deductions

Track 1 plaintiffs LSI and Arch also challenge CBP’s deduction of points related to RREPs. LSI MJAR at 7 (citing AR at 10139–40 (TAB 73.10) (LSI Track 1 Memorandum); 9702–03 (Tab 63) (LSI Self-score Worksheet)); Arch MJAR at 15 (quoting AR at 687 (TAB 18.2) (RFQ Amend. 8)). “The minimum value for an RREP under the RFQ was one million dollars.” Gov’t’s Cross-MJAR and Group 2 Resp. at 61 (citing AR at 687 (Tab 18.2) (RFQ Amend. 8)). “In its evaluation, CBP . . . sought to verify that the projects submitted as RREPs were at least of a value of one million dollars. . . .” Id. RREPs below the $1,000,000 threshold were disregarded by CBP, and those with values greater than $10,000,000 were disfavored. See Gov’t’s Group 2 Reply at 5 (citing AR at 223–25 (TAB 10b¹) (Self-Score Worksheet)) (explaining “the scoring rubric . . . favored projects with lower ‘total’ values,” meaning RREPs of a value between one and ten million dollars were entitled to the most points).

LSI alleges CBP “found that [its] . . . Task 1 RREP 5 (AR [at] 6882), Task 4 RREP 5 (AR [at] 6924), and Task 7 RREP 3 (AR [at] 6961) are the same RREP, . . . [each with a] contract value [of] $758,000, less than the required $1M minimum,” so CBP “removed these RREPs.” LSI MJAR at 7 (citing AR at 10139–40 (TAB 73.10) (LSI Track 1 Memorandum); AR at 9702–03 (Tab 63) (LSI Self-score Worksheet)). According to LSI, “[t]hese RREPs should not have been removed and the deducted points should be restored” because the database used by CBP to validate the value of these RREPs—the Federal Procurement Data System (FPDS)— “does not reflect a full picture,” and LSI could “have provided an explanatory note describing the errors in FPDS” had CBP asked for such. Id. at 7–8. Further, LSI alleges its Task 1 RREP 6, a Department of Commerce (DOC) contract, “should not have been removed.” Id. at 8.

Similarly, according to Arch, each “RREP was to contain, in pertinent part, the associated ‘Requirement or Project Total Value (project total value of work performed if performed as a sub-contractor).’” Arch MJAR at 15 (quoting AR at 687 (TAB 18.2) (RFQ Amend. 8)). Arch therefore “disclosed the value of the Requirement in question, a separate contract line item . . . as $[XXXXXXXXXXXX],” even though the “entire task order was valued at $[XXXXXXXXXXXX].” Id. at 16 (emphasis added). In other words, unlike CBP, Arch interpreted the RFQ as permitting “offerors . . . to rely on specific contract requirements, as opposed to entire task order values.” Id. at 17. Should the Court find CBP’s evaluation of “Arch Systems under Task 5 was . . . consistent with the RFQ,” however, Arch argues there is a “latent ambiguity in the RFQ,” id. at 19, meaning an ambiguity “‘not apparent on the face of the solicitation [or] . . . ascertainable through reasonable or customary care,’” warranting injunctive relief. Id. (citing Input/Output Tech., Inc. v. United States, 44 Fed. Cl. 65, 72 n.10 (1999)). In response to LSI, the government argues “CBP reasonably removed RREPs and their associated points where it could not verify the projects’ claimed values.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 61. Turning to Arch, the government contends Arch’s “extraction of a contract portion was in conflict with the RFQ, which directed quoters to score their RREPs based on the ‘Total Value’ of the ‘Requirement or Project.’” Id. at 46 (quoting AR at 687 (TAB 18.2) (RFQ Amend. 8)). The government also notes “the Q&A process” further “clarified . . . quoters could not break a requirement or project into subparts in order to reach the higher point category.” Id. (citing AR at 722 (TAB 18c¹) (Track 1 Q&A)).

C. Additional Challenges by Track 1 Plaintiffs

Track 1 plaintiffs Arch and GTS challenge three other aspects of this procurement.

Arch first alleges an additional claim related to its argument CBP improperly downgraded its score upon finding its subcontractor [XXXXXXXXXX] is not a small business under NAICS code 541511. Namely, Arch alleges “CBP engaged in disparate treatment during its evaluation of offerors by reducing Arch Systems’ score due to its subcontracting to a purported large business while not deducting points from [awardee Chevo] for the same reason.” Arch MJAR at 25.

In response, the government contends “Arch Systems’ arguments demonstrate its fundamental misunderstanding of the evaluation criteria.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 53–54 (“Because CBP treated all three quoters in strict accordance with the terms of the RFQ as they applied to their distinct proposals, the Court should dismiss the claim.”). Namely, the government clarifies Chevo “awarded [itself] the appropriate score in the first instance, [so] there was no basis for CBP to deduct points.” Id. at 54–55. On the other hand, “Arch Systems . . . submitted RREPs from a large business but had claimed points as though it was a small business,” warranting a point adjustment. Id. at 54.

In its final argument, Arch calls out CBP’s award to Catalina explicitly and notes “Catalina should have been deemed ineligible for award.” Arch Systems MJAR at 30. Arch argues specifically the RFQ required “‘[a] minimum of one (1) RREP must be submitted by the Prime Contractor in each task area.’” Id. (quoting AR at 686 (TAB 18.2) (Solicitation)). According to Arch, the prime contractor “on Catalina’s proposal is ‘Catalina Solutions, LLC,’ ‘a joint venture between Catalina Associates and Versar National Security Solutions LLC (formerly BayFirst Solutions LLC), a subsidiary of Versar, Inc.’” Id. at 30–31 (citing AR at 2226 (TAB 26a) (Catalina Offer); and AR at 10527 (TAB 87) (Catalina Order)). Per Arch, however, “Catalina did not provide ‘[a] minimum of one (1) RREP . . . by the Prime Contractor in each task area.’” Id. at 31 (citing AR at 2244–2338 (TABs 26a–26b²) (Catalina Offer)). In response, the government alleges Arch misunderstands the “meaning given to th[e] term [prime contractor] in the RFQ and the guidance incorporated into it through the Q&A.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 55. Namely, according to the government, CBP made clear this requirement meant “RREP[s] c[ould] come from the small [partner], large [partner], or [joint venture].” Id. (quoting AR (TAB 19.2) (Solicitation Clarifications)) (“[A]s a clarification of answers provided during an earlier round of Q&A, [CBP stated] that ‘CBP will consider [RREPs for] work done and qualifications held individually by each partner to the joint venture as well as any work done by the joint venture itself previously.”). Catalina therefore “fulfilled the RFQ’s requirement of submitting a minimum of one RREP from the prime contractor for each task area.” Id. at 56 (citing AR at 2244–56 (TABs 26a–26b²) (Catalina Offer)).

Finally, plaintiff GTS brings a novel challenge related to both Tracks 1 and 2. GTS argues CBP’s “determination that GTS proposed an inappropriate skill mix for [staffing both tracks] . . . is arbitrary and capricious” because there “is no evidence in the record that . . . any evaluator[] meaningfully considered GTS’s proposed mix of labor categories.” GTS MJAR at 10. Specifically, GTS alleges “[t]here is no substantive discussion in the [source selection authority (SSA)] memo[randum]” regarding CBP’s conclusion “GTS[] proposed an inappropriate skill mix.” Id. GTS likewise argues “there is no evidence elsewhere in the record that the Agency performed a substantive analysis of GTS’s proposed labor mix.” Id. at 16. Further, for Track 2, GTS alleges the record is unclear and inconsistent regarding why GTS was eliminated, meaning CBP’s decision was arbitrary and capricious. GTS Reply at 13 (citing AR at 10161 (TAB 74) (SSA Memorandum)). According to GTS, CBP “seemed not even to understand the labor categories that GTS proposed,” which were intentionally broad to “accommodate the diverse scope of work and open-ended nature of future requirements contemplated by the Solicitation.” GTS MJAR at 19, 22 (citing AR at 627 (TAB 18.2) (Solicitation)).

The government disagrees with GTS’ characterization of CBP’s review. Specifically, according to the government, CBP found “that, rather than tailoring its list of labor categories to be proposed for each task under the [scope of work (SOW)], GTS had ‘used a standardized boiler-plate listing of [labor categories] . . . [and] applied that listing to every task.’” Gov’t’s Cross-MJAR and Resp. to Group 2 at 16, 17 (quoting AR at 9657 (TAB 61.2) (Chamblin E-mail)) (“‘[GTS’] quote provided the same entire catalog of more than [XXX] labor categories for all task areas, which did not demonstrate appropriateness for each task area.’”). Thus, per the government, “CBP correctly excluded GTS from competition” because GTS did not follow the RFQ’s requirement “the skill mix proposed by a quoter needed to be adequate to perform the task under consideration, and not bloated with irrelevant and inapplicable positions.” Id. at 28 (first citing AR at 627, 688–89 (TAB 18.2) (Solicitation); then citing AR at 733 (TAB 18c¹) (Track 1 Q&A); and then citing AR at 747 (TAB 18c²) (Track 2 Q&A)).

Overall, plaintiffs argue these point reductions and eliminations were prejudicial and entitle them to injunctive relief. See LSI MJAR at 8–10; Constellation MJAR at 6–7; Arch Systems MJAR at 9, 33–40; GTS MJAR at 32–33. In contrast, in concluding its Reply, the government opposes injunction and notes “should th[e] Court conclude that the record is not adequate for it to examine the agency’s rationale on any particular issue . . . the proper remedy is remand to CBP for more explanation.” Gov’t’s Group 2 Reply at 32 (citing Fla. Power & Light Co. v. Lorion, 470 U.S. 729, 744 (1985)).

Plaintiffs Ekagra, Unissant, AttainX, CMCI, AOI, and Constellation challenge CBP’s award decision related to Track 2. The Track 2 plaintiffs’ arguments can largely be grouped into the following areas: (1) challenges related to CBP’s alleged application of unstated criterion; (2) challenges related to CBP’s alleged lack of documentation and consideration of RREPs; (3) challenges related to CBP’s alleged disparate treatment of offerors; (4) challenges related to prejudice; and (5) plaintiffs’ requests for a permanent injunction. The Court describes each, as well as the specifics of the parties’ arguments, in turn.

A. Challenges Related to CBP’s Alleged Application of Unstated Criterion

Many plaintiffs argue the agency used unstated, strict evaluation criteria in determining the “similarity” of plaintiffs’ RREPs with the various tasks. AR at 689 (TAB 18.2) (Solicitation) (“The RREPS will be evaluated for similarity in scope and complexity to the task areas identified in the ESB BPA scope of work.”) (emphasis added); Ekagra MJAR at 14–17; Unissant MJAR at 10–18; AttainX MJAR at 10–11; AOI MJAR at 11; CMCI MJAR at 5–9; Constellation MJAR at 5. The government argues “plaintiffs’ arguments are analogous [because they] . . . argue that the agency used unstated evaluation criteria by insisting that the RREPs be ‘identical,’ rather than ‘similar’ to the task areas in the SOW.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 24.

Ekagra argues CBP used unstated evaluation criteria under Task 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. Ekagra MJAR at 14–17 (citing AR at 10088 (TAB 71) (ET TET Consolidated Review)) (“[I]n evaluating Ekagra[’s] second RREP example under Task 3, the Agency deemed the example unacceptable because it did not meet four [performance work statement (PWS)] ‘requirements’ . . . [R]ather than evaluating Ekagra’s RREP 2 to determine whether it was similar in scope and complexity, the Agency turned its bullets into baseline ‘requirements’ and then compared them in rote fashion to Ekagra’s solution.”). Ekagra explains CBP “required quotes to be the same as the list for Task 3 instead of having characteristics in common with the list in Task 3 . . . [and] thus imposed an ‘identical’ requirement rather than a ‘similarity’ requirement.” Id. at 17.

Unissant argues CBP used an unstated evaluation criterion for Tasks 2 – Operational Maintenance Support Service, 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services, 4 –RPA Support Services, and 5 – Security and Privacy Support, see supra Section I.A. Unissant MJAR at 10–18. For Task 2 - Operational Maintenance Support Service, Unissant argues CBP “irrationally determined that Unissant’s RREP did not demonstrate performance that was similar to the scope of the Task 2 requirements” “[d]espite the fact that both RREPs for Task 2 referenced Unissant’s experience performing ‘Operational Maintenance Support Services’” and “explained [Unissant’s] experience providing end-to-end operations and maintenance support for critical IT Systems, including new applications and systems implemented by Unissant [XXX XXXXXXXX] under these contractual efforts.” Id. at 12 (citing AR at 8377 (TAB 45b²) (Unissant Offer)). For Task 3 - Artificial Intelligence/Machine Learning (AI/ML) Support Services, Unissant argues CBP’s “arbitrary and capricious search for key words and specific activities in Unissant’s proposal, which was not included in the evaluation criteria stated in the solicitation, resulted in the Agency improperly determining that Unissant’s RREPs were not of similar scope and complexity to the PWS tasks” even though “Unissant discussed how it had built an AI model (or prototype), including ‘data labeling, ontology and annotations needed for the AI model to run.’” Id. at 13–14 (quoting AR at 8390 (TAB 45b²) (Unisssant Offer)). For Task 4 - RPA Support Services, Unissant argues CBP “irrationally concluded that Unissant had not performed RPA work apparently because Unissant did not use the word ‘robotics’ in its first Primary RREP for this task [but] . . . Unissant addresses each of the actual tasks under this Task [in its RREP], such as governance activities and the ATO Process.” Id. at 16 (comparing AR at 250 (TAB 10b⁵) (Task 2, Solicitation) with AR at 8402–03 (TAB 45b²) (Unissant Offer)). For Task 5 - Security and Privacy Support, Unissant argues “rather than considering whether the offerors provided RREPs which reflected ‘similarity in scope and complexity to the task areas in the’ PWS . . . the Agency applied a different, unstated requirement that the offerors reference each bullet point under the task area,” as “Unissant’s . . . communication or coordination . . . activities [in its proposal] were engrained within every step of the processes that Unissant . . . describe[d]” but Unissant was eliminated nonetheless. Id. at 17–18 (quoting AR at 689 (TAB 18.2) (Solicitation)).

AttainX argues CBP used an unstated evaluation criterion for Task 2 - Operational Maintenance Support Service because “the agency evaluated AttainX as not demonstrating similarity in scope and complexity with the ‘automated notification of issues requirement’” even though AttainX “demonstrated experience with automated notification of issues requiring correction” as “‘[t]he system [in AttainX’s RREP] automatically creates a ticket as soon as it receives an email or voicemail from the system user.’” AttainX MJAR at 10–11 (quoting AR at 10561 (TAB 88c) (AttainX Brief Explanation); and then quoting AR at 1709 (TAB 24b²) (AttainX RREPs)).

CMCI also argues CBP used an unstated evaluation criterion in determining “similarity” for Task 3 and Task 5. CMCI MJAR at 5–9. For Task 3 - Artificial Intelligence/Machine Learning (AI/ML) Support Services, CMCI disputes CBP’s conclusion “CMCI’s proposal failed to establish its experience [d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI . . . [when] CMCI’s proposal did show this experience” and counters its RREP established this experience but used the term “collation” rather than the term “ontology of a program.” Id. at 8 (internal quotations omitted). Also, while CBP concluded “CMCI’s proposal had not demonstrated experience with ‘[a]pplying labeling ontology and annotating data of varying types from various sources according to the ontology,’” CMCI argues it included AI tools and data robots in its proposal. Id. at 8–9. Further, CBP “found lacking CMCI’s experience [e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training,” but according to CMCI, its “proposal discusses its experience utilizing statistical and predictive models, using home-grown and advanced AI/ML tools that allowed CBP to harness real-time, aggregate rich data and gain agility for immediate action.” Id. at 9 (internal quotations omitted). Also, CBP “found that CMCI did not show experience [m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination, identification of any Plan of Action and Milestones (POA&Ms), and support to address them” even though, per CMCI, “CMCI’s work on all phases of Enterprise Life Cycle would encompass promoting applications to live production, which would entail adherence to ATT, ATO processes and mitigations of items collected as part of POAMs.” Id. (internal quotations omitted). For Task 5 - Security and Privacy Support, CMCI highlights CBP concluded CMCI has “no indication of supporting federal security guidelines per the SOW” even though (1) “the RFQ and SOW did not state that ‘supporting federal security guidelines’ was required”; (2) “CMCI’s proposal . . . address[ed] federal security guidelines, discussing its experience with network and Cloud security (in addition to other parts of the write-up pertaining to application security)”; and (3) “federal security guidelines . . . cite the need and standards for security at the system/application level.” Id. at 5–6.

AOI also argues CBP used an unstated evaluation criterion in determining “similarity” for Task 3 - Artificial Intelligence/Machine Learning (AI/ML) Support Services and protests several CBP conclusions related to Task 3’s sub-bullets, see supra Section I.A. AOI MJAR at 11 (citing AR at 684–90 (TAB 18.2) (Solicitation)) (“Under any reading, the Solicitation did not require that a vendor demonstrate identical experience in its RREPs. . .”). Related to its MJAR arguments, AOI filed its Motion to Strike the Duneja Declaration attached to the government’s Cross-MJAR to Group 2 plaintiffs from the record because it provided “extra-record explanations . . . to supplant CBP’s superficial findings.” AOI Mot. to Strike Reply at 1–2.

B. Challenges Related to CBP’s Alleged Lack of Documentation and Consideration of RREPs

Plaintiffs Ekagra and Constellation argue CBP did not properly consider, or document its consideration of, their RREPs. Ekagra MJAR at 9–12; Constellation MJAR at 3–6. Ekagra argues CBP failed to consider its RREP 2 for Task 1 by “ignor[ing] Ekagra’s discussion of code development.” Ekagra MJAR at 9. Constellation likewise argues CBP improperly evaluated its quote for RREP 1, Task 2 and RREPs 1 and 2, Task 3. Constellation MJAR at 3–6.

Although Unissant, AttainX, CMCI, and AOI initially made similar claims, at oral argument they agreed their arguments related to CBP’s lack of documentation are the same as their arguments related to unstated criterion. Tr. at 245:17–22 (“THE COURT: . . . So your argument is related to CBP ignoring the contents for Tasks 2, 3, 4 and 5 . . . [A]ren’t these arguments largely the same as those related to the original arguments of CBP’s alleged use of unstated criteria? [UNISSANT]: Yes, Your Honor.”); Tr. at 246:11–17 (“THE COURT: AttainX, . . . regarding CBP not addressing automating notification of issues requiring correction, . . . does your argument that CBP evaluated the proposal arbitrarily and capriciously largely overlap with the earlier discussed issues of unstated criteria? [ATTAINX]: Yes, Your Honor.”); Tr. at 267:23–268:3 (“THE COURT: . . . [Y]our earlier arguments related to CBP’s alleged application of unstated criteria largely overlapped with these; is that correct? [CMCI]: Yes, Your Honor.”); Tr. at 269:14–18 (“THE COURT: Moving on to AOI, . . . your earlier arguments related to CBP’s alleged application of unstated criteria largely overlap with the arguments related to inadequate documentation; is that correct? [AOI]: Yes, Your Honor.”). The Court will therefore not re-address those arguments.

C. Challenges Related to CBP’s Alleged Disparate Treatment of Offerors

Plaintiffs AttainX, Unissant, and AOI bring disparate treatment claims regarding CBP’s evaluation of their RREPs. AttainX MJAR at 19; Unissant MJAR at 24–25; AOI MJAR at 15– 16. The government responds, “[plaintiffs] cannot prevail on a claim of disparate treatment . . . because [they] cannot show that CBP ‘unreasonably downgraded [their] proposal[s] for deficiencies that were “substantively indistinguishable” or nearly identical from those contained in other proposals.’” Gov’t’s Group 1 Reply at 12, 27–28 (quoting Office Design Grp. v. United States, 951 F.3d 1366, 1373 (Fed. Cir. 2020)).

D. Challenges Related to Prejudice

All plaintiffs argue CBP’s award decision has prejudiced them. Attainx MJAR at 23; Ekagra MJAR at 19; CMCI MJAR at 10; Unissant MJAR at 29; AOI MJAR at 28; Constellation MJAR at 2. The government responds plaintiffs have “‘not [met] the prejudice requirement [as their] allegations, even if proven, would not change the results of the procurement process.’” Gov’t’s Cross-MJAR and Resp to Group 1 at 22 (quoting Seaborn Health Care, Inc. v. United States, 101 Fed. Cl. 42, 48 (2011) (citation omitted)).

E. Challenges Related to Injunction

All plaintiffs argue the Court should enter permanent injunction. Attainx MJAR at 26; Ekagra MJAR at 21; CMCI MJAR at 10; Unissant MJAR at 31–32; AOI MJAR at 29; Constellation MJAR at 6. The government responds none of the plaintiffs have demonstrated entitlement to injunctive relief. Gov’t’s Cross-MJAR and Resp to Group 1 at 71.

A. Bid Protest Jurisdiction and APA Review

The Tucker Act, as amended by the Administrative Dispute Resolution Act (ADRA), provides this court jurisdiction over “action[s] by an interested party objecting to a solicitation by a Federal agency for bids or proposals for a proposed contract or to a proposed award or the award of a contract or any alleged violation of statute or regulation in connection with a procurement or a proposed procurement.” Administrative Dispute Resolution Act, Pub. L. No. 104-320, 28 U.S.C. § 1491(b)(1) (1996). The term “interested party” means “actual or prospective bidders or offerors whose direct economic interest would be affected by the award of the contract or by failure to award the contract.” Am. Fed’n of Gov’t Employees v. United States, 258 F.3d 1294, 1302 (Fed. Cir. 2001). “[T]o prevail in a protest the protester must show not only a significant error in the procurement process, but also that the error prejudiced it.” Data Gen. Corp. v. Johnson, 78 F.3d 1556, 1562 (Fed. Cir. 1996).

The Court evaluates bid protests under the framework laid out in Section 706 of the Administrative Procedure Act. 28 U.S.C. § 1491(b)(4); see also Impresa Construzioni Geom. Domenico Garufi v. United States, 238 F.3d 1324, 1332 (Fed. Cir. 2001); M ATTHEW H.

S OLOMSON , C OURT OF F EDERAL C LAIMS : J URISDICTION , P RACTICE, AND P ROCEDURE § 8-37 (2016) (footnote omitted) (first citing 28 U.S.C. § 1491(b)(4); and then quoting 5 U.S.C. § 706(A), (D)); 28 U.S.C. § 1491(b)(4) (2024) (“In any action under this subsection, the courts shall review the agency’s decision pursuant to the standards set forth in section 706 of title 5.”)). “[T]he proper standard to be applied in bid protest cases is provided by 5 U.S.C. § 706(2)(A): a reviewing court shall set aside the agency action if it is ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’” Banknote Corp. of Am., Inc. v. United States, 365 F.3d 1345, 1350–51 (Fed. Cir. 2004) (quoting Adv. Data Concepts, Inc. v. United States, 216 F.3d 1054, 1057–58 (Fed. Cir. 2000)). An agency’s action is arbitrary, capricious, or an abuse of discretion if “the agency ‘entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or . . . is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.’” Ala. Aircraft Indus., Inc.-Birmingham v. United States, 586 F.3d 1372, 1375 (Fed. Cir. 2009) (quoting Motor Vehicle Mfrs. Ass’n of U.S., Inc. v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)). Since the arbitrary and capricious standard is “highly deferential,” a reviewing court must “sustain an agency action evincing rational reasoning and consideration of relevant factors.” Adv. Data Concepts, 216 F.3d at 1058 (citing Bowman Transp., Inc. v. Ark.-Best Freight Sys., Inc., 419 U.S. 281, 285 (1974)). A court, therefore, will not “substitute its judgment” for the agency’s so long as the agency’s decision was reasonable. Dell Fed. Sys., L.P. v. United States, 906 F.3d 982, 998–99 (Fed. Cir. 2018) (quoting R & W Flammann GmbH v. United States, 339 F.3d 1320, 1322 (Fed. Cir. 2003)).

B. Judgment on the Administrative Record in a Bid Protest

“[Rule] 52.1(c) [of the Rules of the United States Court of Federal Claims (‘RCFC’)] provides for judgment on the administrative record.” Huntsville Times Co. v. United States, 98 Fed. Cl. 100, 104 (2011); see also Bannum, Inc. v. United States, 404 F.3d 1346, 1353–54 (Fed. Cir. 2005). The court may set aside an agency action if plaintiff has proven “either: (1) the procurement official’s decision lacked a rational basis; or (2) the procurement procedure involved a violation of regulation or procedure.” Garufi, 238 F.3d at 1332. The rational basis test requires the court to ask “whether the contracting agency provided a coherent and reasonable explanation of its exercise of discretion.” Dell Fed. Sys., L.P., 906 F.3d at 992 (quoting Banknote Corp. of Am., Inc., 365 F.3d at 1351). “When a challenge is brought on the second ground, the disappointed bidder must show ‘a clear and prejudicial violation of applicable statutes or regulations.’” Garufi, 238 F.3d at 1333 (quoting Kentron Haw., Ltd. v. Warner, 480 F.2d 1166, 1169 (D.C. Cir. 1973)). “[D]e minimis errors do not require the overturning of an award.” Grumman Data Sys. Corp. v. Dalton, 88 F.3d 990, 1000 (Fed. Cir. 1996). A bid protest plaintiff must establish alleged “errors in the procurement process significantly prejudiced” plaintiff by showing “there was a ‘substantial chance’ it would have received the contract award but for the errors.” Bannum, 404 F.3d at 1353 (quoting Info. Tech. & Applications Corp. v. United States, 316 F.3d 1312, 1319 (Fed. Cir. 2003) aff’d, 316 F.3d 1312 (Fed. Cir. 2003)).

C. Permanent Injunction

When deciding whether a permanent injunction is warranted, a court considers: (1) whether . . . the plaintiff has succeeded on the merits of the case; (2) whether the plaintiff will suffer irreparable harm if the court withholds injunctive relief; (3) whether the balance of hardships to the respective parties favors the grant of injunctive relief; and (4) whether it is in the public interest to grant injunctive relief. PGBA, LLC v. United States, 389 F.3d 1219, 1228–29 (Fed. Cir. 2004).

The Court begins its analysis with Track 1—Professional Services. As discussed, supra Section II, the Track 1 parties raise four arguments related to: (1) subcontractor size; (2) RREP value; (3) the ineligibility of awardee Catalina Solutions; and (4) GTS’ labor categories.² The Court discusses each in turn below.

A. Whether CBP’s Use of NAICS Code 541511 to Determine Subcontractor Size

was Arbitrary and Capricious

Track 1 plaintiffs LSI, Arch, and Constellation argue CBP wrongfully found subcontractors used by each plaintiff were “other than small under [North American Industry Classification System (NAICS) code] 541511 and [therefore] made a[n improper] point adjustment.” LSI Reply at 2; see Arch MJAR at 24–25; see Constellation MJAR at 3. Under the solicitation, bidders were entitled to “preferential scores favoring lower percentages of large business utilization.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 61 n.10; see also AR at 686 (TAB 18.2) (Self-Scoring and Document Verification Worksheet). Specifically, CBP “allow[ed] a mix of subcontractor sizes,” Gov’t’s Cross-MJAR and Resp. to Group 2 at 49–50, but “reserved preferential scoring for small businesses under NAICS code 541511.” Gov’t’s Group 2 Reply at 7. LSI, Arch, and Constellation all assert CBP’s “point adjustment[s] [related to their subcontractors] should be reversed,” see LSI MJAR at 4, because the agency improperly “determined that . . . [their] subcontractors . . . [were] not [] small business[es], resulting in losing points on the small business participation score.” Constellation MJAR at 3; see also Arch MJAR at 1, 20–25 (“CBP improperly deducted 375 points based on Arch Systems’ use of a subcontractor that DBP deemed (incorrectly and without authority) to be a large business.”). At oral argument, plaintiffs clarified their primary issue is the FAR makes clear “it’s the size standard associated with the NAICS codes [that matters.] Not the NAICS code per se.” Tr. at 381:6–8 (Constellation). Thus, plaintiffs argue CBP should not have decreased their scores because, although their subcontractors were not self-certified under NAICS code 541511, they were beneath the relevant size threshold. See, e.g., Tr. at 357:14–358:4 (plaintiff spokesperson Unissant). In response, the government contends “CBP did not act arbitrarily or capriciously when it disallowed small business preferential points” for plaintiffs’ subcontractors. Gov’t’s Group 2 Reply at 7 (sentence case omitted). Multiple defendant intervenors, although making distinct legal arguments, echo the government’s sentiment the Court should not disturb CBP’s determinations related to plaintiffs’ subcontractors. See Catalina Cross-MJAR and Resp. at 2–3 (“One offeror (Arch Systems) included RREPs from a large business subcontractor in its proposal and the other (Catalina Solutions) did not.”); Chevo Cross-MJAR and Resp. at 5–6 (“[T]he Group 2 protesters challenge the agency’s decisions related to subcontractor size . . . [the] Court does not have jurisdiction to adjudicate these claims, so they must be dismissed.”). LSI specifically alleges “LSI proposed [XXX XXXXXXXXXX XXXXXXXX XXXXXXXXXXX, XXX (‘XXX’)] as one of its small business subcontractors” and “provided accurate information on the small business and socioeconomic size classification of [XXX].” LSI MJAR at 3 (citing AR at 6810, 6812 (TAB 39a) (LSI Offer)). According to LSI, however, CBP deducted points because it “found that [XXX] was not a small [business under] NAICS 541511 and . . . does not hold a [General Services Administration Multiple Award Schedule (GSA MAS)] contract” even though, per LSI, “[XXX] was small under NAICS 541511 and did not have to hold a GSA MAS contract.” Id. at 4, 5 (citing AR at 6810 (TAB 39a) (LSI Offer)) (“Regarding a requirement to hold a valid GSA MAS contract, the RFQ states that . . . prime contractor[s are] required to have a GSA Contract, [but] the RFQ contains no requirement that subcontractors such as [XXX] have a GSA Contract.”). Concerning NAICS 541511, LSI clarifies “the Agency relied upon the SBA’s Dynamic Small Business Search (‘DSBS’) to verify the size of [XXX],” which “showed that [XXX] was small” under “NAICS [c]ode[s] 541512 . . . and 541513,” codes LSI contends are similar to 541511. Id. at 5; see also LSI Reply at 3 (“The Agency further ignored [XXXXX] Sam.gov entry, which shows that [XXX] is a Woman Owned Small Business . . . under 541512 . . . and 541413.”). LSI ultimately takes issue with CBP’s actions related to [XXX] because “[t]he RFQ . . . d[id] not indicate . . . the Agency would make size determinations of subcontractors as part of the evaluation.” LSI MJAR at 6; see also LSI Reply at 4 (“By incorporating FAR 52.219-8(e)(1) into the RFQ, the Agency gave offerors the impression that they could rely on subcontractors’ written representations of size.”).

2Plaintiff GTS raises challenges to both Track 1 and Track 2. See GTS MJAR. Given the overlapping issues, the Court will address GTS’ arguments related to Track 1 and 2 both in Section V.

Specifically, LSI contends FAR 52.219-8(e)(1), which was incorporated into the RFQ, indicates a small business concern is “a concern . . . that is independently owned and operated, not dominant in the field of operation in which it is bidding on [g]overnment contracts, and qualified as a small business under the size standards in th[e] RFQ.” LSI Reply at 4 (quoting 48 C.F.R. § 52.219-8(e)(1)). [XXX], according to LSI, “met this test.” Id.

Plaintiff Arch makes similar allegations as LSI concerning its subcontractor [XXXXXXXXXX]. See Arch MJAR at 20. Specifically, Arch alleges “CBP . . . reduced Arch Systems’ score based on its assessment that . . . [XXXXXXXXXX][] is a large business” even though “[t]he RFQ ma[de] no mention of any purported requirements or restrictions related to an offeror’s subcontractor’s size.”³ Id. at 20–21 (citing AR at 689 (TAB 18.2) (Solicitation)). Arch alleges it is “improper for CBP to conduct a size determination” as such actions are exclusively within the authority of the SBA. Id. at 22. Further, like LSI, Arch alleges its subcontractor “is not a large business” under the NAICS codes it planned to assign to [XXXXXXXXXX], which are separate from “the NAICS code assigned . . . by the government” to Arch’s general contract (i.e., NAICS code 541511). Id. at 23–24 (citing 13 C.F.R. § 121.410); see also Arch Reply at 10 (alleging CBP’s “size determination was factually incorrect[] because Arch Systems’ subcontractor . . . is small under the NAICS code for the subcontract Arch Systems would issue to it”). Arch likewise argues the RFQ’s incorporation of FAR 52.219-8, which permits contractors to “rely on the representations of their subcontractors” contradicts “that CBP would be validating subcontractor size.” Arch Reply at 11–12 (citing AR at 631 (TAB 18.2) (Solicitation); 48 C.F.R. § 52.219-8(e)(1)). Both LSI and Arch clarify they are not requesting the Court make a size determination, rather they are “arguing that CBP lacked the authority to make a determination” of their subcontractor’s size. Id. at 13; see also LSI Reply at 5.

Plaintiff Constellation makes similar arguments, stating “CBP was wrong . . . for several reasons” when it “determined that one of Constellation’s subcontractors (Trandes Corporation) was not a small business, resulting in [Constellation] losing points on the small business participation score.” Constellation MJAR at 3. Specifically, per Constellation, “small business size status was a matter of self-certification,” meaning CBP was “not authorized or empowered to make that determination.” Id. (emphasis omitted). Counsel for Constellation argued at oral argument NAICS code 541511 in the RFQ is relevant only to the extent “the [FAR] regulation[s] . . . say[] that the offeror has to . . . self-certify that it is a small business under the size standard associated with the procurement[’s NAICS code. It does not matter] what the . . . description of the NAICS code says” or whether the (sub)contractor is actually self-certified thereunder. See Tr. at 381:10–16.

3Arch, as with its RREP challenge, infra, alleges even if CBP is correct [XXXXXXXXXX] is not a small business for purposes of this Solicitation, Arch is entitled to some points—rather than zero—as “the RFQ does not contemplate that an offeror subcontracting to a large business will receive zero points for the large business’ work.” Arch MJAR at 29.

The government asserts the three plaintiffs’ arguments are a “repetition” of one another and “fail[] for the same reasons.” Gov’t’s Group 2 Reply at 21. Specifically, the government asserts “CBP did not act arbitrarily when it reserved preferential scoring for small businesses under NAICS code 541511, which is the code assigned to this solicitation.” Id. at 7 (citing AR at 476 (TAB 14.2) (Solicitation Cover Page)). According to the government, NAICS code 541511, which “was limited to businesses with annual receipts under $34 million” at “the time . . . of the RFQ” is different than similar NAICS codes, meaning CBP’s “decision to validate ‘small business’ status against SBA databases showing NAICS code profiles and GSA MAS contract holders was reasonable.” Id.; Gov’t’s Cross-MJAR and Resp. to Group 2 at 59 (“CBP verified . . . claim[s regarding] small business preferential scoring in two ways. First, it looked to see if the subcontractor . . . holds a GSA MAS contract . . . [then CBP] attempted to verify [the] business size via DSBS.”); see Tr. at 365:21–366:12. The government argues plaintiffs were not entitled to preferential scoring as neither they nor their subcontractors represented the subcontractors as “small business[es] with NAICS [c]ode 541511.” Gov’t’s Group 2 Reply at 8, 18 (“[B]ecause the only size applicable at the solicitation stage was NAICS code 541511, the relevant question for CBP was whether the entity performing the work in the RREP was small under NAICS code 541511, not whether it might be considered small under a different NAICS code. . . .”). The government further argues it was not arbitrary and capricious to “decline[] to accept other NAICS codes for performed work or future work for proposed subcontractors” because CBP is without authority to conduct size determinations, such as by comparing a subcontractors’ alleged size to the specifications of NAICS code 541511.4 Gov’t’s Group 2 Reply at 8, 18. Indeed, in response to LSI’s claim CBP’s “determination that TES was not small violated the SBA’s rules” because the SBA “is the only entity that could make such a finding,” LSI Reply at 5, the government clarifies CBP merely “verified whether a business [identified as small by a bidder] had been designated by the SBA as small”—CBP did not itself make size determinations.

4The government also notes, even if it wanted to independently judge subcontractor size, it lacked the “information for comparison” to make the “difficult[] . . . size comparisons across NAICS codes” plaintiffs request. Gov’t’s Group 2 Reply at 8, 19. As the government points out, it would need the “business receipts” of each subcontractor self-certified under codes other than 541511 to perform this comparison because, for example, “NAICS code 541513 has a dollar threshold of $37.0 million, which is above the $34 million size standard for small businesses under NAICS code 541511.” Id. Thus, self-certification under a code similar to 541511 does not necessarily mean a business meets the requirements of 541511. See also Tr. at 370:4–13 (“[GOVERNMENT:] [T]he difference in [541]519 [compared to 541511 is] it has a 34 million [dollar] cutoff, but it also has an exception of 150 employees. . . . THE COURT: [541]519 is not apples-to-apples with [541]511? [GOVERNMENT:] Correct.”); Tr. at 365:22– 24 (“[GOVERNMENT:] [541]511 is custom computer programming services. Whereas, [541]512 is custom systems design services.”).

Gov’t’s Group 2 Reply at 19. Finally, to the extent plaintiff Arch argues “CBP’s acceptance of RREPs performed by [XXXXXXXXXX] means CBP also should have accepted [XXXXXXXXXX] as a small business,” the government states, “the two analyses [(i.e., RREP and small business)] are separate,” meaning “nothing in the RFQ compels” CBP to accept [XXXXXXXXXX] as a small business. Id. at 9.

1. Whether CBP Improperly Made a Size Determination As noted, supra, plaintiffs LSI, Arch, and Constellation allege CBP “improper[ly] . . .

conduct[ed] a size determination,” an act solely within the SBA’s jurisdiction, when deeming their subcontractors other-than-small businesses for the purposes of this solicitation. Arch MJAR at 22; see LSI MJAR at 6; see Constellation MJAR at 3. In response, the government claims CBP did not make size determinations, rather CBP “verified whether a business [identified as small by a bidder] had been designated by the SBA as small.” Gov’t’s Group 2 Reply at 19. At oral argument, the government reiterated this point—distinguishing between conducting a size determination and verifying a self-certification against the SBA’s designations—and explicitly agreed “size determinations are exclusively the jurisdiction of the SBA.” Tr. at 379:14–380:1.

When conducting a size determination, the SBA “counts the receipts, employees, or other measure[s] of size of the [entity] whose size is at issue and all of its domestic and foreign affiliates.” See 13 C.F.R. § 121.103(a)(6). Despite arguing CBP wrongfully “conducted a size determination” by verifying the subcontractor size status listed in their offers, plaintiffs do not allege CBP undertook any calculation approximating the SBA’s procedures as described. See Arch MJAR at 22; LSI MJAR at 6; Constellation MJAR at 3. Indeed, plaintiffs could not make this argument as CBP merely looked to various government databases, including the SBA’s Dynamic Small Business Search, to verify the size information (i.e., the NAICS codes) included in offers. See Gov’t’s Cross-MJAR and Resp. to Group 2 at 59. CBP, as agencies often do, verified the size information provided by offerors as the agency said it would in the RFQ, which noted CBP would review quotes for “compliance [and] accuracy,” AR at 689 (TAB 18.2) (RFQ Amendment 8). See, e.g., Veterans Electric, LLC, B-413198, 2016 CPD P 231, at *3 (Comp. Gen. Aug. 26, 2016) (noting the contracting officer “verified . . . several of the NAICS codes the [offeror] . . . ha[d] listed”); SDR Architects, B-412498, 2016 CPD P 63, at *1 (Comp. Gen. Feb. 25, 2016) (“After proposals were submitted the contract specialist queried the System for Award Management (SAM) in order to verify [the offeror’s] NAICS code.”). Thus, while it is true only the SBA may conduct size determinations, see 4 C.F.R. § 21.5(b)(1), CBP did not encroach upon the SBA’s jurisdiction.⁵ See also Small Business Act, 15 U.S.C. § 637(b)(7) (2023). Rather, CBP took advantage of the SBA’s expertise to ensure its award decision reflected the small business set aside designated on the cover of the RFQ. See AR at 476 (TAB 14.2) (Solicitation Cover Page). This standard-practice verification of offeror-provided information is therefore neither arbitrary nor capricious. Banknote, 365 F.3d at 1350–51 (“[T]he proper standard to be applied in bid protest cases is provided by 5 U.S.C. § 706(2)(A): a reviewing court shall set aside the agency action if it is ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’”). While plaintiffs dispute the propriety of CBP requiring subcontractors be self-certified under NAICS code 541511 in order to earn preferential scoring, see infra, plaintiffs do not allege CBP’s method for verifying NAICS self-certification was arbitrary and capricious. See Arch MJAR; LSI MJAR; Constellation MJAR. The Court therefore next turns to plaintiffs’ allegation CBP incorrectly awarded preferential scoring only to those “companies that have self-certified that they can perform under the 511 NAICS code.” Tr. at 368:2–4 (the government); Tr. at 383:5–384:10 (“[CONSTELLATION:] It’s not mandatory that a contractor register themselves as complying with a particular NAICS code in an SBA database.”).

5Chevo filed a motion for partial dismissal for lack of subject matter jurisdiction related to Arch’s and LSI’s arguments about CBP’s alleged size determinations. Chevo argues Arch and LSI “ask the Court to rule [its] . . . subcontractors are small businesses . . . [but this Court cannot make size determinations given] SBA has exclusive jurisdiction to determine a business’s size.” Chevo Mot. to Dismiss at 6. Chevo reasons “Arch Systems’ and LSI’s requests for the Court to do so must be dismissed.” Id. Here, the Court is not making size determinations of Arch and LSI; as such, and because the Court is granting Chevo’s cross-MJAR, the Court finds as moot Chevo’s motion for partial dismissal for lack of subject matter jurisdiction.

2. Whether CBP’s Requirement Subcontractors Self-Certify under NAICS Code 541511 to Receive Preferential Scoring was Arbitrary and Capricious

The RFQ cover sheet states, “this acquisition is . . . set aside . . . 100% for . . . small businesses [under] NAICS[] 541511.” AR at 476 (TAB 14.2) (Solicitation Cover Page). The parties therefore do not dispute the cover sheet of the RFQ set this solicitation aside under NAICS code 541511, Tr. at 381:21–24 (“THE COURT: And you agree that the solicitation notes acquisition is set-aside [for] small business[es] under a specific NAICS code [i.e., NAICS code 541511][?] [CONSTELLATION:] [Y]es.”); Tr. at 364:11–12 (“[GOVERNMENT:] [NAICS code 541511] is the size standard that the [g]overnment has chosen for the offerors.”), which covers “custom computer programming services.” See U.S. S MALL B US . ADMIN., TABLE OF S MALL B USINESS S IZE S TANDARDS (Mar. 17, 2023) (https://www.sba.gov/sites/sbagov/files/2023-03/Table%20of%20Size%20Standards_Effective%20March%2017%2C%202023%20%281%29 %20%281%29_0.pdf). Rather, they disagree as to what a set-aside under 541511 means.

Specifically, plaintiffs argue, “because the solicitation listed [541]511, the solicitation is not being set aside for companies that have self-certified . . . under the [541]511 NAICS code. The solicitation was set aside for small businesses [of any kind] under [the] $34 million [size threshold established by that code].” Tr. at 368:1–12. Plaintiffs concede, however, as “the solicitation has the . . . NAICS code number” on its cover, “in hindsight . . . [it] would have been a great idea” to self-certify before submitting an offer. Tr. at 367:14–19 (plaintiff spokesperson Unissant). The government responds, plaintiffs’ argument regarding the meaning of the NAICS code is “unpersuasive where the solicitation is explicitly targeted to small businesses [self-certified] under NAICS code 541511.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 49; see also Tr. at 365:22–24 (“[GOVERNMENT:] [541]511 is custom computer programming services[, which is different from similar codes such as] [541]512[, which] is custom systems design services.”).

At issue here is the plain meaning of the RFQ cover sheet. See AR at 476 (TAB 14.2) (Solicitation Cover Page) (“[T]his acquisition is . . . set aside . . . 100% for . . . small businesses [under] NAICS[] 541511.”). “As in all . . . construction cases, we begin with the language” in issue. Barnhart v. Sigmon Coal Co., 534 U.S. 438, 450 (2002); ANTONIN S CALIA & B RYAN A.

GARNER , R EADING LAW : THE INTERPRETATION OF LEGAL TEXTS 56 (2012) (“The words of a governing text are of paramount concern, and what they convey, in their context, is what the text means.”). This text is unambiguous. See Momenta Pharm., Inc. v. Amphastar Pharm., Inc., 686 F.3d 1348, 1354 (Fed. Cir. 2012) (quoting Robinson v. Shell Oil Co., 519 U.S. 337, 340 (1997) (“The [first step in interpreting [text] is to determine whether the language at issue has a plain and unambiguous meaning with regard to the particular dispute in the case.’”)). To the extent CBP decided to award preferential scoring to small businesses, the cover sheet sets forth the type of small business eligible for preferred treatment—those self-certified under NAICS code 541511. See AR at 476 (TAB 14.2) (Solicitation Cover Page). Although plaintiffs urge the Court to read 541511 broadly to include companies self-certified under similar NAICS codes, see Tr. at 353:15–354:2, the Court is in no position to “add[] to what the text states.” S CALIA & GARNER , supra, at 93. The RFQ plainly sets the solicitation, including preferential scoring, aside for small businesses self-certified under NAICS code 541511. Barnhart, 534 at 450; Momenta, 686 F.3d at 1354 (quoting Robinson, 519 U.S. at 340) (“‘Our inquiry must cease if the . . . language is unambiguous.’”); S CALIA & GARNER , supra, at 56.

Turning next to applicable regulations, plaintiffs argue FAR 19 suggests contractors need not self-certify under NAICS code 541511 as “[t]he import of the NAICS code is [solely] the size standard that it assigns to the contract.” See Tr. at 373:4–5. Outside the size standard, the code itself, according to plaintiffs, see, e.g., Tr. at 371:8–25, is of no importance. See Tr. at 368:6–9 (“[Given the set-aside under 541511,] [t]he solicitation was set aside for small businesses under $34 million. A jet engine repair company that has 30 million in revenue could have submitted a proposal. They would have been evaluated badly, . . . but they were eligible[.]”). FAR 19.102(b)(1) explains, in a procurement, “contracting officers shall assign one NAICS code and corresponding size standard to all solicitations . . . .” FAR 19.102(b)(1). This provision continues, “[t]he contracting officer shall determine the appropriate NAICS code by classifying the product or service being acquired in the [] industry that best describes the principle purpose of the supply or service being acquired.” Id. FAR 19.301-1(a)(1)(i)(A) further provides, “[t]o be eligible for award as a small business concern . . . an offeror is required to represent in good faith . . . it meets the small business size standard corresponding to the [NAICS] code identified in the solicitation.” FAR 19.301-1(a)(1)(i)(A).

Although these FAR provisions discuss the assignment and applicability of NAICS codes to a solicitation, they fail to support plaintiffs’ position NAICS codes merely represent size standards. See FAR 19.102(b)(1); FAR 19.301-1(a)(1)(i)(A). Indeed, while FAR 19.102(b)(2) requires agencies to “assign one NAICS code and corresponding size standard” to each solicitation, there is no indication the size standard is the only important aspect of the code. FAR 19.102(b)(2) (emphasis added). To the contrary, the size standard is called out as separate and distinct from the NAICS code itself, suggesting each carries independent significance. See id.⁶ Further, the requirement the CO assign a NAICS code according to the “industry that best describes the principle purpose” of the contract suggests the selected industry—not just the size standard associated with the NAICS code—is relevant to a small business set aside. Compare FAR 19.102(b)(2) with Tr. at 361:20–22 (“[CONSTELLATON:] The only import in a set-aside solicitation like this of that NAICS code is the associated size standard.”) and Tr. at 368:6–7 (“[CONSTELLATION:] [Even though NAICS code 541511 was selected by CBP here,] [a] jet engine repair company that has 30 million in revenue could have submitted a proposal.”). There is also no suggestion in these provisions, which merely require a NAICS code and size standard be selected based upon the contract size and relevant industry, an agency cannot require offerors and their subcontractors self-certify under the selected NAICS code to be eligible for preferential scoring. This aligns with the government’s contention even similar NAICS codes can be materially different, meaning agencies are free to use self-certification under the relevant NAICS code as an eligibility requirement. See Gov’t’s Group 2 Reply at 7; Tr. at 369:21–370:14 (“[GOVERNMENT:] I want[] to point out the difference in [541]519 [relative to 541511.] It has a 34 million cutoff, but it also has an exception of 150 employees. . . . THE COURT: [541]519 is not apples-to-apples with [541]511? [GOVERNMENT:] Correct.”); Tr. at 365:22– 23 (“[GOVERNMENT:] [541]511 is custom computer programming services[, which is different from similar codes such as] [541]512[, which] is custom systems design services.”). Further, despite plaintiffs’ contention NAICS codes are relevant only for the size standard they set, there is a significant string of pre-award bid protests in which offerors challenge the industry of the NAICS codes selected by agencies for solicitations. See, e.g., Consolidated Safety Svcs., Inc. v. United States, 167 Fed Cl. 543, 546 (2023) (“This pre-award bid protest action centers on whether the government reasonably assigned the [NAICS] code that best fits the solicitation at issue.”); see also InGenesis, Inc. v. United States, 104 Fed. Cl. 43, 52 (2012); Arcata Assocs., Inc. v. United States, 110 Fed. Cl. 290, 302 (2013). These cases highlight the significance of NAICS codes apart from the size standards they set, including because they convey industry information. Id. at 555 (quoting Brown v. Plata, 563 U.S. 493, 562 (2011) (Scalia, J., dissenting)) (explaining an “agency cannot justify its selection of a particular NAICS code on the grounds that it fits the procurement in some general sense or that the selection is ‘good enough for government work.’”).

6NAICS code 541511, for instance, covers only businesses engaged in “custom computer programming services” with average annual receipts equal to or below $34 million. See supra, U.S. SMALL BUS. ADMIN., T ABLE OF SMALL BUSINESS S IZE STANDARDS (Mar. 17, 2023) (sentence case omitted). Similar NAICS codes, such as 541512, have a size standard equal to or below $34 million but cover other industries (e.g., “computer systems design services” for 541512). Id. (sentence case omitted).

Additionally, while FAR 19.301-1(a) limits eligibility “for award as a small business concern” to those contractors able to “represent in good faith . . . [they] meet[] the small business size standard corresponding to the [NAICS] code identified in the solicitation,” this provision does not suggest agencies are unable to impose further restrictions on award eligibility, including a requirement subcontractors be self-certified under the relevant NAICS code to receive preferential scoring. FAR 19.301-1(a). Indeed, FAR 8.405-2(c), under which this solicitation was issued, contemplates procuring agencies will specify the “evaluation criteria” in RFQs. FAR 8.405-2(c); see also FAR 8.405-2(d) (“The ordering activity shall evaluate all responses received using the evaluation criteria provided to the schedule contractors.”); see AR at 478 (TAB 14.2) (Solicitation). There is therefore no indication FAR 19-301 prevents agencies from requiring offerors and their subcontractors to be self-certified under a particular relevant code. See FAR 19.301-1(a). The provision merely imposes a threshold eligibility requirement on offerors. See id.

Despite their insistence NAICS code 541511 is of no consequence beyond “the size standard . . . it assigns to the contract,” Tr. at 373:4–5, plaintiffs are also unable to point the Court to a case supporting the proposition agencies cannot limit awards or preferential scoring to (sub)contractors self-certified under particular NAICS codes. See Tr. at 376:8–378:18. While plaintiffs briefly discussed Veterans Electric at oral argument, the case does not support plaintiffs’ contentions. Tr. at 438:6–22; Veterans Electric, LLC, B-413198, 2016 CPD P 231, at *3 (Comp. Gen. Aug. 26, 2016).⁷ In Veterans Electric, the solicitation was “set aside for . . . small businesses under [NAICS] code 238210.” Id. at *1. The awardee, however, “was not registered . . . under the relevant [] code.” Id. at *2. Recognizing at the outset “[t]he evaluation of an offeror’s proposal is a matter within the agency’s discretion,” the GAO determined there is “no basis to conclude . . . [the awardee] was required to certify that it was small under the specific NAICS code relevant to the solicitation.” Id. at *3; see also SDR Architects, B-412498, 2016 CPD P 63, at *1 (Comp. Gen. Feb. 25, 2016) (“[W]e are aware of no statutory or regulatory requirement that [an offeror] have the particular NAICS code identified in the solicitation as its primary code.”); S4, Inc., B-299817, at *8 (Comp. Gen. Aug. 23, 2007) (concluding the plaintiff could not “point to any requirement in statute or regulation that mandates listing . . . the precise NAICS code applicable to a procurement” in an offer); High Plains Computing, Inc., B-409736.2, at *6–*7 (Comp. Gen. Dec. 22, 2014) (stating “there [i]s no apparent statutory or regulatory requirement for the NAICS code in a solicitation to be listed in an offeror’s” offer). While at first blush this and related cases appear to support plaintiffs’ contention an agency cannot reduce offerors’ scores because subcontractors are not certified under “the specific NAICS code relevant to the solicitation,” the GAO’s reasoning cuts against plaintiffs’ suggestion. Veterans Electric, LLC, B-413198 at *4. Veterans Electric stands for the proposition it is within an agency’s discretion to accept (or not) offers from contractors not certified under the relevant NAICS code. GAO’s recognition of: (1) agency discretion when evaluating offers; and (2) the lack of authority permitting or prohibiting agencies from requiring certification under particular NAICS codes supports the government’s position here. See id. While the GAO permitted the agency in Veterans Electric to award contracts to an offeror who had not self-certified under the NAICS code relevant to the solicitation, plaintiffs have not proffered even one case supporting the opposite proposition, namely that agencies cannot require offerors or subcontractors be self-certified under the relevant code. Id. Rather, as recognized by the Federal Circuit and reiterated by the GAO in Veterans Electric, agencies are “entitled to exercise discretion upon a broad range of issues confronting them in the procurement process,” Impresa Construzioni Geom. Domenico Garufi v. United States, 238 F.3d 1324, 1332 (Fed. Cir. 2001) (citation and internal quotations omitted); see also Veterans Electric, LLC, B-413198 at *2, including in deciding whether a NAICS code listed in an RFQ represents more than a size standard.

Thus, neither caselaw nor applicable regulations provide grounds on which to find CBP’s decision to award preferential scoring only to those offerors whose subcontractors self-certified under NAICS code 541511 arbitrary and capricious. Banknote, 365 F.3d at 1350–51 (“[T]he proper standard to be applied in bid protest cases is provided by 5 U.S.C. § 706(2)(A): a reviewing court shall set aside the agency action if it is ‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’”). All caselaw and regulatory language proffered by plaintiffs, including the language of FAR 19, suggests agencies have discretion to decide whether NAICS codes have greater significance than “the size standard . . . [they] assign[,]” Tr. at 373:4–5. See, e.g., Veterans Electric, LLC, B-413198, at *3; Consolidated Safety Svcs., 167 Fed. Cl. at 546. CBP therefore “evinc[ed] rational reasoning” in limiting preferential scoring to those subcontractors self-certified under NAICS 541511, Adv. Data Concepts, 216 F.3d at 1058 (citing Bowman Transp., 419 U.S. at 285), the code for which the solicitation is explicitly set aside. AR at 476 (TAB 14.2) (Solicitation Cover Page). Holding otherwise would require the Court to ignore the plain language of the solicitation, which sets the solicitation aside for businesses certified under NAICS code 541511, see supra, and require CBP to engage in the difficult task of determining which NAICS codes are sufficiently comparable to 541511 to warrant award, a line drawing problem which would likely lead to additional litigation.

7GAO decisions are not binding on this court.

3. Whether Blue & Gold Prohibits Plaintiffs from Challenging CPB’s Reliance on NAICS Code 541511

As noted supra, FAR 19.102(b)(1) requires “contracting officers [to] assign one NAICS code and corresponding size standard to all solicitations.” FAR 19.102(b)(1). In this case, CBP assigned one NAICS code—541511—to this solicitation. AR at 476 (Tab 14.2) (Solicitation Cover Page). It did not, however, explicitly assign a size standard, leaving the box labeled “SIZE STANDARD” on the RFQ cover sheet blank. See id. Although plaintiffs do not explicitly allege this omission led to ambiguity, at oral argument plaintiffs suggested the solicitation was unclear regarding whether CBP intended to “validate that [offerors] me[]t the size standard . . . by looking at [] NAICS code[s].” Tr. at 391:21–392:10. In other words, plaintiffs appear to argue the inclusion of a NAICS code, which incorporates a size standard, in tandem with the omitted size standard left offerors guessing as to whether self-certification under the NAICS code would be used to verify size. See id. The Court therefore addresses whether plaintiffs can raise challenges related to RFQ clarity arising from the omission of a clear size standard at this stage in the solicitation process.

In Blue & Gold, the Federal Circuit held:

[A] party who has the opportunity to object to the terms of a government solicitation containing a patent error and fails to do so prior to the close of the bidding process waives its ability to raise the same objection subsequently in a bid protest action in the Court of Federal Claims.

Blue & Gold, Fleet, L.P. v. United States, 492 F.3d 1308, 1313 (Fed. Cir. 2007). In M.R.

Pittman, the Federal Circuit applied Blue & Gold in a situation akin to this case. M.R. Pittman Grp., LLC v. United States, 68 F.4th 1275 (Fed. Cir. 2023). There, the SAM.gov webpage linking to the solicitation at issue stated, “this is a 100% Small Business Set Aside procurement . . . under NAICS code: 811310.” Id. at 1278. “The solicitation itself[, however,] did not include a reference to NAICS Code 811310,” although it did “incorporate by reference [FAR] 52.219-6,” which “warns that ‘[o]ffers are solicited only from small businesses concerns.’” Id. (quoting FAR 52.219-6). Plaintiff M.R. Pittman was the lowest bidder for the contract, but the agency determined the plaintiff “did not qualify as a small business under NAICS [c]ode 811310 and was thus ineligible for the award.” Id. This court held the plaintiff could not bring its challenge “pursuant to the Blue & Gold waiver rule” as it had “failed to object [to the omission of the NAICS code in the solicitation] prior to the close of the bidding process.” Id. at 1277 (citing M.R. Pittman Grp., 154 Fed. Cl. at 244 (citing Blue & Gold, 492 F.3d 1308)). The Federal Circuit agreed, holding “the missing NAICS code—when considered with the solicitation’s webpage or the FAR provision incorporated in the solicitation—constituted a patent omission from the solicitation.” Id. at 1283. Thus, according to the Federal Circuit, “M.R. Pittman waived its protest grounds under the Blue & Gold waiver rule.” Id. at 1284.

In this case, plaintiffs appear to suggest the solicitation—which includes a NAICS code but not a size standard—did not put offerors on notice CBP would verify size by checking whether subcontractors were self-certified under NAICS code 541511 (rather than simply below the associated size threshold) before awarding preferential scoring. Tr. at 387:7–12 (“[CONSTELLATION:] Your honor, I’m not taking issue with the verification. THE COURT: So you agree that CBP does have allowance to verify? [CONSTELLATION:] Yes, your honor. I’m just taking issue with how they verified.”); Tr. at 391:21–392:10 (“THE COURT: What about if [the RFQ] said CBP will verify [whether] all offerors seeking points for small business [subcontracting] fall under the 511 NAICS code? [CONSTELLATION:] If the [g]overment said we are going to validate that you meet the size standard . . . [a]nd if they said we are going to validate it by looking at the NAICS code, and a company didn’t have the NAICS code, then the validation process would be reasonable.”). To the extent plaintiffs rely on the absence of a size standard in the RFQ to make this argument, however, M.R. Pittman suggests this is a “patent omission from the solicitation,” meaning plaintiffs have “waived [their] protest grounds under the Blue & Gold waiver rule.” 68 F.4th at 1283–84. The presence of the NAICS code and absence of a size standard on the solicitation cover sheet, AR at 476 (Tab 14.2) (Solicitation Cover Page), “created an ambiguity that a reasonably diligent contractor would seek to clarify.” M.R. Pittman, 68 F.4th at 1284. Plaintiffs cannot now object the solicitation was unclear as to whether and how the NAICS code would be used to verify size; plaintiffs had “the opportunity to object to the terms” of the solicitation “prior to the close of the bidding process” and chose not to do so. Blue & Gold, 492 F.3d at 1313. Plaintiffs have therefore waived their “protest grounds under the Blue & Gold waiver rule” as—to the extent the absence of an explicit size standard made unclear whether the NAICS code stood merely for a size standard or something more—this was a “patent omission” plaintiffs should have sought to clarify earlier in the solicitation process. M.R. Pittman, 68 F.4th at 1283–84.⁸

8To the extent plaintiffs allege CBP acted improperly because it required subcontractors to have GSA MAS contracts, see LSI MJAR at 4, 5 (citing AR at 6810), and ignored the solicitation’s incorporation of FAR 52.219-8(e), id. at 6 (quoting 48 C.F.R. § 52.219-8(e)(1)), the Court confirmed the government acted properly at oral argument. First, the government confirmed subcontractors were “not required to have GSA MAS contracts,” rather CBP used the existence of GSA MAS contracts (if any) as one method by which to verify whether a subcontractor was self-certified under NAICS code 541511. Tr. at 390:9–13. Thus, to the extent plaintiffs argue CBP wrongfully required subcontractors have GSA MAS contracts, there is no evidence this is the case. See id. Further, in briefing plaintiffs appeared to argue FAR 52.219-8(e), which permits contractors to “rely on the [size] representations of their subcontractors,” Arch Reply at 11–12 (citing AR at 631 (TAB 18.2) (Solicitation); 48 C.F.R. § 52.219-8(e)(1)), immunized offerors from point adjustments stemming from the agency’s verification of subcontractor size representations. At oral argument, however, plaintiffs admitted the incorporation of FAR 52.219-8(e) “does not require the [a]gency to accept [subcontractor’s] written [size] representations without verification.” Tr. at 386:25–

4. Whether Arch Systems Experienced Disparate Treatment Regarding Subcontractor Size

In an argument closely related to its allegations regarding subcontractor size, plaintiff Arch also contends “CBP engaged in disparate treatment during its evaluation of offerors by reducing Arch’s score due to its subcontracting to a purported large business while not deducting points from other offerors for the same reason.” Arch MJAR at 25. Arch argues disparate treatment occurs “[w]here an agency ‘unreasonably downgrade[s] [an offeror’s] proposal for deficiencies that were “substantially indistinguishable” or nearly identical [to] those contained in other proposals’ or where ‘the agency inconsistently applied objective solicitation requirements’ between offerors.” Id. (quoting Office Design Grp. v. United States, 951 F.3d 1366, 1372 (Fed. Cir. 2020) (additional citations omitted)). Arch alleges this occurred here, because CBP gave awardee Chevo “10 points for the use of a large business subcontractor,” id. at 26 (citing AR at 9159 (TAB 54.2) (Chevo Self-Score Worksheet)), while “Arch Systems received zero points for its use of a (purportedly) large business subcontractor on the same task.” Id. (citing AR at 11924–25 (TAB 154.a) (Arch Self-Score Worksheet)).

In response, the government contends “Arch Systems’ arguments demonstrate its fundamental misunderstanding of the evaluation criteria.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 53–54 (“Because CBP treated all three quoters in strict accordance with the terms of the RFQ as they applied to their distinct proposals, the Court should dismiss the claim.”). Namely, the government clarifies Chevo “awarded [itself] the appropriate score in the first instance, [so] there was no basis for CBP to deduct points.” Id. at 54–55. On the other hand, “Arch Systems . . . submitted RREPs from a large business but had claimed points as though it was a small business,” warranting a point adjustment. Id. at 54. The government notes, per the RFQ, CBP was permitted to either leave self-scores alone or decrease them to zero, see Gov’t’s Cross-MJAR and Resp. to Group 2 at 53; see Tr. at 394:16–395:13. Thus, the government reduced to zero all of Arch’s incorrectly scored boxes on its self-score work sheet. Tr. at 394:16–395:13. Defendant-intervenor Chevo agrees with the government and notes, “unlike Arch Systems, Chevo did not claim points for small business usage when it was instead relying on a large business.” Chevo Cross-MJAR and Resp. at 9.

To succeed on a disparate treatment claim, a “protestor must show that the agency unreasonably downgraded its proposal for deficiencies that were ‘substantively indistinguishable’ or nearly identical from those contained in other proposals” or “that the agency inconsistently applied objective solicitation requirements between it and other offerors, such as proposal page limits . . . or submission deadlines.” Office Design Grp., 951 F.3d at 1372 (quoting Enhanced Veterans Sols, Inc. v. United States, 131 Fed. Cl. 565, 588 (2017) (additional citations omitted)). Here, Arch systems alleges this test is met as awardee Chevo received “10 points for the use of a large business subcontractor” while “Arch Systems received zero points for its use of a (purportedly) large business subcontractor on the same task.” Arch MJAR at 26. Although Chevo and Arch received different point totals for similarly situated other-than-small 387:9. These arguments therefore have no impact on the Court’s consideration of plaintiffs’ subcontractor size arguments.

subcontractors, the government contends, Gov’t’s Cross-MJAR and Resp. to Group 2 at 54–55, there was no disparate treatment as CBP neither “unreasonably downgraded” Arch’s proposal for “deficiencies . . . substantively indistinguishable” from those in Chevo’s solicitation nor “inconsistently applied objective solicitation requirements.” Office Design Grp., 951 F.3d at 1372.

At the outset, Arch conceded at oral argument Chevo’s solicitation was accurate, Tr. at 395:22–24 (“THE COURT: Do you agree that Chevo awarded itself the appropriate score? [ARCH SYSTEMS:] I do.”), meaning Arch and Chevo could not have “substantively indistinguishable” deficiencies. Office Design Grp., 951 F.3d at 1372. Indeed, Chevo had no deficiency at all. See AR at 9159 (TAB 54.2) (Chevo Self-Score Worksheet). Further, to the extent Arch argues CBP “inconsistently applied objective solicitation requirements” by downgrading Arch’s incorrect score to zero while permitting Chevo to receive 10 points for its large subcontractor, the solicitation language at issue is “[q]uoters’ self-scores can either remain the same or decrease.” See AR at 617 (Tab 17.2) (RFQ Evaluation Methodology). In applying this scoring criteria to Chevo, CBP found “Chevo . . . appropriately calculated and claimed 10 points [for a large business subcontractor],” Gov’t’s Cross-MJAR and Resp. to Group 2 at 54, meaning its score should “remain the same.” AR at 617 (Tab 17.2) (RFQ Evaluation Methodology). On the other hand, Arch “claimed a small business classification” for its subcontractor, which CBP deemed large, thus requiring CBP to decrease Arch’s self-score.

Gov’t’s Cross-MJAR and Resp. to Group 2 at 55. CBP therefore did not disparately apply the requirement scores can either remain the same or decrease, rather it evenhandedly applied this mandate from the RFQ to two differently situated offerors. Office Design Grp., 951 F.3d at 1372.

Finally, to the extent Arch takes issue with CBP’s decision to reduce Arch’s improper self-score (375) to zero rather than adjusting it to the amount Arch would have been entitled to for its use of a large subcontractor had Arch properly self-scored (185), see Arch MJAR at 27, the government contends it did not have “authority to write a [new score] in” the box in which Arch incorrectly scored itself. Tr. at 395:1–13. Rather, CBP—per the government—could “decrease [self-scores] by the points that were not eligible [e.g., points for small business participation] but could not grant [points] in other categor[ies] [e.g., points for large business participation].” Id.⁹ Thus, although Chevo received points for the use of a large subcontractor while Arch did not, this is because the offerors were “not similarly situated.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 55. As Arch admitted at oral argument, Chevo properly assessed itself points for using a large subcontractor, prompting CBP to leave its score alone. See Tr. at 396:22–24. Arch, on the other hand, awarded itself points for the use of a small subcontractor when, in CBP’s view, the subcontractor was large. Gov’t’s Cross-MJAR and Resp. to Group 2 at 54. Thus, CBP adjusted Arch’s score to the extent possible given its limited authority to do so and removed Arch’s points in the small business category (e.g., decreased Arch’s small business subcontractor score to zero). See AR at 617 (Tab 17.2) (RFQ Evaluation Methodology). CBP could not, however, add points elsewhere (e.g., the large subcontractor category), meaning Arch’s subcontracting score for the contractor at issue fell to zero. See id. This is an objective and even application of the RFQ’s requirements, meaning no disparate treatment between Arch and Chevo occurred. See Office Design Grp., 951 F.3d at 1372.

9A hypothetical is useful in understanding the RFQ language in issue. Imagine an offeror who, believing its subcontractor is a small business, awards itself 100 points in the small business subcontractor category and zero points in the large business subcontractor category on its self-score worksheet. Upon agency review, however, the agency determines the offeror’s subcontractor is large, not small. The agency, bound by the RFQ requirement “[q]uoters’ self-scores can either remain the same or decrease,” AR at 617 (Tab 17.2) (RFQ Evaluation Methodology), reduces the offeror’s small business subcontractor self-score to zero. The agency cannot, however, add points to the offeror’s large business subcontractor self-score as it only has authority to leave scores “the same or decrease” them. Id. The offeror’s score thus remains zero, even if—had the offeror properly scored itself—it could have received some amount of large business subcontractor points.

B. Whether CBP’s Verification of RREP Value using the Federal Procurement

Data System was Arbitrary and Capricious

Track 1 plaintiffs¹⁰ LSI and Arch also challenge CBP’s deduction of points related to RREPs. LSI MJAR at 7; Arch MJAR at 14. “The minimum value for an RREP under the RFQ was one million dollars.” Gov’t’s Cross-MJAR and Group 2 Resp. at 61 (citing AR at 687) (TAB 18.2) (RFQ Amend. 8). “In its evaluation, CBP . . . sought to verify that the projects submitted as RREPs were at least of a value of one million dollars. . . .” Id. RREPs below the $1,000,000 threshold were disregarded by CBP, and those with values greater than $10,000,000 were disfavored. See Gov’t’s Group 2 Reply at 5 (citing AR at 223–25 (TAB 10b¹) (Self-Score Worksheet)) (explaining “the scoring rubric . . . favored projects with lower ‘total’ values,” meaning RREPs of a value between one and ten million dollars were entitled to the most points). LSI alleges CBP “found that [its] . . . Task 1 RREP 5 (AR [at] 6882), Task 4 RREP 5 (AR [at] 6924), and Task 7 RREP 3 (AR [at] 6961) are the same RREP, . . . [each with a] contract value [of] $758,000, less than the required $1M minimum,” so CBP “removed these RREPs.” LSI MJAR at 7 (citing AR at 10139–40 (TAB 73.10) (LSI Track 1 Memorandum); AR at 9702–03 (Tab 63) (LSI Self-score Worksheet)). According to LSI, “[t]hese RREPs should not have been removed and the deducted points should be restored” because the database used by CBP to validate the value of these RREPs—the Federal Procurement Data System (FPDS)— “does not reflect a full picture,” and LSI could “have provided an explanatory note describing the errors in FPDS” had CBP asked for such. Id. at 7–8. Further, LSI alleges its Task 1 RREP 6, a Department of Commerce (DOC) contract, “should not have been removed.” Id. at 8.

Specifically, LSI argues CBP did not list FPDS validation “as a requirement in the RF[Q],” so LSI did not provide the contract number for Task 1 RREP 6 in the form required to identify a DOC contract on FPDS. Id. (“As is often the case for DOC contracts, FPDS users must prepend the letters ‘DOC’ to the beginning of the contract number before searching FPDS . . .”). LSI clarifies it “has not alleged that it was unaware of the $1M minimum or that the Agency was going to verify that the projects submitted were at least $1M,” LSI Reply at 7, rather LSI argues all four removed RREPs should be restored as their value verifiably exceeds $1,000,000, and LSI could have provided additional information required to prove such to CBP had the RFQ identified FPDS as CBP’s verification system of choice. See LSI MJAR at 7–8.

10To the extent Constellation argues in its MJAR “CBP could not ‘locate’ several of the Constellation team’s RREPs in FPDS [because its] . . . commercial project[s] [are by their nature not] in FPDS,” Constellation MJAR at 4, this argument fails because Constellation’s experiences are in fact non-commercial, government projects, namely its work for the Naval Information Warfare Systems. See Gov’t’s Cross-MJAR and Resp. to Group 2 at 39–40 (citing AR at 3831–39) (TAB 31b2) (Constellation Offer). Constellation does not refute this in its Reply. See Constellation Reply.

Arch likewise asserts “CBP’s deduction of points . . . based on the size of one of its RREP references was improper.” Arch MJAR at 14. Arch notes “[t]he RFQ instructed quoters to submit at least two, but not more than seven, RREPs per task similar in scope and complexity as identified in the . . . scope of work.” Id. at 15 (citing AR at 686 (TAB 18.2) (RFQ Amend. 8)). According to Arch, each “RREP was to contain, in pertinent part, the associated ‘Requirement or Project Total Value (project total value of work performed if performed as a sub-contractor).’” Id. (quoting AR at 687 (TAB 18.2) (RFQ Amend. 8) (emphasis omitted)).

Arch accordingly “disclosed the value of the Requirement in question, a separate contract line item . . . as $[XXXXXXXXXXXX],” even though the “entire task order was valued at $[XXXXXXXXXXXXXX]” Id. at 16 (emphasis added). In other words, Arch interpreted the RFQ as permitting “offerors . . . to rely on specific contract requirements, as opposed to entire task order values.” Id. at 17. Arch argues “the Agency’s Q&A” and “RFQ[] Amendment 008” “reinforce [its] reasonable interpretation that a quoter could submit an RREP . . . reflective of a specific requirement’s value, as opposed to project total value.” Id. at 18; see also Arch Reply at 5–6, 8 (arguing the government’s proffered Q&A “offer[] clarity” as to the question of whether offerors could “submit RREPs based on ‘Requirement or Project Total Value’ and pointing to Q&A “explicitly recogniz[ing] that a bidder could submit an RREP . . . reflective of a specific requirement’s value”) (emphasis omitted). Should the Court find CBP’s evaluation of “Arch Systems under Task 5 was . . . consistent with the RFQ,” however, Arch argues there is a “latent ambiguity in the RFQ,” id. at 19, meaning an ambiguity “‘not apparent on the face of the solicitation [or] . . . ascertainable through reasonable or customary care,’” warranting injunctive relief. Id. (quoting Input/Output Tech., Inc. v. United States, 44 Fed. Cl. 65, 72 n.10 (1999)). In response to LSI, the government argues “CBP reasonably removed RREPs and their associated points where it could not verify the projects’ claimed values.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 61 (sentence case omitted). Specifically, the government argues “[i]t was perfectly reasonable and obvious . . . for CBP to use FPDS to validate federal contract values” as it is “widely-recognized as ‘a comprehensive, web-based tool for agencies to report contract actions’ . . . and is the authoritative source for procurement data, including dollars obligated in contract actions.” Id. at 62 (citing FAR § 4.602(a); then citing Contract Data, GEN. S ERV. ADMIN., sam.gov/content/contract-data (last accessed Jan. 1, 2024)). The government asserts “CBP . . . acted reasonably and in conformity with the RFQ when it excluded” LSI’s Task Concerning the DOC RREP, the government notes “LSI was on notice that its quote would be subject to validation and should have known that FPDS is the authoritative site for federal contract confirmation,” meaning LSI “should have provided additional information or instructions to CBP” as “LSI was aware of a disconnect between FPDS’s search function and the contract information that it provided.” Id. at 63. According to the government, “CBP acted reasonably in conducting its verification [by] searching the contract number provided by LSI[] and acting on the results.” Id. Further, the government notes “any error . . . [on the government’s behalf related to the DOC contract] would be harmless” because, “[e]ven with the additional points” from this contract as correctly scored, “LSI would remain well below the cut-off mark for awardees.” Id.

1RREP 5, Task 4 RREP 5, and Task 7 RREP 3 because “FPDS showed a contract value of only $758,000.” Id. (citing AR at 10582 (TAB 88j) (CBP Explanation of Award Decision)).

Turning to Arch, the government contends Arch’s “extraction of a contract portion was in conflict with the RFQ, which directed quoters to score their RREPs based on the ‘Total Value’ of the ‘Requirement or Project.’” Gov’t’s Cross-MJAR and Resp. to Group 2 at 46 (quoting AR at 687 (TAB 18.2) (RFQ Amend. 8)). The government also notes “the Q&A process” further “clarified . . . quoters could not break a requirement or project into subparts in order to reach the higher point category.” Id. (citing AR at 722 (TAB 18c¹) (Track 1 Q&A)). Specifically, the government asserts “CBP’s . . . responses” to Q&A “explicitly forbid the project splitting Arch Systems attempted,” meaning CBP “acted rationally and in accordance with the RFQ (as clarified in the Q&A) when it deducted . . . 75 points claimed by Arch Systems.” Id. at 48 (citing AR at 722 (TAB 18c¹) (Track 1 Q&A)) (noting: (1) CBP will not accept particular option years of a contract as their own task order / experience project; and (2) CBP will not consider evaluating an RREP “just over $500k” on a “$35M multi-sub-task contract”). Indeed, the government contends “its scoring rubric—intended to distinguish smaller contracts/subcontracts from medium and large contracts/subcontracts—would . . . be[] meaningless . . . if large multimillion dollar contracts simply could be split . . . to qualify for preferential scoring.” Gov’t’s Group 2 Reply at 4 (citing NVT Techs., Inc. v. United States, 370 F.3d 1153, 1159 (Fed. Cir. 2004)). Further, to the extent Arch argues “CBP at least should have awarded it the 50 points eligible for projects in the $10–20 million range,” the government notes the RFQ permitted “CBP . . . [to] verify scores but . . . []not award points not previously claimed by the quoters.” Id. at 6; see Arch MJAR at 27–30 (“[E]ven if the Agency was theoretically correct in downgrading Arch Systems’ quoted based on the value of the Task 5 RREP 5 task order . . . Arch Systems’ quote still should have received [a higher score than CBP calculated.]”); see also supra Section V.A.4. Finally, addressing Arch’s claim of a latent ambiguity, the government notes there is no “ambiguity at all,” Gov’t’s Cross-MJAR and Resp. to Group 2 at 48, because Arch’s “belief was unreasonable” given the “robust Q&A regarding the valuation of a ‘requirement or project.’” Id. at 48–49. The government notes, in the absence of such an ambiguity, Arch’s claim is “tantamount to a challenge to the terms of the solicitation, which Arch Systems has forfeited by failing to raise prior to submitting its quote.” Id. at 48 (citing Blue & Gold Fleet, 492 F.3d at 131 (Fed. Cir. 2007). 1. Whether CBP Acted Arbitrarily and Capriciously When it Verified LSI’s RREPs on FPDS Using the Information LSI Provided LSI’s claim regarding validation of RREP value resembles Track 1 plaintiffs’ arguments related to verification of subcontractor size, see supra Section V.A. Specifically, LSI’s core argument is CBP did not put offerors on notice CBP planned to use FPDS to validate RREP values, so LSI did not provide all information potentially necessary for CBP to adequately search for and verify its RREPs. Tr. at 404:14–22 (“THE COURT: So [CBP] should have found [your RREP contracts], and they made a wrong conclusion? [LSI:] They couldn’t find the contract. If they would have found the contract, it would have showed it was above a million dollars. Because they apparently didn’t understand [our RREP submissions] and since we didn’t know FPDS was going to be used, if we would have, we would have put something in our proposal that said for this contract, you have to put DOC in front to get a [Department of] Commerce contract.”). In particular, for Task 1 RREP 6, which LSI performed for the Department of Commerce (DOC), LSI alleges because it “did not know the Agency was going to use FPDS validation, LSI provided accurate contract information based on the manner the DOC defines its [contract] numbers . . . [but did not tell CBP] FPDS users must prepend the letters ‘DOC’ to the beginning of the contract number before searching FPDS.” LSI Reply at 6. Thus, CBP was unable to find LSI’s Task 1 RREP 6 on FPDS. See LSI MJAR at 7–8. As a result, according to LSI, CBP improperly “removed these RREPs” when it was unable to validate their values. See id. at 7 (citing AR at 10139–40 (TAB 73.10) (LSI Track 1 Memorandum); AR at 9702–03 (TAB 63) (LSI Self-score Worksheet)). In response, the government notes throughout the RFQ and Q&A “the [g]overnment put [p]laintiffs on notice that their claims would be validated.” Tr. at 406:8–25. Specifically, the government states “[f]irst, the RFQ talks about requirements[,] . . . [i]t discusses that all of the information is going to be verified and validated . . . . And then the[] Q&As . . . talk[] exactly about the dollar amount. So everybody was on notice that the 1 million range was a key number and that that was going to be validated . . . . [And] FPDS . . . is the primary authoritative database for searching for this type of information.” Id. Thus, the government argues “CBP . . . acted reasonably and in conformity with the RFQ when it excluded” LSI’s RREPs for which it could not validate a value above the $1,000,000 threshold. Gov’t’s Cross-MJAR and Resp. to Group 2 at 62. An agency decision should be set aside if it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Banknote Corp. of Am., 365 F.3d at 1350– 51. This standard permits a “reviewing court [to] set aside a procurement action if: (1) the procurement official’s decision lacked a rational basis; or (2) the procurement procedure involved a violation of regulation or procedure.” Centech Group, Inc. v. United States, 554 F.3d 1029, 1037 (Fed. Cir. 2009) (citing Garufi, 238 F.3d at 1332). Here, CBP used FPDS to “validate the values of the identified [f]ederal contracts” in LSI’s RREPs using “the contract identification information provided by LSI.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 59, 63. The RFQ, although not explicitly calling out FPDS, see Tr. at 406:8–25 (“[GOVERNMENT:] [The solicitation] does not refer specifically to FPDS . . .”), warned offerors “the [g]overnment will compile a ranking of quotes by reviewing the quotations for compliance, accuracy, and completion of all required areas.” AR at 689 (TAB 18.2) (RFQ Amend. 8) (emphasis added). LSI concedes this in its briefing, noting “LSI has not alleged that it was unaware . . . the Agency was going to verify that projects submitted were at least $1M.” LSI Reply at 7. Offerors, including LSI, were therefore on notice CBP would validate the information provided. See id. Further, FPDS is a “‘comprehensive, web-based tool for agencies to report contract actions’” and remains “the authoritative source for procurement data.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 62 (first quoting FAR § 4.602(a); and then citing Contract Data, GEN. S ERV. ADMIN ., https://sam.gov/content/contract-data (last accessed Jan. 30, 2024)); see also Bluewater Mgmt Grp. LLC v. United States, 150 Fed. Cl. 588, 617 (2020) (emphasis added) (noting the agency at issue “reviewed the information provided [by the awardee and] the information in FPDS” in evaluating offers); Dynetics, Inc. v. United States, No. 18-481C, 2018 WL 2439295 at *2 (Fed. Cl. 2018) (permitting an agency to verify relevant experience via multiple methods, including FPDS); Evolver Inc.; Armed Forces Svcs. Corp., B-413559.2 at *4 (Comp. Gen. Dec. 21, 2016) (noting it was “reasonably related to the agency’s needs” to award points only to firms capable of providing “FPDS report[s]” with certain information related to relevant experience). Indeed, the FAR requires “[f]ederal agencies . . . to report [contracts] to FPDS.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 62 (citing FAR § 4.606). Thus, to the extent LSI complains CBP “should have told [offerors] they were using FPDS,” Tr. at 408:19– 20, CBP expressly stated in the RFQ it would “review [quotes] . . . for compliance, accuracy, and completion of all required areas,” AR at 689 (TAB 18.2) (RFQ Amend. 8), and did so by using “the authoritative source for procurement data.” Contract Data, GEN. S ERV. ADMIN., https://sam.gov/content/contract-data (last accessed Jan. 30, 2024)). Conducting this review using “the contract . . . information provided by LSI,” Gov’t’s Cross-MJAR and Resp. to Group 2 at 63, is neither arbitrary nor capricious as agencies “are ‘entitled to exercise discretion upon a broad range of issues confronting them’ in the procurement process,” Garufi, 238 F.3d at 1332, and LSI has failed to shoulder the “heavy burden” of “showing . . . the award decision ‘had no rational basis.’” Id. Although there may have been several other possible routes of verification, CBP is able to “exercise discretion,” id., in the evaluation process, including in validating information provided by offerors. See Bluewater Mgmt, 150 Fed. Cl. at 617 (emphasis added) (noting the agency at issue “reviewed the information provided [by the awardee and] the information in FPDS” in evaluating offers); Banknote Corp. of Am. v. United States, 365 F.3d 1345, 1350–51 (Fed. Cir. 2004) (explaining agency action should be set aside if “‘arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.’”) (citing Adv. Data Concepts, 216 F.3d at 1057–58).

2. Whether CBP Acted Arbitrarily and Capriciously When it Removed Arch’s RREP

Plaintiff Arch’s primary argument is “CBP’s deduction of points . . . based on the size of one of its RREP references was improper,” Arch MJAR at 14, because “a quoter could submit an RREP . . . reflective of a specific requirement’s value, as opposed to project total value.” Id. at 18. The government counters, the RFQ and Q&A made clear offerors could not “take what is a large contract and break it up into discrete subtasks in order to get into a specific dollar range,” Tr. at 416:2–24, so Arch’s attempt at “carv[ing] out a single contract line item from a much larger contract so it could award itself the maximum number of points under the scoring rubric” was improper. Gov’t’s Cross-MJAR and Resp. to Group 2 at 46.

The parties agree the language in issue is as follows: “[f]or each RREP, offerors’ quotes were to include . . . the ‘Requirement or Project Total Value (project total value of work performed if performed as a subcontractor.)’” Arch MJAR at 5 (quoting AR at 687 (TAB 18.2) (Solicitation)) (italics emphasis added; bolding omitted). According to Arch, this RFQ language permits offerors to submit either “requirement or project total value,” with the parenthetical merely placing “additional requirements . . . on subcontractors[,] who . . . need to include . . . the project total value” in their RREP. Tr. at 410: 7–25. On the other hand, the government contends the parenthetical means “if [an offeror] performed [work] as a subcontractor,” it should “provide the value of subcontracted work in lieu of the requirement total value,” and “if you were the prime [contractor], you would put in the . . . total value.” Tr. at 411:7–11. The government further clarifies “the Q&A [on this point] was very clear.” Tr. at 411:12–19.

Indeed, when asked during Q&A:

Could an RREP’s Project Value just account for the level of effort associated with completing the relevant work to that RREP’s task on a larger contract? For instance, on a $35M multi-sub-task contract that contains a discreet [artificial intelligence/machine learning] sub-task/requirement just over $500k, would [CBP] consider evaluating an RREP that reflects scope and complexity?[,] CBP replied “No.” See AR at 722 (TAB 18c¹) (Track 1 Q&A). In other words, when explicitly

asked whether CBP would consider a discreet subtask of relevant work from a larger contract, CBP said “no.” Id. The government further made clear during Q&A it would not consider parts of contracts for RREP value purposes, declining to consider “each option year of a contract” on its own. See id. To the extent there was any ambiguity in the RFQ’s requirement offerors submit RREPs stating the “[r]equirement or [p]roject [t]otal [v]alue (project total value of work performed if a subcontractor),” AR at 687 (TAB 18.2) (RFQ Amend. 8), the Court agrees CBP’s Q&A responses make clear offerors were not permitted to submit a subtask from a larger contract. See AR at 722 (TAB 18c¹) (Track 1 Q&A). Indeed, were offerors able to do so, the government is correct “its scoring rubric—intended to distinguish smaller contracts/subcontracts from medium and large contracts/subcontracts . . . would . . . be[] meaningless.” Gov’t’s Group 2 Reply at 4; Tr. at 416:2–24 (“[GOVERNMENT:] The question [in the Q&A] is . . . ‘how do I get into that sweet spot where I get [the full] 75 points?’, which we believe is exactly what Arch Systems was looking at. We believe this is clear, along with the language in the RFQ to say, no, we are looking at project total value or requirement total value. . . . The blanket no [in the Q&A to the question of whether you can use a portion of a contract as an RREP] means that you can’t take what is a large contract and break it up into discrete subtasks in order to get into a specific dollar range.”). CBP’s preference for contractors with experience satisfactorily performing contracts between one and ten million dollars, see AR at 223–25 (TAB 10b¹) (Self-Score Worksheet), would be moot were offerors permitted to break large contracts into discrete subtasks. CBP was therefore not arbitrary or capricious when it deducted points from Arch for using a portion of a larger contract in contravention of the RFQ as clarified by the Q&A.11 Banknote Corp., 365 F.3d at 1350–51 (explaining agency action should be set aside if “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”).

C. Whether Arch Systems Has Standing to Challenge CBP’s Award to Catalina Solutions

Plaintiff Arch alleges CBP’s award to Catalina was improper because “Catalina should have been deemed ineligible.” Arch MJAR at 30. Arch points out the RFQ required “[a] minimum of one (1) RREP [] be submitted by the Prime Contractor in each task area.” Id.

(quoting AR at 686 (TAB 18.2) (Solicitation)) (emphasis omitted). According to Arch, the prime contractor “on Catalina’s proposal is ‘Catalina Solutions, LLC,’ ‘a joint venture between Catalina Associates and Versar National Security Solutions LLC (formerly BayFirst Solutions LLC), a subsidiary of Versar, Inc.’” Id. at 31 (citing AR at 2226 (TAB 26a) (Catalina Offer); AR at 10527 (TAB 87) (Catalina Order)). Per Arch, however, “Catalina did not provide ‘[a] minimum of one (1) RREP . . . by the Prime Contractor in each task area.’” Id. (citing AR at 2244–2338 (TAB 26b²) (Catalina Offer)). Indeed, Arch states, “[w]ith only [three] exceptions, every RREP reference provided by Catalina in its entire quote was from BayFirst.” Id. (citing AR at 2244–2338 (Tab 26b²) (Catalina Offer)). Arch purports the GAO’s decision in AttainX, in which the GAO noted “SBA regulations require the [procuring] agency to evaluate each joint venture member individually when the joint venture itself does not demonstrate it has the required experience,” shows applicable regulations require CBP to “consider[] Catalina Associates’ experience,” which Arch alleges CBP did not. Id. at 31–32, n.8 (quoting AttainX, Inc. 2023 WL 1860802 at *7 (Comp. Gen. Jan. 23, 2023)). Arch therefore argues Catalina was improperly selected for award.

11To the extent Arch contends, “even if the Agency was theoretically correct in downgrading Arch Systems’ quoted based on the value of the Task 5 RREP 5 task order . . . Arch Systems quote still should have received [a higher score than CBP calculated],” Arch MJAR at 27–30 (emphasis omitted), Arch’s argument fails as CBP could not award points to an offeror the offeror had not claimed for itself in the self-scoring worksheet. See supra Section V.A.4.

The government responds Arch misunderstands the “meaning given to th[e] term [prime contractor] in the RFQ and the guidance incorporated into it through the Q&A.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 55. Namely, according to the government, CBP made clear this requirement meant “RREP[s] c[ould] come from the small [partner], large [partner], or [joint venture].” Id. at 55–56 (quoting AR at 749 (TAB 19.2) (Solicitation Clarifications)) (“[A]s a clarification of answers provided during an earlier round of Q&A, [CBP stated] that ‘CBP will consider [RREPs for] work done and qualifications held individually by each partner to the joint venture as well as any work done by the joint venture itself previously.”); see Tr. at 421:9–22. Catalina therefore “fulfilled the RFQ’s requirement of submitting a minimum of one RREP from the prime contractor for each task area.” Id. at 56 (citing AR at 2244–56 (TAB 26a) (Catalina Offer)). Discussing AttainX, the government notes “GAO decisions are not binding . . . on this [c]ourt,” Arch “failed to raise a challenge pre-award” despite knowing CBP planned to “consider RREPs from either joint venture partner as satisfying the prime contractor requirement,” and CBP “satisfied AttainX’s interpretation of 13 C.F.R § 125.8(c)” by evaluating both Versar and Catalina Associates. Gov’t’s Group 2 Reply at 13–15. Catalina agrees the government’s procurement process followed applicable regulations and notes in its cross-MJAR, “[t]he issue is governed by 13 C.F.R. § 125.8,” which clarifies “‘a procuring activity must consider work done . . . individually by each partner to the joint venture as well as any work done by the joint venture itself previously’” but does not necessitate a review of RREPs by the joint venture as well as its individual partners. Catalina Cross-MJAR and Resp. at 4 (quoting 13 C.F.R. § 125.8) (emphasis added).

At oral argument, the Court sua sponte raised the issue of whether Arch has Article III standing to contest CBP’s award to Catalina Solutions without alleging disparate treatment. See Tr. at 419:6–13 (“THE COURT: Generally speaking . . . does Arch have standing to challenge Catalina’s award without arguing something like disparate treatment? [CATALINA:] No, your honor, I would say not in terms of [their challenge] doesn’t relate to their own evaluation, and they are not here to step into the shoes of the Agency and sort of relitigate the Agency’s decisions on a matter that doesn’t relate to their own proposal.”); Tr. at 419:16–420:2 (“[ARCH SYSTEMS:] Certainly Arch has an interest in . . . eliminat[ing] those who fall above it in terms of the rank ordering. We have 11,000 points. We believe that under the [position] most deferential to the [g]overnment[] . . . this Court will give us 235 more points. . . . Having Catalina Solutions at 11,475 moves us . . . . THE COURT: So it hurts you in the point scoring [for Catalina to not be cut from the solicitation]. [ARCH SYSTEMS:] Correct.”). The government and Catalina contend Arch does not have standing as Arch’s challenge of Catalina’s award “doesn’t relate to [Arch’s] own evaluation.” Tr. at 419:6–13. Arch disagrees and states, although there is “no guarantee that even getting inside of [the top] eight” offerors would have guaranteed an award, Tr. at 420:7–14, “Arch has an interest in . . . eliminat[ing] those who fall above it in terms of the rank ordering.” Tr. at 419:16–420:3. As standing is “an indispensable part of [a] plaintiff’s case[]” and “is a threshold jurisdictional issue,” the Court first addresses whether Arch has Article III standing to challenge CBP’s award to Catalina. See Lujan v. Defs. of Wildlife, 504 U.S. 555, 561; see also Myers Investigative & Sec. Servs., Inc. v. United States, 275 F.3d 1366, 1369 (Fed. Cir. 2002) (citing Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 102–04 (1998)), abrogated on other grounds by CACI, Inc.-Fed. v. United States, 67 F.4th 1145 (Fed. Cir. 2023).

“To establish a case or controversy, a party invoking federal jurisdiction must meet the ‘irreducible constitutional minimum of standing.’” Allgenesis Biotherapuetics Inc v. Cloudbreak Therapeutics, LLC, 85 F.4th 1377, 1380 (Fed. Cir. 2023) (quoting Lujan, 504 U.S. at 560).

“[T]o establish standing, a plaintiff must show (i) that he suffered an injury in fact that is concrete, particularized, and actual or imminent; (ii) that the injury was likely caused by the defendant; and (iii) that the injury would likely be redressed by judicial relief.” TransUnion LLC v. Ramirez, 594 U.S. 413, 423 (2021) (citing Lujan, 504 U.S. at 560–61). Most relevant here, where the government and defendant-intervenor allege Catalina’s award had no impact on Arch’s unsuccessful offer, is the injury in fact requirement. See Tr. at 419:6–13. To properly allege an injury in fact, a plaintiff “must have suffered . . . an invasion of a legally protected interest which is (a) concrete and particularized . . . and (b) ‘actual or imminent, not “conjectural” or “hypothetical.”’” Lujan, 504 U.S. at 560 (citing Allen v. Wright, 468 U.S. 737, 751 (1984); and then quoting Whitmore v. Arkansas, 495 U.S. 149, 155 (1990)) (additional citations omitted). As applied to bid protests before this court,¹² the Article III injury in fact inquiry focuses on whether plaintiff has “allege[d] facts demonstrating that the government committed prejudicial error.” Superior Waste Management LLC v. United States, 169 Fed. Cl. 239, 274 (2024); CACI, Inc-Federal, 67 F.4th at 1153 (noting “the issue of prejudice is . . . jurisdictional [when] it implicates Article III considerations”); Labatt Food Svc., Inc. v. United States, 577 F.3d 1375, 1380 (Fed. Cir. 2009) (finding no standing where there was “no showing of how the government’s error caused [plaintiff] to suffer disparate treatment or particularized harm”). The federal courts do not have a “freewheeling power to hold defendants accountable for legal infractions.” TransUnion, 594 U.S. at 427 (emphasis added) (citation omitted) (noting a plaintiff who “sues . . . not seeking to remedy any harm to herself but instead merely seeking to ensure a defendant’s ‘compliance with regulatory law’ (and, of course, to obtain some money . . .)” has not suffered an injury in fact). Rather, Article III permits the courts to “redress harms that defendants cause plaintiffs.” Id.

Here, Arch alleges CBP improperly evaluated Catalina, resulting in Catalina receiving an award when, according to Arch, “Catalina should have been deemed ineligible.” Arch MJAR at 30. Arch does not, however, clearly articulate the concrete injury it suffered from Catalina’s apparent errant award; rather it generally alleges Arch was “prejudiced by the Agency’s arbitrary and capricious evaluation of quotes.” Arch Reply at 29. Further, although Arch contends it “has an interest in . . . eliminat[ing] those who fall above it in terms of the rank ordering,” Tr. at 419:16–420:3, it concedes CBP only awarded seven out of eight possible Track 1 contracts, meaning had Catalina been eliminated during CBP’s initial evaluation process, there is “no guarantee that even getting inside of [the top] eight” would have resulted in an award. Tr. at 420:7–14. While there is no dispute Arch is an “actual bidder, a fact the administrative record also supports,” Arch’s failure to identify its specific injury from Catalina’s award renders its “allegations related to prejudice [] thin and conclusory to the extent they are made at all.” Superior Waste Management, 169 Fed. Cl. at 273 (citing American Relocation Connections, LLC v. United States, 789 Fed. Appx. 221, 227 (Fed. Cir. 2019)); Labatt Food Svc., Inc., 577 F.3d at 1380 (declining to “create a rule that, to an unsuccessful but economically interested offeror in a bid protest, any error is harmful”). Arch cannot—and indeed does not—allege “but for the government’s [alleged] error” in evaluating Catalina, Arch “had a substantial chance . . . of winning the contract[,]” because Arch could have won a contract even with CBP’s award to Catalina. Labatt Food Svc., Inc., 577 F.3d at 1380; see Tr. at 420:7–14. Had CBP deemed Arch worthy of award, the unawarded eighth contract could have gone to Arch, even without the Court “step[ping] into the shoes of the Agency and . . . relitigate[ing] the Agency’s decisions on a matter that does” not “relate to [Arch’s] . . . proposal.” Tr. at 419:10–13. Arch therefore is unable to allege prejudicial error as there is no indication, but for Catalina’s award, Arch had a substantial chance of winning the contract. Superior Waste Management, 169 Fed. Cl. at 273 (citing American Relocation Connections, 789 Fed. Appx. at 227). Thus, to the extent Arch alleges an injury from CBP’s award to Catalina (and it is not clear Arch does so), the injury is highly speculative and hypothetical in contravention of the Supreme Court’s pronouncements in Lujan and TransUnion; Arch merely alleges the existence of a “legal infraction[]” without pointing to its own concrete injury resulting therefrom.¹³ See TransUnion, 594 U.S. at 427; Lujan, 504 U.S. at 560. Arch therefore lacks Article III standing to challenge CBP’s award to Catalina.¹⁴

12For an extended discussion of standing and its relationship to the Court of Federal Claims, see Superior Waste Management. Superior Waste Management LLC v. United States, 169 Fed. Cl. 239, 252–65 (2024).

D. Whether CBP’s Elimination of GTS for Inappropriate Skill Mix Was Arbitrary and Capricious

Plaintiff GTS brings overlapping Track 1 and 2 claims. The solicitation permitted CBP to find a quote “unacceptable if the . . . labor categories” proposed by the offeror for each task “[we]re not appropriate.” AR at 688 (TAB 18.2) (Evaluation Criteria). GTS argues CBP’s “determination that GTS proposed an inappropriate skill mix [in its labor categories] . . . is arbitrary and capricious” because there “is no evidence in the record that . . . any evaluator[] meaningfully considered GTS’s proposed mix of labor categories.” GTS MJAR at 10.

13The Court notes the lack of injury in fact necessitates the absence of causation and redressability, meaning Arch cannot adequately allege any of the three prongs of Article III standing. See TransUnion, 594 U.S. at 423 (citing Lujan, 504 U.S. at 560–61).

14The Court notes on 16 February 2024, Catalina filed a motion for leave to file a supplemental brief further supporting its contention Arch is incorrect Catalina is ineligible for award. Catalina’s Mot. for Leave to File, ECF No. 165. On 19 February 2024, Arch filed a Response to Catalina’s Motion, Arch’s Resp., ECF No. 168, and on 21 February 2024, Catalina filed its Reply, ECF No. 174. In its proposed supplemental filing, ECF No. 165-1 at 2, Catalina argues the GAO’s recent decision in Maxmed “helps to clarify the AttainX decision and further illustrates why Arch Systems’s position is incorrect under . . . 13 C.F.R. § 125.8(e).” Id. (citing Maxmed Healthcare, Inc., B-421623.5, 2024 WL 359201 (Jan. 25, 2024)). Specifically, Catalina states in Maxmed the GAO “noted that the absence of specific RF[Q] language requiring a past performance submission by both joint venture members is not equivalent to a conclusion that the agency will ignore the experience of either joint venture member.” Id. As the Court does not reach the substance of Arch’s argument, the Court does not consider the GAO’s Maxmed decision and accordingly denies, see infra, Catalina’s Motion, ECF No. 165, at its discretion. See RCFC 7.2(d) (acknowledging the Court’s discretion to grant or deny requests to file with leave of court); cf. Nesselrode v. United States, 127 Fed. Cl. 421, 434 (2016) (recognizing the Court’s discretion to “grant[] or den[y] a motion for leave to amend” a complaint).

Specifically, related to Track 1, GTS alleges “[t]here is no substantive discussion in the [source selection authority (SSA)] memo[randum]” regarding CBP’s conclusion “GTS proposed an inappropriate skill mix.” Id. GTS likewise argues “there is no evidence elsewhere in the record that the Agency performed a substantive analysis of GTS’s proposed labor mix.” Id. at 16. To the extent the government relies on “the Consensus Evaluation form for GTS’s proposal” and “an email from the evaluation chair . . . to the contracting officer” as evidence of CBP’s analysis in the record for Track 1, GTS states these documents “are devoid of any substantive analysis of the appropriateness of GTS’s labor categories for the tasks to be performed.” Id. at 16–18. GTS maintains a similar argument for Track 2. See GTS Reply at 1 (“[T]he record fails to support the conclusion[] . . . that GTS proposed an inappropriate skill mix for the PS and ET tracks.”). Further, for Track 2, GTS alleges, although the SSA memorandum “states that GTS’s quote was eliminated from further consideration in [Track 2] because its skill mix was inappropriate,” id. at 13 (citing AR at 10161 (TAB 74) (SSA Memorandum)), the remainder of the “evaluation record shows that GTS was not considered for further evaluation because the TET found that GTS did not have a fair and reasonable price.” Id. at 13. This, according to GTS, signals CBP’s decision was arbitrary and capricious because “the lack of supporting documentation in the record makes it impossible to know why” GTS was eliminated from consideration for Track 2. Id. at 19.

Ultimately, according to GTS, CBP “seemed not even to understand the labor categories that GTS proposed,” GTS MJAR at 10, which were intentionally broad to “accommodate the diverse scope of work and open-ended nature of future requirements contemplated by the Solicitation.” Id. at 19, 22 (citing AR at 627 (TAB 18.2) (Solicitation)).

The government disagrees with GTS’ characterization of CBP’s review. Specifically, according to the government, CBP found “that, rather than tailoring its list of labor categories to be proposed for each task under the SOW, GTS had ‘used a standardized boiler-plate listing of [labor categories] . . . [and] applied that listing to every task.’” Gov’t’s Cross-MJAR and Resp. to Group 2 at 16, 17 (quoting AR at 9657 (TAB 61.2) (Chamblin E-mail)) (“‘[GTS’ quote provided the same entire catalog of more than [XXX] labor categories for all task areas, which did not demonstrate appropriateness for each task area.’”); Tr. at 329:16–24 (“[GOVERNMENT:] The RF[Q] did not require them to propose all of the categories [available to them.] . . . It required the opposite. You got to propose appropriate categories. That means a fit to the requirements.”). Thus, per the government, “CBP correctly excluded GTS from competition” because GTS did not follow the RFQ’s requirement “the skill mix proposed by a quoter needed to be adequate to perform the task under consideration, and not bloated with irrelevant and inapplicable positions.” Id. at 27–28 (first citing AR at 627, 688–89 (TAB 18.2) (Solicitation); then citing AR at 733 (TAB 18c¹) (Track 1 Q&A); and then citing AR at 747 (TAB 18c²) (Track 2 Q&A)) (emphasis added). Indeed, to the extent other “quoters used substantially all of their labor categories or also had overlapping sections” in their proposals, the government clarifies “other companies had fewer labor categories to draw from,” making the potential for overlap “among the tasks in the respective SOWs” more likely and more reasonable. Gov’t’s Group 2 Reply at 23 (citing AR at 2550 (TAB 26c) (Catalina Proposal)). On the other hand, “even at its lowest point . . . , GTS’s quote included scores upon scores of [inappropriate] labor categories.” Id. Further, to the extent GTS challenges the clarity of CBP’s decision making in the AR, the government acknowledges “this is a messy record,” but clarifies “the standard is not a perfect record.” Tr. at 324:23–325:3. According to the government, CBP made clear, particularly via an email correspondence between the CO and evaluators, GTS was eliminated because “they shotgunned their list of labor categories” and were “way out of step with their competitors.” Tr. at 312:15–25.

In reply, GTS notes there “were two separate evaluations”—one for Track 1 and one for Track 2. Tr. at 333:24–25. Thus, even if the email cited by the government is sufficient explanation for the track it relates to (Track 1), there is insufficient explanation of CBP’s decision to eliminate GTS from consideration for Track 2. See id.; GTS Reply at 1 (explaining the email relates to Track 1). Further, GTS argues it included a large quantity of appropriate labor categories, Tr. at 319:23–320:1, because the descriptions for each task in Tracks 1 and 2 specified “the delineated tasks are ‘not limited to’ the information listed,” meaning additional labor categories of all kinds could be relevant. GTS MJAR at 23.

An agency decision should be set aside if it is “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” Banknote, 365 F.3d at 1350. This standard permits a “reviewing court set aside a procurement action if: (1) the procurement official’s decision lacked a rational basis; or (2) the procurement procedure involved a violation of regulation or procedure.” Centech Group, 554 F.3d at 1037 (citing Garufi, 238 F.3d at 1332). Specifically, an agency decision lacks a rational basis and is therefore arbitrary and capricious if the agency “entirely failed to consider an important aspect of the problem [or] offered an explanation for its decision that runs counter to the evidence before the agency.” Ala. Aircraft Indust., Inc.- Birmingham v. United States, 586 F.3d 1372, 1375 (Fed. Cir. 2009) (quoting Motor Vehicle Mfrs. Ass’n V. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983)); Garufi, 238 F.3d at 1332 (internal citation omitted) (“[T]he test for reviewing courts is to determine whether the contracting agency provided a coherent and reasonable explanation of its exercise of discretion.”). “[T]he disappointed bidder bears a ‘heavy burden’ of showing that the award decision ‘had no rational basis.’” Garufi, 238 F.3d at 1333 (quoting Saratoga Dev. Corp. v. United States, 21 F.3d 445, 456 (D.C. Cir. 1994)).

The solicitation states:

The Government will evaluate quotations by reviewing the labor categories and rates for all tasks for all performance periods. The Government may determine that a Quoter is unacceptable if the rates are not reasonable or if the labor categories are not appropriate for the tasks.

AR at 688 (TAB 18.2) (Evaluation Criteria) (emphasis added). In proposing labor categories for each of the thirteen tasks in Track 1 and Track 2, GTS proposed [XXX] for ten, and no less than [XX] for the remaining three. See Gov’t’s Cross-MJAR and Resp. to Group 2 at 30. In contrast, all other offerors proposed “a few dozen” labor categories per task, putting GTS “far outside the standard deviation of labor categories.” Tr. at 327:4–17. Further, not only did GTS propose significantly more labor categories per task than other offerors, more importantly, CBP found GTS did not narrow its proposed categories to those “appropriate for the tasks,” AR at 688 (TAB 18.2) (Evaluation Criteria). See, e.g., AR at 9657 (TAB 61.2) (Chamblin E-mail) (noting “[m]ost of the [labor categories] proposed are inappropriate”); Tr. at 331:5–10 (“[GOVERNMENT:] CBP . . . look[ed] at th[ese proposed labor categories] and [saw they were] so dramatically out of step [with the tasks, including by listing] scores more proposed labor categories [than any other offeror.]”). As an example, the government cites Track 1 Task 5, for which GTS proposed “an [XXXXXXXXXXXX XXXXXXXXXX] [and] a [XXXXX XXXXXXXXXXX],” despite the task involving “advising and providing subject matter expertise on Acquisition Strategy.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 35 (citing AR at 9657 (TAB 61.2) (Chamblin E-mail)). Indeed, at oral argument, GTS was unable to explain how the exact same “[XXX] [labor categories were] appropriate for 10 out of the 13” distinct tasks, noting only “[t]hat’s the way it came out.” Tr. at 329:7–10; see Gov’t’s Cross-MJAR and Resp. to Group 2 at 30 (“GTS does not dispute that it quoted the very same [XXX] labor categories for each and every one of the seven tasks.”) (emphasis added). To CBP, however, GTS’ regurgitation of dozens of labor categories for 13 tasks each requiring separate skill sets signaled a disregard of the requirement to propose an “appropriate skill mix for the tasks” and instead exemplified a proposal submitted “without considering the work to be completed.” AR at 9655 (TAB 61.1) (GTS Consensus Evaluation Form). As “the minutiae of the procurement process . . . involve[s] discretionary determinations of procurement officials that a court will not second guess,” E.W. Bliss Co. v. United States, 77 F.3d 445, 449 (Fed. Cir. 1996), the Court is unable to find CBP acted irrationally, arbitrarily, or capriciously, Garufi, 238 F.3d at 1333 (quoting Saratoga Dev. Corp., 21 F.3d 445 at 456 (“[T]he disappointed bidder bears a ‘heavy burden’ of showing that the award decision ‘had no rational basis.’”)), in determining GTS’ dozens of overlapping and often identical labor categories proposed for 13 separate tasks included at least some inappropriate categories for the task requirements. See Gov’t’s Cross-MJAR and Resp. to Group 2 at 22 (citing AR at 10164 (TAB 74) (SSA Memorandum)).

Further, to the extent GTS argues there “is no evidence in the record that . . . any evaluator[] meaningfully considered GTS’s proposed mix of labor categories,” GTS MJAR at 10, this argument is unavailing for Track 1. As noted by the government, not only did CBP warn during Q&A offerors could not “just put [all of their labor categories] in” each task proposal, Tr. at 330:4–8, for Track 1, the record contains explicit evidence stating, “in reviewing the [labor categories] for Garud, they have used a standardized boiler-plate listing of [labor categories] and have applied that listing to every task area. . . . All Tasks/RREPs – Proposed [labor categories] are ‘boiler-plate’ and . . . not appropriate.” AR at 9657 (TAB 61.2) (Chamblin E-mail); Gov’t’s Cross-MJAR and Resp. to Group 2 at 31; see also Tr. at 330:9–13. Further, for Track 1, CBP reiterated this reasoning in GTS’ Consensus Evaluation Form, noting the labor categories for Track 1 are not appropriate because GTS “did not provide the appropriate skill mix for the tasks and appears to have provided a list of labor categories without considering the work to be completed.” AR at 9655 (TAB 61.1) (GTS Consensus Evaluation Form). CBP therefore clearly explained its reasoning regarding GTS’ Track 1 proposal. The Court is thus in no position to “second guess,” E.W. Bliss, 77 F.3d at 449, the agency’s decision where, as here, CBP “provided a coherent and reasonable explanation of its exercise of discretion.” Garufi, 238 F.3d at 1333. Concerning Track 2, although the government now claims the TET took issue with GTS’ “inappropriate skill mix,” GTS emphasizes “three separate times . . . the technical panel documented its findings in the record with respect to [Track 2] and stated” GTS “failed in fair and reasonable pricing.” Tr. at 335:1–8. In response, the government first notes “labor categories and pricing are intertwined” because “[q]uoters were required to propose labor categories, and the labor rates to be charged, on the pricing sheets of their quotations.” Gov’t’s Group 2 Reply at 24–25 (citation omitted). Next, the government clarifies any confusion “was resolved by the SSA Memorandum,” which explicitly stated GTS was eliminated from Track 2 for proposing an inappropriate skill mix. Id. at 24 (citing AR at 10161, 10164 (TAB 74) (Award Decision Memorandum)). In contrast to CBP’s decision for GTS’ Track 1 labor categories, for which the agency’s thought process is consistently written in multiple locations, the record for CBP’s evaluation of GTS’ Track 2 labor categories is not “coherent” and does not contain a “reasonable explanation” for the agency’s “exercise of discretion.” Garufi, 238 F.3d at 1333 (citations omitted). Rather, “[t]he evaluation record” for Track 2 “is inconsistent,” GTS Reply at 13, with portions of the record stating GTS was eliminated for lack of “fair and reasonable pricing,” see AR at 9651 (TAB 60) (Track 2 Consensus Evaluation Form); AR at 10103, 10106 (TAB 71) (TET Consolidated Review), and other record documents identifying GTS’ failure as an inappropriate skill mix. See AR at 10161 (TAB 74) (Award Decision Memorandum). This failure to explain CBP’s decision to eliminate GTS from the award process for Track 2 is arbitrary and capricious. Garufi, 238 F.3d at 1333. As such, the Court must next determine whether CBP’s failure to explain its reasoning when deeming GTS ineligible for Track 2 award was prejudicial. See Alfa Laval Separation, Inc. v. United States, 175 F.3d 1365, 1367 (Fed. Cir. 1999) (citing Statistica, Inc. v. Christopher, 102 F.3d 1577, 1581 (Fed. Cir. 1996)) (“To prevail in a bid protest, a protester must show a significant, prejudicial error in the procurement process.”).

1. Whether CBP’s Failure to Adequately Explain its Decision to Eliminate GTS from Track 2 was Prejudicial

As noted, supra, CBP’s failure to adequately explain its reason for disqualifying GTS from Track 2 was arbitrary and capricious. “To prevail in a bid protest,” however, a protestor must not only show the government acted irrationally, it must also “show a significant, prejudicial error in the procurement process.” Alfa Laval Separation, 175 F.3d at 1367 (citing Statistica, Inc., 102 F.3d at 1581). “[S]mall errors made by the procuring agency are not sufficient grounds for rejecting an entire procurement” as “[o]verturning awards on de minimis errors wastes resources and time, and is needlessly disruptive of procurement activities and governmental programs and operations.” Grumman Data Systems Corp. v. Widnall, 15 F.3d 1044, 1048 (Fed. Cir. 1994) (quoting Lockheed Missiles & Space Co. v. Bentsen, 4 F.3d 955, 960 (Fed. Cir. 1993); and then quoting Andersen Consulting Co. v. United States, 959 F.2d 929, 932 (Fed. Cir. 1992)). Here, CBP improperly conflated two related issues in evaluating GTS’ Track 2 labor categories—pricing and skill mix. See Gov’t’s Group 2 Reply at 24 (noting the relationship between labor categories and pricing); Tr. at 324:23–325:3 (“[GOVERNMENT:] We don’t dispute that this is a messy record, but the standard is not a perfect record. The standard is whether the Court can follow the rationale of the Agency.”). Ultimately, however, CBP clarified in the SSA Memorandum, AR at 10161 (TAB 74) (Award Decision Memorandum), GTS was eliminated from Track 2 for “shotgunn[ing] massive lists of potential employee categories [without] making [any] attempt to fit the proposed categories to the tasks.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 29 (citations omitted). Further, GTS was on notice of what CBP’s Track 2 decision regarding inappropriate skill mix meant in the SSA Memorandum given the explanation for Track 1, which noted GTS “did not provide the appropriate skill mix for the tasks and appears to have provided a list of labor categories without considering the work to be completed.” AR at 9655 (TAB 61.1) (GTS Consensus Evaluation Form). Thus, although GTS claims it “was prejudiced . . . because, but for these errors, GTS . . . would have been selected as an awardee,” GTS MJAR at 32, it appears CBP’s failure to clearly document its decision to eliminate GTS from Track 2 consideration was a “[s]mall error . . . [and is therefore] not sufficient grounds for rejecting an entire procurement.” Grumman Data Systems Corp., 15 F.3d at 1048 (citation omitted). GTS provides no evidence “but for these [failures to clearly document why GTS was eliminated], GTS . . . would have been selected as an awardee.” GTS MJAR at 32. Indeed, as noted in the SSA Memorandum, AR at 10161 (TAB 74) (Award Decision Memorandum), CBP found GTS proposed an inappropriate skill mix for Track 2, as with Track 1, because it listed more than [XX] overlapping labor categories, see Gov’t’s Cross-MJAR and Resp. to Group 2 at 30, per task. The Court need not “needlessly disrupt[ the] procurement activities and governmental programs and operations” of CBP, Grumman Data Systems Corp., 15 F.3d at 1048 (citation omitted), where CBP ultimately provided its reason in the record (i.e., the SSA Memorandum) and put GTS on notice regarding its reasoning through the documents related to Track 1. GTS has therefore failed to “show a significant, prejudicial error in the procurement process.” Alfa Laval Separation, 175 F.3d at 1367.

The Court will next assess Track 2—Emerging Technology. As discussed, supra Section I, this contract involves novel applications of AI. Tr. at 74:13–15 (“[GOVERNMENT]: [T]his is the type of work that the agency is looking to procure in the future as emerging technology, so novel applications of AI.”). At oral argument, the government explained the quoters’ past experience must be relatively recent for CBP to consider it “similar,” as this contract is for emerging technology. Tr. at 74:20–24 (“[GOVERNMENT]: I think just by the nature of the work that [the agency is] seeking to procure, [work experience] would have to be relatively recent. So AI systems in the 1990s are not going to cut it in terms of similarity in scope and complexity.”). As also discussed, supra Section III, the Track 2 parties raise five arguments related to: (1) challenges related to CBP’s alleged application of unstated criterion; (2) challenges related to CBP’s alleged lack of documentation and consideration of RREPs; (3) challenges related to CBP’s alleged disparate treatment of offerors; (4) challenges related to prejudice and RFQ’s language on the elimination of offerors related to dissimilar RREPs; and (5) challenges related to permanent injunction. The Court discusses each in turn below.

A. Whether CBP Applied Unstated Criterion in Its Assessment of RREPs

Many plaintiffs argue the agency used unstated, strict evaluation criteria in determining the “similarity” of plaintiffs’ RREPs with the various tasks. AR at 689 (Solicitation); see supra Section III. The Solicitation stated:

The RREPS will be evaluated for similarity in scope and complexity to the task areas identified in the ESB BPA scope of work. If the Government determines any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated. Any additional RREP submissions will not be substituted by the Government for any RREP submissions that are deemed not relevant.

AR at 689 (TAB 18.2) (Solicitation) (emphasis added). Plaintiffs assert CBP erroneously interpretated the Tasks’ use of the word “shall” as creating requirements and argue CBP needed only evaluate whether RREPs were “similar”—not whether the RREPs met each of the Tasks’ “requirements.” Tr. at 31:20–32:3 (“[PLAINTIFF SPOKESPERSON ATTAINX]: [T]he language in the evaluation methodology is very clear that it’s to be evaluated for similarity in scope, and what the agency has just said is we went down a checklist, and if you didn’t hit this particular point in this particular way in a ‘shall’ statement of work requirement, that you weren’t deemed similar in scope and complexity, and that to me reads that they were looking for identical past performance, not similar past performance.”). The government responds CBP evaluated those tasks with “shall” language more stringently and offerors should have understood “shall” as connoting required subtasks. Tr. at 30:17–22 (“THE COURT: Should offerors have read this language in conjunction with the evaluation language that you just read to infer stringent requirements for each task? [GOVERNMENT]: For the ‘shall’ tasks, yes. There were some [including Task 4] that were identified as ‘may’ that would be less stringent.”). The government also responds these protestors should have raised their challenges to the terms of the Solicitation prior to award and therefore have “forfeited the issue by failing to raise it prior to submitting . . . quote[s].” Gov’t’s Cross-MJAR and Resp. to Group 1 at 36 (citing Blue & Gold Fleet, L.P. v. United States, 492 F.3d 1308, 1313 (Fed. Cir. 2007)).

At the outset, the Court notes the government explained, and plaintiffs acknowledged, if one required RREP was not “similar in scope and complexity” to the relevant task described in the RFQ, the offeror was ineligible for award for the entire track. Tr. at 228:15–22 (“[THE COURT:] [W]hen an offeror submitted an RREP that is not similar in scope and complexity, CBP would eliminate the offeror from award. Is it from award for the entire contract or just that a specific task? . . . [THE GOVERNMENT]: From a full track. . . .”); see also Tr. at 19:8–14 (THE COURT: . . . [W]hat do you think it means, “The quotation will be ineligible for award and will not be further evaluated?” That allows the Government still some discretionary wiggle room? [PLAINTIFF SPOKESPERSON UNISSANT]: The words mean what they say. They will be ineligible for award.”). It was the duty of the offeror to submit a quote “similar in scope and complexity,” AR at 689 (TAB 18.2) (Solicitation), to be considered eligible for award. See Structural Assocs., Inc./Comfort Sys. USA (Syracuse) Joint Venture v. United States, 89 Fed. Cl. 735, 744 (Fed. Cir. 2009) (“Plaintiff’s failure to provide more detailed information [in its bid] is chargeable to it alone.”). The Court agrees with the government offerors “make tradeoffs all the time” in submitting the best proposal possible to meet the terms of the Solicitation. Tr. at 83:17– 23 (“[THE GOVERNMENT:] Offerors make these trade-offs all the time. They have to determine how close their experience is, and . . . if that’s the experience that’s closer, but it’s going to get them less points because it’s . . . over the dollar threshold for the most points, that’s just part of bidding on the solicitation.”).

Despite the plain language of the Solicitation, at oral argument Unissant called attention to the Solicitation’s use of the word “will”—“[i]f the Government determines any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated,” AR at 689 (TAB 18.2) (Solicitation) (emphasis added)—and argued “will” indicates CBP had discretion to not eliminate an offeror if the quote was not similar in scope and complexity. Tr. at 19:3–13 (“[UNISSANT]: I understand what the language says, Your Honor, and I think it gives the Government the discretion to determine, in its evaluation process, that they will not move forward with someone because they’ve found dissimilarity. . . . THE COURT: So what do you think it means, ‘The quotation will be ineligible for award and will not be further evaluated’? That allows the Government still some discretionary wiggle room? [UNISSANT]: The words mean what they say. They will be ineligible for award.”). The Court disagrees with this reading of the Solicitation’s language because CBP did not have “discretion” to further evaluate once CBP determined an RREP not to be similar in scope and complexity under the plain terms of the Solicitation. AR at 689 (TAB 18.2) (Solicitation); see CGS-ASP Security, JV, LLC v. United States, 162 Fed. Cl. 783, 805 n.4 (2022) (quoting Bankers Ins. Co. v. Fla. Residential Prop. & Cas. Joint Underwriting Assoc., 137 F.3d 1293, 1298 (11th Cir. 1998) (“‘Will’ . . . can be a[] mandatory word[].”). The Court notes, however, in some cases CBP further evaluated some offerors’ proposals even when their required RREPs failed; the government did not give clear reasoning as to why the TET did this. Tr. at 27:22–25 (“[GOVERNMENT]: I’m not sure what the evaluation team’s rationale was there. It’s possible that . . . they wanted to provide the additional notation for some.”). For clarity, the Court proposed a definition from a Third Circuit case, Obasi Invest. Ltd., to further define “similar” at oral argument: “Similar[]”. . . is most aptly defined as “[h]aving a marked resemblance or likeness; of a like nature or kind.” Similar, Oxford English Dictionary 59 (1st ed. 1933). True, there are varying degrees of likeness that might be described as “similar.” See In re Bernard L. Madoff Inv. Sec. LLC, 773 F.3d 411, 419 (2d Cir. 2014) (“In ordinary usage, the word ‘similar’ means ‘having characteristics in common,’ or ‘alike in substance or essentials.’” (quoting Webster’s New Int’l Dictionary 2120 (3d ed. 1993))) . . . A commonsense example explains why this is so. We might describe a “sedan” as similar to a “truck”—both are vehicles, after all. But an ordinary English speaker would not say a sedan is similar to a “light-duty pickup truck.” The use of a narrowing term of art that distinguishes one class of trucks from others connotes a likeness of specific functions—beyond basics like personal transportation. Obasi Invest. Ltd. v. Tibet Pharms, 931 F.3d 179, 185–86 (3rd Cir. 2019). All parties agreed to this definition of “similar.” Tr. at 35:2–5 (“THE COURT: Would you agree with Judge Hardiman’s sentiments there regarding how the word ‘similar’ can be applied? [PLAINTIFF SPOKESPERSON:] Yes, Your Honor . . .”). Given this contract involves emerging technology and the development of AI, the RREPs submitted necessarily needed to address relatively recent experience or else the experience would not be “similar in scope and complexity.” Tr. at 74:20– 24 (“[GOVERNMENT]: I think just by the nature of the work that [the agency is] seeking to procure, [work experience] would have to be relatively recent. So AI systems in the 1990s are not going to cut it in terms of similarity in scope and complexity.”). 1. Ekagra’s RREP 2, Task 3 Ekagra argues CBP used unstated evaluation criteria under Task 3—Artificial Intelligence/Machine Learning (AI/ML) Support Services, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. Ekagra MJAR at 14–17 (“[I]n evaluating Ekagra second RREP example under Task 3, the Agency deemed the example unacceptable because it did not meet four PWS ‘requirements’ . . . [R]ather than evaluating Ekagra’s RREP 2 to determine whether it was similar in scope and complexity, the Agency turned its bullets into baseline ‘requirements’ and then compared them in rote fashion to Ekagra’s solution.”). Ekagra explains CBP “required quotes to be the same as the list for Task 3 instead of having characteristics in common with the list in Task 3 . . . [and] thus imposed an ‘identical’ requirement rather than a ‘similarity’ requirement.” Id. at 17. The relevant language of Ekagra’s RREP 2, Task 3 and CBP’s corresponding evaluation is as follows:

Compare RREP 2. AR at 4734–35 (TAB 34b¹) (Ekagra Proposal).

With CBP’s Evaluation. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

[XXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX]

•RREP 2 Task 3 - No – The vendor did not address the following requirement in the RFP scope for Task 3:

o Developing an approach to labeling, annotation, and ontology development to support application of AI at CBP.

o Applying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.

o Establishing and managing a centralized library of annotated data for use across the organization for AI model training.

o Managing and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.

Compare AR at 4734–35 (TAB 34b¹) (Ekagra Proposal) with AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

For the Task 3 subtask describing the “[d]evelop[ment] [of] an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP,” see AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), Ekagra argues its “RREP did address an approach to labeling and annotation [because] an AI / ML tool cannot exist without it.” Ekagra MJAR at 17. For the Task 3 subtask describing “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology,” see AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), Ekagra states “[t]he only way to ensure that an AI or ML program works is to ensure that it knows what data to look for. And the only way to do that is to tag and label data samples.” Ekagra MJAR at 17. For the Task 3 subtask describing “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training,” see AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), Ekagra argues: “labeled and tagged data . . . would only be (maximally) useful if contained in a centralized repository.” Ekagra MJAR at 17. For the Task 3 subtask describing “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), Ekagra conceded its RREP 2 for Task 3 “does not address . . . the Authority to Operate.” Ekagra MJAR at 18. The government responds Ekagra’s RREP “does not describe an experience in which Ekagra developed software” but rather “describes Ekagra’s work . . . [XXXXXXX X XXXX XXXXXXXX XXXXXXXX XXXXXXX].”

Gov’t’s Cross-MJAR and Resp. to Group 1 at 37. The government reasons: “The agency concluded that [XXXXXXXX XXXXXXXXX XXXXXX X XXXXXXXXX XXXX XXXXXXXX XXXX XX] not the same as developing software.” Id. at 39.

The Federal Circuit has held when a protestor’s challenges “deal with the minutiae of the procurement process in such matters as technical ratings . . . , which involve discretionary determinations of procurement officials . . . [,] a court will not second guess” those determinations. E.W. Bliss Co. v. United States, 77 F.3d 445, 449 (Fed. Cir. 1996). The Federal Circuit has explained “[p]rocurement officials have substantial discretion¹⁵ to determine which proposal represents the best value for the government.” Id. The Federal Circuit has also explained the standard for determining whether an agency applied unstated criterion: if the agency “used a significantly different basis in evaluating the proposals than was disclosed.” Banknote Corp. of Am., Inc. v. United States, 56 Fed. Cl. 377, 387 (2003), aff’d, 365 F.3d 1345 (Fed. Cir. 2004).

15At oral argument, the government argued CBP was entitled to “Triple Whammy Deference.” Tr. at 290:10–16 (“[GOVERNMENT]: I think some of the cases that we’ve cited for the technical deference points, there’s this concept of the triple whammy of deference, which is not a great term, but it’s easy enough to find in Westlaw where it says if this is a technical matter within the agency’s expertise, the Court is . . . most deferential.”). The term “Triple Whammy Deference” comes from Judge Block’s decision in Overstreet Elec. Co. v. United States and only applies in negotiated procurement cases. Overstreet Elec. Co. v. United States, 59 Fed. Cl. 99, 117 (2003) (“[T]his court must defer to the SSA’s decision absent a showing that the SSA’s evaluation was arbitrary, capricious or contrary to law. . . . But that is not all. This standard becomes more deferential when dealing with a negotiated procurement such as the one in this case. . . . And when a procurement involves performance standards, which is what this court also faces, a court must grant even more deference to the evaluator’s decision. . . . In other words, Overstreet must overcome a triple whammy of deference by demonstrating by a preponderance of the evidence that the SSA lacked any rational basis to grade Boldt as ‘Exceptional.’”) (emphasis added). Here, CBP did not have a negotiated procurement but rather “two (2) multiple award Blanket Purchase Agreements (BPAs) . . . utilizing the General Service Administration (GSA) Multiple Award Schedule (MAS) Information Technology Category, Small Business Community.” AR at 627 (TAB 18.2) (Solicitation). As such, the Court will not assess CBP’s actions using “Triple Whammy Deference” as described in Overstreet. Overstreet Elec. Co., 59 Fed. Cl. at 117.

Plaintiff Ekagra agreed CBP found its RREP 2, Task 3 lacking for four reasons. Tr. at 63:19–23 (“THE COURT: . . . [B]eginning with Task 3, RREP 2, so to confirm at the onset, CBP found your RREP was lacking [four of] the seven items in the bulleted task description list, correct? [EKAGRA]: Yes.”). Per the Solicitation, if “any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated.” AR at 689 (TAB 18.2) (Solicitation). The Court therefore need only find CBP rationally assigned one of these four failures to Ekagra’s RREP 2 for Task 3 for Ekagra to have been properly eliminated from award. See id. The Court agrees with the government regarding the Solicitation’s inclusion of “shall” connoting required subtasks within the Tasks. Tr. at 30:17–22.

CBP first found Ekagra’s RREP not similar in scope and complexity because it did “not address . . . [d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Ekagra argues its “RREP did address an approach to labeling and annotation” by virtue of including “an AI / ML tool.” Ekagra MJAR at 17. At oral argument, Ekagra conceded its description of its approach to labeling and ontology “was not all that clear.” Tr. at 68:24–69:6 (“[EKAGRA]: I think we missed a word there. I apologize. ‘A sophisticated AI machine learning tool such as the one in this RREP.’ . . . I don’t think I was all that clear in the brief on that point, but to hone in more specifically, I do think that there probably are AI or machine learning tools that could exist without that, but that is a fundamental aspect typically of AI machine learning.”). Ekagra ultimately agreed its Task 3, RREP 2 does not state it addresses developing an approach to labeling, annotation, and ontology development explicitly. Tr. at 69:13–20 (“[THE COURT]: . . . [D]id Ekagra explicitly address past experience akin to ‘developing an approach to labeling, annotation, and ontology development to support CBP’? [EKAGRA]: Is Your Honor asking did they say the words? They did not say the words.”); see also Tr. at 71:7–8 (“[EKAGRA]: Again, Your Honor is correct that . . . these terms[, ‘ontology,’] are not explicitly stated.”). Ekagra at oral argument also argued CBP “should infer” from its proposal that it meets the requirements for Task 3. Tr. at 72:18–20 (“[THE COURT:] [T]he Government should infer that? [EKAGRA]: The Government should infer that or explain why it is not going to infer that.”); see also Tr. at 70:3–5 (“[EKAGRA:] [T]he way this is explained, the Government would have to recognize that ontology’s labeling and annotation were a fundamental part of it.”) (emphasis added). Ekagra pointed to the following language in RREP 2 at oral argument as being similar to the subtask: [XXXX XXXXXXXXX XXX XXXXXXX XX XXXXXXX XXXXX XXX XXXXXXX X XXXXXXX XXXXXXXXX XX XXXX XXXXXXXXXXX XXX XXXXXXXXXXX XXXXXXXXXX XXXXXX X XXXXXXX XXXXXXXXX] Tr. at 70:9–12; AR at 4734–35 (TAB 34b¹) (Ekagra Proposal). Ekagra explained “in order to do the activities throughout an entire federal network with differing agendas and differing requirements, annotation and labeling of data would be required.” Tr. at 70:13–17 (emphasis added).

Here, the challenges raised by Ekagra are those “deal[ing] with the minutiae of the procurement process,” E.W. Bliss Co., 77 F.3d at 449, as they contest CPB’s technical assessment of whether certain past performance examples are “similar in scope and complexity.” AR at 689 (TAB 18.2) (Solicitation). In this technical review, CBP is entitled to “substantial discretion.” E.W. Bliss Co., 77 F.3d at 449.

First, CBP rationally found Ekagra’s RREP 2 did “not address . . . [d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), because nowhere in the RREP does it describe development of such an approach. AR at 4734–35 (TAB 34b¹) (Ekagra Proposal). Ekagra’s RREP 2 only describes having “an AI / ML tool.” Id. Ekagra concedes its RREP does not explicitly address an “approach to labeling, annotation, and ontology development.” Tr. at 69:1–20; see also AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). CBP was not required to “infer,” Tr. at 72:18–20 (Ekagra), Ekagra’s RREP addresses “[d]eveloping an approach to labeling, annotation, and ontology development,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); offerors are responsible for submitting well-written, detailed proposals. Structural Assocs., 89 Fed. Cl. at 744. CBP assessed Ekagra’s RREP 2 in accordance with the terms of the Solicitation, determining whether the RREP was “similar in scope and complexity” to Task 3, and concluded the RREP lacked a discussion of “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” See AR at 10089 (TAB 71) (Track 2 TET Consolidated Review) (“The vendor did not address the following requirement in the RFP scope for Task 3: [d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP.”). CBP did not apply an unstated criterion because it followed the terms of the Solicitation by assessing whether the RREP was similar in scope and complexity in its analysis. AR at 689 (TAB 18.2) (Solicitation); AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP did not “use[] a significantly different basis in evaluating the proposals than was disclosed”—it held RREP 2 to the terms of the Solicitation and found discussion of similar experience lacking. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Ekagra ultimately challenges the “severity” of CBP’s assessment of its RREP 2, but its challenge falls flat because CBP followed the terms of the Solicitation in its assessment of how similar the RREP was. CBP’s assessment does not rise to the level of applying “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. Further, Ekagra’s challenges “deal with the minutiae of the procurement process.” E.W. Bliss Co., 77 F.3d at 449. As such, CBP rationally found Ekagra’s RREP 2 failed on this subtask; the Court “will not second guess” the agency’s determination. Id.; Impresa Construzioni Geom. Domenico Garufi v. United States, 238 F.3d 1324, 1332 (Fed. Cir. 2001) (“[T]he procurement official’s decision [did not] lack[] a rational basis”).

CBP next found Ekagra’s RREP was not similar in scope and complexity because it did “not address . . . [a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Ekagra argues its RREP 2 meets this subtask because RREP 2 describes an “AI or ML program . . . that [] knows what data to look for,” meaning the program can “tag and label data samples.” Ekagra MJAR at 17. At oral argument, Ekagra stated its RREP 2 contains language related to “apply[ing] labeling ontology for CBP and annotat[ing] data of varying types,” specifically: [XXXXX XXXXX XXXXXXXXXX XXXXXXXXXXX XXXXXXXXXX XXX XXXXXXXX XXXX XX X XXXXXXX XXXXXXXXX XXX XXX XXXXXXXXXXXX XXXXXXXXXX] Tr. at 71:12–14 (Ekagra). Ekagra further pointed to the following language in its RREP: [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXX] Tr. at 71:15–18; see AR at 4734–35 (TAB 34b¹) (Ekagra Proposal). Nowhere in the RREP, however, does Ekagra explicitly describe “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). While the RREP describes “normaliz[ing,] validat[ing], and analyz[ing] data,” Tr. at 71:12–14 (Ekagra); see also AR at 4734–35 (TAB 34b¹) (Ekagra Proposal), these skills are not the same as “[a]pplying labeling ontology . . . and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As agreed by the parties at oral argument, “ontology” is defined as: a knowledge structure consisting of terminologies (topics), their definitions, and relational information within one or multiple domains. This semantically represented information can be used for downstream tasks, such as document classification and recommendation systems.

N.C. Santosa et al., Automating Computer Science Ontology Extension with Classification Techniques, 9 IEEE ACCESS 161815 (2021); Tr. at 67:22–68:8 (“THE COURT: In a recent IEEE article, . . . [a] team of computer science faculty described ontology as follows: ‘A knowledge structure consisting of terminologies or topics, their definitions and relational information within one or more multiple domains, this semantically represented information can be used for downstream tasks such as document classification and recommendation systems.’ Would you agree with that? [GOVERNMENT]: That sounds fair, yes. THE COURT: . . . [plaintiff spokesperson], would you agree with that as well? [PLAINTIFF SPOKESPERSON EKAGRA]: Yes, Your Honor.); see also Tr. at 67:5–7 (“[THE COURT]: . . . Do you agree with that [IEEE ontology] definition? [EKAGRA]: I think that’s a good summary of what that means.”).

The process of [XXXXXXXXXXXXXXX XXXXXXXXXXXXX XXX XXXXXXXXXXX XXXXX] Tr. at 71:12–14 (Ekagra); see also AR at 4734–35 (TAB 34b¹) (Ekagra Proposal), is different from “[a]pplying labeling ontology . . . and annotat[ing] data of varying types from various sources according to the ontology,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), because an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Ekagra’s experience [XXXXXXXXXXXXX XXXXXXXXXXX XXX XXXXXXXXX XXXX] is less complex than what Task 3 calls for, which requests the application of “labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Ekagra’s MJAR only notes its RREP has “an AI / ML tool,” which, Ekagra argues, inherently “knows what data to look for” and “tag[s] and label[s] data samples.” Ekagra MJAR at 17; AR at 4734 (TAB 34b¹) (Ekagra Proposal). As with the first subtask, CBP was not required to infer the RREP addresses “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology” when the RREP fails to include any explicit discussion of such experience. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As such, CBP assessed Ekagra’s RREP 2 in accordance with the terms of the Solicitation and found it was “not similar in scope and complexity” to Task 3, as the RREP lacked a discussion of “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP did not apply an unstated criterion. Id.

Third, CBP found Ekagra’s RREP was not similar in scope and complexity because it did not address “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). For this Task 3 subtask, Ekagra argues its AI/ML model inherently had these features because “labeled and tagged data . . . would only be (maximally) useful if contained in a centralized repository.” Ekagra MJAR at 17. At oral argument, Ekagra pointed to the following in its RREP: [XXXX XXXXXXXXX XXX XXXXXXX XX XXXXXXX XXXXX XXX XXXXXXX X XXXXXXX XXXXXXXXX XX XXXXX XXXX XXXXXXXXXXX XXX XXXXXXXXXXX XXXXXXXXXX XXXXXX X XXXXXXX XXXXXXXX XX XXXXXX XXXX XXXXX XXXXXXXXX XXXX XXXX XXXXXXX XXXXXXXX XX XXXXXXXXX XXX XXX XXXXXXXXXXX XXXXXXXXX XXXXXXXXXXX XXX XXXXXXXX XXXXXXXXX XXXXXXXXX XX XXXXX XXX XXXXXXX] Tr. at 71:20– 72:11; AR at 4734 (TAB 34b¹) (Ekagra Proposal). Ekagra’s RREP does not explicitly address “[e]stablishing and managing a centralized library of annotated data,” rather it merely includes a description of an AI/ML model. AR at 4734–35 (TAB 34b¹) (Ekagra Proposal); Ekagra MJAR at 17. Ekagra argues CBP should have “infer[red]” from its description Ekagra had a “centralized repository” and therefore should have deemed its RREP similar in scope and complexity. See Ekagra MJAR at 17 (“Common sense dictates it would make no sense for that labeled and tagged data to exist in disparate libraries. It would only be (maximally) useful if contained in a centralized repository.”) (emphasis added); Tr. at 72:18–20 (“THE COURT: So the Government should infer that? [EKAGRA]: The Government should infer that or explain why it is not going to infer that.”). As with the other subtasks, CBP was not required to infer Ekagra’s RREP addresses “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training” when the RREP does not discuss such skills; offerors are responsible for submitting well-written, detailed proposals. See supra; Structural Assocs., 89 Fed. Cl. at 744; AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Second, even if CBP inferred Ekagra had a “centralized repository” for its data from its AI/ML model description, merely having a “centralized repository” is not the same as “[e]stablishing and managing a centralized library of annotated data,” which requires ongoing management and administration of annotated data. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As such, CBP assessed Ekagra’s RREP 2 in accordance with the terms of the Solicitation and found it was “not similar in scope and complexity” to Task 3, AR at 689 (TAB 18.2) (Solicitation), as the RREP lacked a discussion of “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP did not apply an unstated criterion. Id.

Finally, Ekagra conceded for the Task 3 subtask describing a prototype for “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), its RREP 2 “does not address . . . the Authority to Operate,” which is granted to “systems [once they] become operational and . . . produc[e] live data.” Ekagra MJAR at 18. This omission alone, which indicates Ekagra’s RREP was not “operational,” id., is enough for CBP to determine Ekagra’s RREP 2 was not similar to Task 3 because RREP 2 fails to address one of the Task 3 subtasks entirely. As the Solicitation stated, if “any of the . . . RREPs are not similar in scope and complexity, [CBP would deem] the quotation . . . ineligible for award,” so CBP was justified in determining an RREP missing experience addressing a key subtask was dissimilar. AR at 689 (TAB 18.2) (Solicitation). CBP thus need not find additional problems with Ekagra’s RREPs beyond this omission for CBP to eliminate Ekagra under the terms of the Solicitation. Id.

Accordingly, for each of the subtasks CBP found Ekagra to be missing for Task 3, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. Ekagra’s challenges “deal with the minutiae of the procurement process”; as such, the Court “will not second guess” the agency’s determination. E.W. Bliss Co., 77 F.3d at 449; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

2. Unissant’s RREPs¹⁶

a. Unissant’s RREP 1, Task 2

Unissant argues CBP used an unstated evaluation criterion for Task 2 – Operational Maintenance Support Service, see supra Section I.A, in assessing its RREPs 1 and 2 and determining scope and complexity. Unissant MJAR at 12. Unissant argues CBP “irrationally determined that Unissant’s RREP did not demonstrate performance that was similar to the scope of the Task 2 requirements” “[d]espite the fact that both RREPs for Task 2 referenced Unissant’s experience performing ‘Operational Maintenance Support Services’” and “providing end-to-end operations and maintenance support for critical IT Systems, including new applications and systems implemented by Unissant [XXX XXX XXXXX] under these contractual efforts.” Id.

The relevant language of Unissant’s RREP 1 for Task 2, and CBP’s corresponding evaluation, is as follows: Compare RREP 1. AR at 8377–78 (TAB 45b²) (Unissant’s Proposal).

16In its MJAR, Unissant also relied on description from its CPARs/CPRs as support for its RREPs meeting Task requirements. Unissant MJAR at 10, 12. At oral argument, the government confirmed there was no requirement for CBP to consider the CPARs/CPRs as part of its evaluation of the RREPs; Unissant ultimately conceded this point, agreeing with the government. Tr. at 47:3–10 (“THE COURT: In determining whether the RREPs were similar, was CBP required to look at the CPR, [] the contractor performance records, or the CPARs, Contractor Performance Assessment Reporting System, for degrees of similarity? . . . [THE GOVERNMENT]: No.”); Tr. at 228:3–8 (“[UNISSANT]: Your Honor, we did not find any specific language in the RFQ that would require them to consider [CPARs/CPRs]. THE COURT: [] So you agree with the Government, then, that it was not required? [UNISSANT]: That’s correct, Your Honor.”).

With CBP’s

Evaluation. AR at

10090 (TAB 71)

Compare AR at 8377–80 (TAB 45b²) (Unissant Proposal) with AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

(Track 2 TET Consolidated

First, CBP found Unissant’s RREP 1 not similar in scope and complexity because it “did not address [the] following Task 2 requirement in RREP 1: . . . [experience] perform[ing] defect resolution and corrective maintenance categorized as break/fix.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). For this Task 2 subtask, Unissant argues its RREP 1 “detailed its use of specific applications to ‘implement visibility, health, and optimization standards into the [client’s] ecosystem’ . . . which included system failure repairs to ‘return the asset to a fully functional state.’” Unissant MJAR at 19–20 (quoting AR at 8377–78 (TAB 45b²) (Unissant’s Proposal)). Unissant further asserts its RREP 1 “specified that it integrated certain [XXXXXXX XXXXXXX XXX XXXXXXXXXXXXX XXXXXXXXXXXXX XXXXXXXXXX XXXXXXXXXX XXXXX XX XXXXXX XXXX XXX XXXX XXXXXX XXXXXXXX XXX XXXX XXXXXXXXXXX XX XXXXXXX XXXXXXXXXXX] Id. at 20 (citing AR at 8377–78 (TAB 45b²) (Unissant’s Proposal)). The government responds, “Unissant’s RREP 1 for Task 2 is unresponsive to the SOW requirements because it largely describes configuration management (CM) and inventory management processes, not the required operational maintenance support activities.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 58–59 (citation and internal quotations marks omitted). The government explains configuration management— “the process of planning, approving, and implementing changes to IT systems and applications”—is distinct from “[operational maintenance] support,” which involves “identifying and resolving potential problems before they impact users.” Id. at 59.

Here, Unissant’s RREP 1 lacks any discussion of “perform[ing] defect resolution and corrective maintenance categorized as break/fix.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Unissant’s RREP 1 only describes:

[XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX]

AR at 8377–78 (TAB 45b²) (Unissant’s Proposal). This RREP largely details an approach to “configuration management,” AR at 8377–78 (TAB 45b²) (Unissant’s Proposal), which is distinct from performing defect resolution and corrective maintenance in the operational maintenance context. AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation) (“The Contractor shall perform operational maintenance support activities for the newly developed and enhanced applications developed by the Contractor.”). At oral argument, the Court, referencing the government’s MJAR, noted Unissant’s RREP 1 was akin to configuration management which is the process of “‘addressing the process of planning, approving, and implementing changes to IT systems and applications,’ rather than operational maintenance support, [which] . . . involve[s] identifying and resolving potential problems before they impact users.” Tr. at 92:18–24. Unissant responded by reading from its RREP language involving configuration management processes. Tr. at 93:9–17 (“[UNISSANT:] [W]e leverage the [XXXXXXX XXX XXXXXXXX XXXXXXXXXXXXX XX XXXXXXXXX XXXXXX XXXXXXXXXX XX XXX XXX XXXXXXXXX] Then there’s consolidation, maintenance permitting the customer to possess a better understanding of the complex infrastructure and allowing the ability to reduce market-ready and minimum time, the average time to repair a system failure and return the asset to a fully functional state. That’s responding to a problem and restoring it to a fully functional state.”).

Unissant conceded at oral argument its RREPs do not explicitly address the subtask of “perform[ing] defect resolution and corrective maintenance categorized as break/fix.” Tr. at 92:1–13 (“THE COURT: . . . [W]here in Unissant’s RREP does Unissant discuss past experience with ‘performing defect resolution and corrective maintenance,’ categorized as ‘break/fix’? [UNISSANT]: Again, those specific words, ‘break/fix,’ don’t appear . . .”); AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). When asked to point to specific language in RREP 1 meeting Task 2, Unissant struggled to highlight specific language and “paraphrased” in its response. Tr. at 94:1–5 (“THE COURT: I’m wondering where it is in the RREP. [UNISSANT]: I was paraphrasing . . .”). The configuration services in RREP 1 demonstrate experience with “addressing the process of planning, approving, and implementing changes to IT systems and applications” rather than Task 2’s operational maintenance support services, “[which] . . . involve identifying and resolving potential problems before they impact users.” Tr. at 92:18–24 (the Court) (internal citations omitted). Unissant did not refute its RREP 1 primarily describes configuration management services, and such services are distinct from operational maintenance support services. Tr. at 92:18–93:19. CBP sufficiently explained why RREP 1 failed to meet Task 2, for failing to discuss this subtask entirely: “The vendor did not address . . . perform[ing] defect resolution and corrective maintenance categorized as break/fix.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). CBP assessed Unissant’s RREP 1 in accordance with the terms of the Solicitation, determining whether the RREP was “similar in scope and complexity” to Task 2, and concluded the RREP lacked a discussion of “perform[ing] defect resolution and corrective maintenance categorized as break/fix.” See id. CBP did not apply an unstated criterion because it followed the terms of the Solicitation by assessing whether the RREP was similar in scope and complexity. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; see AR at 689 (TAB 18.2) (Solicitation); AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

Second, for the Task 2 subtask to “collaborate [with the government] and identify what corrective action is required,” Unissant argues its RREP 1 “addressed . . . collaboration with the respective agency teams on developing and deploying fixes [and] Unissant’s experience providing multi-tiered service desk and all other necessary support to meet its customers’ needs at agreed-upon performance and service level standards.” Unissant MJAR at 20 (emphasis added). The government does not respond in its cross-MJAR directly to Unissant’s argument it addressed collaboration, but notes RREP 1 fails entirely given CBP found it to fail on other subtasks. See Gov’t’s Cross-MJAR and Resp. to Group 1 at 61. At oral argument, Unissant explained its approach to “collaboration” was “inherent” in its help desk support described in RREP 1. Tr. at 92:10–17 (quoting AR at 8377–78 (TAB 45b²) (Unissant’s Proposal)) (“THE COURT: But what about specifically collaborating and required corrective action? [UNISSANT]: That’s inherent in the help desk support, including, . . . ‘Unissant provides IT infrastructure and operations support to ensure and verify the service delivery and the structure complies with all equitable regulations, policies, procedures, standards, et cetera.’”). Unissant’s RREP 1 states the following about its help desk: “Our IT support and services are provided through a multi-tiered Service Desk, subject to agreed-upon performance and service level standards.” AR at 83778 (TAB 45b²) (Unissant’s Proposal).

Unissant’s RREP 1 does not address “collaboration” explicitly. See id. All RREP 1 states is Unissant has a multi-tiered Service Desk by which it provides IT support and services. Id. CBP was not required to determine whether Unissant’s multi-tiered Service Desk “inherently” addressed “collaborat[ing] [with the government] and identify[ing] what corrective action is required,” because offerors are responsible for submitting well-written, detailed proposals. Structural Assocs., 89 Fed. Cl. at 744. CBP need not infer certain “collaboration” experience from RREP 1 when Unissant’s RREP 1 does not describe the Task 2 subtask of “collaboration”—only “Service Desk”-related services. AR at 8377–78 (TAB 45b²) (Unissant’s Proposal). CBP, in following the terms of the Solicitation, sufficiently explained RREP 1 was not similar for its failure to address this subtask. AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation) (“The Government and the Contractor shall collaborate and identify what corrective action is required.”). CBP did not apply unstated criterion because it followed the terms of the Solicitation by assessing and determining the RREP was not similar in scope and complexity in its analysis. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; see AR at 689 (TAB 18.2) (Solicitation); AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

Third, for the Task 2 subtask to “provide 24x7x365, on-call remote support to corrective issues on applications . . . requiring 24x7x365 service,” Unissant states: “‘Our IT support and services are provided through a multi-tiered Service Desk, subject to agreed-upon performance and service level standards. We provide access to IT support for identified High Value Assets in accordance with mission requirements.’” Unissant MJAR at 25 (quoting AR at 8378 (TAB 45b²) (Unissant’s Proposal)). Unissant concedes its RREP 1 does not explicitly describe 24x7x365 support. Id. at 25 (“[Unissant] did not call out 24x7x365 support explicitly.”). The government responds “[a]lthough RREP 1 mentioned a multi-tiered Service Desk, . . . there is no indication . . . this was provided 24x7x365 or that it focused on the more complex O&M tasks raised by Task 2, rather than the simpler issues generally attended to by a Service Desk.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 61 (citation omitted). The government reasons: “Unissant’s RREP 1 under Task 2 makes no reference to, or demonstration of, O&M support activities, and CBP correctly concluded that it was therefore not responsive to the requirements of Task 2.” Id.

As with the prior subtask, Unissant concedes its RREP 1 does not address “24x7x365 support.” Unissant MJAR at 25 (“[Unissant] did not call out 24x7x365 support explicitly.”). All RREP 1 states is Unissant has a multi-tiered Service Desk by which it provides IT support and services. AR at 8377–78 (TAB 45b²) (Unissant’s Proposal). CBP was not required to infer Unissant’s multi-tiered Service Desk “provide[d] 24x7x365, on-call remote support to corrective issues on applications . . . requiring 24x7x365 service,” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review), because offerors are responsible for submitting well-written, detailed proposals. Structural Assocs., 89 Fed. Cl. at 744. Unissant’s challenge fails because RREP 1 does not describe “24x7x365 support,” and CBP, in following the terms of the Solicitation, determined RREP 1 was not similar to Task 2, which called for “24x7x365, on-call remote support to corrective issues.” AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation) (“The Government and the Contractor shall collaborate and identify what corrective action is required. The Contractor shall provide 24x7x365, on-call remote support to corrective issues on applications identified by CBP as requiring 24x7x365 service.”). CBP did not apply an unstated criterion because it followed the terms of the Solicitation by assessing whether the RREP was similar in scope and complexity in its analysis. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; see AR at 689 (TAB 18.2) (Solicitation); AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

Accordingly, for each of the subtasks CBP determined Unissant was missing for Task 2, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 1 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 19.2) (Solicitation) (explaining if “any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated.”); Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

b. Unissant’s RREP 2, Task 2

Unissant argues CBP used an unstated evaluation criterion for Task 2 – Operational Maintenance Support Service, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. Unissant MJAR at 12. The relevant language of Unissant’s Task 2 RREP 2, and CBP’s corresponding evaluation, is as follows:

Compare RREP 2. AR at 8379–80 (TAB 45b²) (Unissant Proposal).

With CBP’s Evaluation. AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

[XXX XXXXXX] was also responsible for the operations and maintenance to support the modernization [XX XXX XXXX XXXXXXXXX XXX XXXXXXX XXXXXXXXXXXXXXXXXXXXX XXX XXXXX] systems administrators support network operations by monitoring the applications and integration points to ensure continuous performance of the systems. As these front-line systems support [XXXXXXXX XX XXX XXXXX], it was imperative that network and system performance was maintained. We developed monitoring and alerting for all systems, including host and application-level alerting[X XXXXX XXXXXXXX XXXXX XXX XX XXXXXXXXXX] All network and connection issues were sent as alert messages for system administrators to resolve based on priority and criticality. We also developed data dictionaries and user manuals to support user training. Unissant: 2nd look

•RREP 1 & 2 Task 2: No – The vendor did not address in following Task 2 requirements in RREP 1 or 2: o The Contractor shall perform defect resolution and corrective maintenance categorized as break/fix.

o The Government and the Contractor shall collaborate and identify what corrective action is required.

o The Contractor shall provide 24x7x365, on-call remote support to corrective issues on applications identified by CBP as requiring 24x7x365 service.

Compare AR at 8379–80 (TAB 45b²) (Unissant Proposal) with AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

First, for the Task 2 subtask to “perform defect resolution and corrective maintenance categorized as break/fix,” Unissant argues its RREP 2: “similarly described the defect monitoring and resolution process it developed . . . [XX XXXXXXXXXX XXXX XXXXX XXXXXXX XXXXXXXX] which included ‘monitoring and alerting for all systems, including host and application-level alerting,’ and a process for sending ‘[a]ll network and connection issues [ ] as alert messages for system administrators to resolve based on priority and criticality.’” Unissant MJAR at 20 (citing AR at 8379–80 (TAB 45b²) (Unissant Proposal)). Unissant also highlighted [XXX XXXXX XXXXXXX] administrators’ support of network operations, who “‘ensur[e] continuous performance of the systems.’” Id. at 12 (quoting AR at 8379 (TAB 45b²) (Unissant Proposal)). The government responds: “Unissant’s description of [XXX XXXXXXX XXXX] in RREP 2 demonstrates maintenance of the network and system performance but not O&M for software applications development, as called for by Task 2.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 62. The government explains, “Unissant stated that, under RREP 2, ‘all network and connection issues were sent as alert messages for system administrators to resolve based on priority and criticality’” and distinguishes this from an “approach to handling Task 2’s requirement to perform defect resolution and corrective maintenance categorized as break/fix.” Id. at 62–63 (citing AR at 8379 (TAB 45b²) (Unissant Proposal); AR at 248 (TAB 10b⁵) (Task 2, Solicitation)).

As with RREP 1, Unissant’s RREP 2 lacks any discussion of “perform[ing] defect resolution and corrective maintenance categorized as break/fix.” AR at 10090 (TAB 71) (Track We developed monitoring and alerting for all systems, including host and application-level alerting[X XXXXX XXXXXXXX XXXXX XXX XX XXXXXXXXXX] All network and connection issues were sent as alert messages for system administrators to resolve based on priority and criticality. We also developed data dictionaries and user manuals to support user training. All network and connection issues were sent as alert messages for system administrators to resolve based on priority and criticality. We also developed data dictionaries and user manuals to support user training.

2TET Consolidated Review). Unissant’s RREP 2 only states:

AR at 8379 (TAB 45b²) (Unissant Proposal). RREP 2 largely describes network maintenance, characterized by “monitoring [the system] and alerting” system administrators by message for ultimate resolution. See id. As discussed supra, Task 2 called for performing defect resolution and operational maintenance, which involves providing ongoing technical support during and after the development of applications. See AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation) (“The Contractor shall perform operational maintenance support activities for the newly developed and enhanced applications developed by the Contractor.”). Here, RREP 2 is not on point because it describes a monitoring and alert system, see AR at 8379 (TAB 45b²) (Unissant Proposal), while Task 2 calls for ongoing operational maintenance, which involves “identifying and resolving potential problems before they impact users.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 59. CBP, in following the terms of the Solicitation, rationally determined RREP 2

was not similar given its lack of discussion of operational maintenance services. AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation).

Second, for the Task 2 subtask to “collaborate [with the government] and identify what corrective action is required,” see AR at 10090 (TAB 71) (Track 2 TET Consolidated Review), Unissant provides the same argument for its RREP 2 as for its RREP 1, namely RREP 2 “addressed . . . collaboration with the respective agency teams on developing and deploying fixes [and] Unissant’s experience providing multi-tiered service desk and all other necessary support to meet its customers’ needs at agreed-upon performance and service level standards.” Unissant MJAR at 20 (citing AR at 8377–80 (TAB 45b²) (Unissant Proposal, RREPs 1 and 2 for Task 2)). The government responds Unissant does not quote any portion of RREP 2 related to this subtask and notes Unissant’s briefing “appears to quote from RREP 2 exactly once” in support of another subtask. Gov’t’s Cross-MJAR and Resp. to Group 1 at 63.

Unissant, failing to cite to any specific language from RREP 2 in its MJAR, provides little to no explanation as to why RREP 2 meets this Task 2 subtask. See Unissant MJAR at 20. Unissant merely reiterates the same argument for its RREP 2 as it does for its RREP 1, supra. Unissant MJAR at 20 (citing AR at 8377–80 (TAB 45b²) (Unissant Proposal, RREPs 1 and 2 for Task 2)). RREP 2 provides no references on the ability to “collaborat[e] and identify what corrective action is required.” AR at 8379–80 (TAB 45b²) (Unissant Proposal). Here, CBP rationally found Unissant’s RREP 2 failed on this subtask because it contains no explanation regarding how its RREP supports this subtask; the Court “will not second guess” CBP’s determination. Id.; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Third, for the Task 2 subtask to “provide 24x7x365, on-call remote support to corrective issues on applications identified by CBP as requiring 24x7x365 service,” see AR at 10090 (TAB 71) (Track 2 TET Consolidated Review), Unissant argues RREP 2 “addressed . . . Unissant’s experience providing multi-tiered service desk and all other necessary support to meet its customers’ needs at agreed-upon performance and service level standards.” Unissant MJAR at 20 (citing AR at 8377–80 (TAB 45b²) (Unissant Proposal, RREPs 1 and 2 for Task 2)). Similar to the prior subtask, the government responds, Unissant does not quote any portion of RREP 2 related to “provid[ing] 24x7x365, on-call remote support,” and “appears to quote from RREP 2 exactly once,” for another subtask. See Gov’t’s Cross-MJAR and Resp. to Group 1 at 63. Unissant’s RREP 2 does not address providing 24x7x365, on-call remote support. AR at 8379–80 (TAB 45b²) (Unissant Proposal). In its MJAR, Unissant does not cite to specific language from RREP 2 to support having experience providing such support. See Unissant MJAR. Given RREP 2 contains no explanation of how it supports this subtask, CBP’s assessment Unissant failed on this subtask was rational. Id.; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Accordingly, for the subtasks CBP determined Unissant was missing for Task 2, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 2 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 18.2) (Solicitation); Garufi, 238 F.3d at 1332.

c. Unissant’s RREP 1, Task 3

Unissant argues CBP used an unstated evaluation criterion for Task 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services, see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. Unissant MJAR at 13. Unissant states its RREP “addressed Unissant’s experience supporting [XXX XXXXXX XX XXXXX XXXXXXXXXXX XXXXXXX XX XXX XXXXXXXXXXXXXX XX XXXXXXXX XXX XXXXXXXX XXXX XXX XX XXX XXXX] Id. The relevant language of Unissant’s RREP 1 for Task 3, and CBP’s corresponding evaluation is as follows:

Compare RREP 1. AR at 8390–91 (TAB 45b²) (Unissant’s Proposal).

With CBP’s Evaluation. AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

Based on a series of recommendations from Unissant, [XXX XXXXX XXXXXXXXX XXXX XXXXXXXXX XXXXXXXXXX XXX XXXXXXXXX XXXXXXX XX XXX XXXXXXXX XXXXXXXX XXXX XXXX XXX XXX XXX XXXXXXXX XXX XXXX XXXXXXXXXXX XX XXX XX XXXXXXX XXXXXXXXXX XXXXXX XXXX XX XXXXXXX XXX XXX

XXXXXXXXX XXX XXXXXXXX XXX XXXXXXXXXX XX XXXXXXXXXX]. . . .

Instead of sifting through a large volume of historic tickets to determine the priority which could have potentially taken months to complete, Unissant built a Natural Language Processing (NLP) based AI model to categorize the incidents and prioritize them based on various criteria provided by the customer. Unissant built the initial data labeling, ontology and annotations needed for the AI model to run. [XXX XXXXXXXX XXX XXXX XX XXXXX XXXXXXXXXX XXXXX XXX XXXX XXXXXXXX] is currently extending this AI model to then route the requests intelligently to the correct group for processing and action. . . .

RREP 1 & 2 Task 3: No – The vendor did not address the following requirements:

o Developing an approach to labeling, annotation, and ontology development to support application of AI at CBP.

o Applying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.

o Establishing and managing a centralized library of annotated data for use across the organization for AI model training.

Compare AR at 8390–91 (TAB 45b²) (Unissant’s Proposal) with AR at 10090 (TAB 71) (Track First, CBP found Unissant’s RREP 1 not similar in scope and complexity because it did not address “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). To meet this subtask, Unissant argues it “discussed how it had built an AI model (or prototype), including ‘data labeling, ontology and annotations needed for the AI model to run.’” Unissant MJAR at 13 (citing AR at 8390 (TAB 45b²) (Unissant Proposal)). Unissant notes it has experience “developing ‘Natural Language Processing (NLP) based AI models,’ which were further based on ‘initial data labeling, ontology and annotations’ that Unissant built out and trained on.” Id. at 21 (citing AR at 8390–91 (TAB 45b²) (Unissant Proposal)). The government does not respond directly to Unissant on this subtask in its Cross-MJAR but argues Unissant’s RREP 1 example of “buil[ding] the initial data labeling, ontology and annotations needed for the AI model to run” [XXX XXX XXX XXXX XXXX] is not reflective of what Task 3 stated. See Gov’t’s Cross-MJAR and Resp. to Group 1 at 64. At oral argument, the government stressed the “burden was on the quoter to put . . . information [related to developing an approach to labeling, annotation, and ontology] in [its proposal] and to add the terms that were going to definitely explain [to CBP the offeror] had the relevant experience to show that they could support CBP’s needs.” Tr. at 107:6–9.

2TET Consolidated Review).

Unissant’s RREP 1 states:

Unissant built a Natural Language Processing (NLP) based AI model to categorize the incidents and prioritize them based on various criteria provided by the customer.

Unissant built the initial data labeling, ontology and annotations needed for the AI model to run.

AR at 8390 (TAB 45b²) (Unissant’s Proposal). Unissant built this model [XXX XXXXX XXXXX XXXXX XXXXXXXX] to “sift[] through a large volume of historic tickets.” Id.

Here, RREP 1 uses many of the “buzz words” contained in Task 3, such as “data labeling,” “ontology,” “annotations,” and “AI.” Id. at 8390–91 (TAB 45b²) (Unissant’s Proposal). RREP 1, however, does not explicitly discuss the “[d]eveloping [of] an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” See AR at 10090 (TAB 71) (Track 2 TET Consolidated Review) (emphasis added). RREP 1 involves building a model for a ticketed help desk, which is distinct from “support[ing the] application of AI at CBP” via the development of a method for “labeling, annotat[ing], and ontology.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). This experience describes building a simple help desk model for the isolated function of “sifting through a large volume of historic tickets,” AR at 8390–91 (TAB 45b²) (Unissant’s Proposal), whereas the Task called for the development of an advanced “approach to labeling, annotation, and ontology development to support application of AI at CBP.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). This help desk experience does not demonstrate a sophisticated ongoing approach to supporting AI across an organization. Compare AR at 8390–91 (TAB 45b²) (Unissant’s Proposal) with AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Given this RREP contains a different experience than the one requested by the subtask, CBP rationally found Unissant’s RREP 1 failed on this subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Second, for the Task 3 subtask, “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology,” see AR at 10090 (TAB 71) (Track 2 TET Consolidated Review), Unissant argues it “clearly addresses data labeling and development of an ontology” via RREP 1’s language, “Unissant built the initial data labeling, ontology and annotations needed for the AI model to run.” Unissant MJAR at 14 (citing AR at 8390 (TAB 45b²) (Unissant Proposal)). The government responds Unissant’s “example [XXXXXXX XX XXXXXXXXX XXX XXXXXXX XXXX XXXXXXXXX XXXXXXXX XXX XXXXXXXXXXX XXX XXX XXX XXXX XXXX] does not reflect ‘[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.’” Gov’t’s Cross-MJAR and Resp. to Group 1 at 64.

As with the prior subtask, Unissant points to its experience “buil[ding] the initial data labeling, ontology and annotations needed for the AI model to run,” Unissant MJAR at 14 (citing AR at 8390 (TAB 45b²) (Unissant Proposal)), to demonstrate its ability to “[a]pply[] labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). As discussed in Section VI.A.1, an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Unissant’s experience “buil[ding] the initial data labeling, ontology and annotations needed for the AI model to run,” Unissant MJAR at 14 (citing AR at 8390 (TAB 45b²) (Unissant Proposal), does not explicitly address “[a]pplying labeling ontology . . . and annotat[ing] data of varying types from various sources according to the ontology” and is less complex than what Task 3 calls for in terms of applying a labeling ontology. AR at 8390–91 (TAB 45b²) (Unissant’s Proposal). RREP 1’s model for a ticketed help desk is especially distinct from “[a]pplying labeling ontology . . . and annotat[ing] data of varying types,” as the help desk provides the simple service of ticketed notification rather than the advanced process of working within a sophisticated knowledge structure. AR at 10090 (TAB 71) (Track 2 TET Consolidated Review); see Santosa et al. at 161815. CBP rationally found Unissant’s RREP 1 failed on this subtask as the RREP contains a different experience than the one requested by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”). For the Task 3 subtask describing “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training,” Unissant concedes its RREP 1 did “not . . . expressly use[] the term ‘centralized library.’” Unissant MJAR at 14; Tr. at 101:3–6 (“THE COURT: . . . [Y]ou concede the proposal does not explicitly address having a centralized library? [UNISSANT]: That’s right, Your Honor.”). The government responds, Unissant’s RREP 1 does not reflect “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 63 (citation and internal quotations omitted). Here, as conceded at oral argument, Unissant’s RREP does not explicitly address establishing and managing a centralized library. Unissant MJAR at 14; Tr. at 101:3–6. As such, CBP did not “use[] a significantly different basis in evaluating the proposals than was disclosed,” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345, in determining RREP 1 was not similar to the subtask. AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Accordingly, for each of the subtasks CBP determined Unissant was missing for Task 3, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 1 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 18.2) (Solicitation); Garufi, 238 F.3d at 1332. d. Unissant’s RREP 2, Task 3 Unissant argues CBP used an unstated evaluation criterion for Task 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. Unissant MJAR at 13–14. This RREP 2 “addressed the performance of [XXXXXXXXXXX XXXXXXXXXXXXXX XXXXXXXX XXXXXXXXXX XXXX XXXXXXXXXX XX XXX XX XXX XXX XXXXXXX XXXXXXXXXXXX XXXXXXXX] Id. The relevant language of Unissant’s RREP 2, for Task 3, and CBP’s corresponding evaluation, is as follows:

Compare Unissant’s RREP 2. AR at 8392–93 (TAB 45b²) (Unissant Proposal).

With CBP’s Evaluation. AR

at 10090 (TAB 71) (Track 2 TET Consolidated Review).

[XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXX]. We made the data readable by developing a web portal and simple user interface (UI) with multiple views and tabs for searching data and describing relationships between elements of the supply chain. . . . [XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXX]. This enables the analyst to analyze and assemble data from various sources all in one platform. The program improves the speed and efficiency of analysis by providing searchable access to indices of large data sets and can process text, audio, and image files. . . . We leverage modern AI techniques to translate audio, video, and images to text and then apply keyword extraction algorithms that highlight the valuable information found in RREP 1 & 2 Task 3: No – The vendor did not address the following requirements:

o Developing an approach to labeling, annotation, and ontology development to support application of AI at CBP.

o Applying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.

o Establishing and managing a centralized library of annotated data for use across the organization for AI model training.

the text. [XXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX]. We also worked on speech to text functionality, researching speech to text packages to extend the keyword extraction process to audio files, creating initial image classification models to extend the keyword extraction process to images, and creating initial speech to text transcription models to extend the keyword extraction process to audio files. We also implemented cross-network query functionality and added the keyword extraction output to the article metadata to improve our search accuracy.

Compare AR at 8392–93 (TAB 45b²) (Unissant Proposal) with AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

First, CBP found Unissant’s RREP 2 not similar in scope and complexity because it did not address “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). For this subtask, Unissant argues [XXX XXXXXXXXXXXXX XXXXXXXXX XXXXX XX XX XXXXXXXXXXXXXXX XXX XXXXXXXX XXXXXXXXXXX XXXXXXXXX XXXXXX XXX XXXXXXX XXXX XXXXXXXX XX XXX XXXXX] Unissant MJAR at 14 (citing AR at 8392 (TAB 45b²) (Unissant Proposal)). Unissant argues, “[t]he model/prototype ‘assemble[s] data’ and maintains it within a single platform, which is searchable and creates datasets from multiple sources.” Id. (citing AR at 8392–93 (TAB 45b²) (Unissant Proposal)). Unissant explains “[t]hese prototypes were used for demonstration and training prior to being released.” Id. (citing AR at 8393 (TAB 45b²) (Unissant Proposal)). The government does not address the specifics of Unissant’s RREP 2 for Task 3 but argues “[t]he short of it is that Unissant cannot now fix through attorney argument and appeals to inference what it should have done in its initial submission.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 65.

Here, Unissant’s RREP 2 presents experience “collect[ing], understand[ing], annotat[ing], stor[ing], and connect[ing] the data ingested by the AI.” AR at 8392 (TAB 45b²) (Unissant Proposal). This experience is different from “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP” because the latter involves the establishment of a sophisticated approach to data organization. See supra; AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Indeed, as agreed by the parties at oral argument, see, e.g., Tr. at 67:22–68:8, an ontology is a sophisticated “knowledge structure” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Although Unissant’s RREP 2 shows experience using AI to collect and handle complex data sets, it fails to explain Unissant’s work with such an organizational structure. See AR at 8392–93 (TAB 45b²) (Unissant Proposal). Given Unissant describes a different experience than the one requested by the Task, the RREP is not similar to the Task. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345.; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Second, for the Task 3 subtask describing “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology,” see AR at 10090 (TAB 71) (Track 2 TET Consolidated Review), Unissant argues it “clearly addresses data labeling and development of an ontology.” Unissant MJAR at 14 (citing AR at 8392 (TAB 45b²) (Unissant Proposal)) (describing the “AI model/prototype ‘to collect, understand, annotate, store, and connect data’”). Unissant explains its “proposal also detail[ed] its experience developing and building out a[n] NLP based AI solution using ‘keyword extraction algorithms’ and further creating and testing transcription models that cover images, speech, and text file types based on various data sources for accurate and predictive results.” Id. at 21 (citing AR at 8392–93 (TAB 45b²) (Unissant Proposal)).

As with the prior subtask, Unissant points to its experience creating an “AI model/prototype ‘to collect, understand, annotate, store, and connect data,’” Unissant MJAR at 14 (citing AR at 8390 (TAB 45b²) (Unissant Proposal)), to demonstrate its ability to “[a]pply[] labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Per Section VI.A.1, an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Unissant’s experience does not explicitly address “[a]pplying labeling ontology . . . and annotat[ing] data of varying types from various sources according to the ontology” and is less complex than what Task 3 calls for in terms of applying a labeling ontology as it involved prototyping a model to collect and label data rather than the creation of an ontological system, see supra. AR at 8390–91 (TAB 45b²) (Unissant’s Proposal). As such, CBP rationally found Unissant’s RREP 2 failed on this subtask because the RREP contains a different experience than the one requested by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345.; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Third, for the Task 3 subtask describing “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training,” Unissant notes its RREPs did “not . . . expressly use[] the term ‘centralized library.’” Unissant MJAR at 14. As with RREP 1, Unissant does not explicitly address establishing and managing a centralized library in RREP 2. Id. Given RREP 2’s lack of discussion of a centralized library, it was rational for CBP to find RREP 2 failed on this subtask for lack of similarity. Id.; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”). Accordingly, for each of the subtasks CBP evaluated Unissant as missing for Task 3, CBP rationally evaluated each in accordance with the Solicitation and did not rise to the level of applying “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 2 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 18.2) (Solicitation); Garufi, 238 F.3d at 1332. e. Unissant’s RREP 1, Task 4

Unissant argues CBP used an unstated evaluation criterion for Task 4 –RPA Support Services, see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. Unissant MJAR at 15–16. The relevant language of Unissant’s RREP 1 for Task 4, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at 8402–8403 (TAB 45b²) (Unissant’s Proposal).

With CBP’s Evaluation. AR

at 10090 (TAB 71) (Track 2 TET Consolidated Review).

Based on a series of recommendations from Unissant, [XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX] Once the concept was proven through a POC, Unissant was asked to initiate the first set of automation. [XXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXX] Whether the request is through an email or through a phone call or auto detection of the backend monitoring

[XXXXXXXXXXXXXXXXXXXXXXXXXXX] and routed the ticket to the appropriate workstream. In addition, the automation provided the necessary basic information of the ticket to the help desk personnel whom the incident was routed to. In the process of implementing this solution, the Unissant team coordinated with [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXX] to ensure all required governance activities and policies are followed and will have a smooth ATO when moved to production. In addition, Unissant worked iteratively with the Mission Stakeholder and the CIO to ensure the implemented process automation is satisfactory and performs as expected.

RREP 1 Task 4: No – There

is no evidence of any RPA

work being performed on the contract.

Compare AR at 8402–8403 (TAB 45b²) (Unissant’s Proposal) with AR at 10090 (TAB 71) (Track 2 TET Consolidated Review).

Unissant argues “the Agency irrationally concluded that Unissant had not performed RPA work apparently because Unissant did not use the word ‘robotics’ in its first Primary RREP for this task” even though “in its RREP, Unissant addresses each of the actual tasks under this Task, such as governance activities and the ATO Process.” Unissant’s MJAR at 16. The government responds: “RPA is a specific type of automation that uses software robots to automate repetitive, rules-based tasks.” Gov’t’s Cross-MJAR and Resp. for Group 1 at 66. Given this specific definition of RPA, the government reasons, “Unissant’s RREP lacked sufficient information for the TET to determine the precise automation methodology Unissant employed in that instance.” Id. The government also took issue with “Unissant us[ing] the same exact example to illustrate both RPA (RREP 1 Task 4) and AI/ML (RREP 1 Task 3).” Id. at 67; Tr. at 115:23–116:1 (“[THE GOVERNMENT:] Task 4 is specifically looking for robotic process automation, and they submitted their same experience from Task 3 that was artificial intelligence based.”).

At oral argument, Unissant asserted CBP treated the bullets in Task 4 as requirements even though Task 4 uses the word “may,” meaning CBP treated permissive (“may”) and mandatory (“shall”) language the same without explanation. Tr. at 85:9–16 (“[THE COURT:] While Task 4 does use the phrase ‘may include,’ how do you account for the use of the word ‘shall’ in connection with the other tasks that are bulleted? [UNISSANT]: . . . If it says you shall perform that, that’s a performance requirement. It’s not a proposal submission requirement.”); Tr. at 89:15–19 (“[UNISSANT:] [T]here’s no difference in the way they looked at Task 4 compared to the others. It just said, did you meet this requirement, yes or no? That’s all that is reflected in the agency evaluation.”). The government responded “may” and “shall” are different, and these specific word choices helped CBP determine the “degree of similarity” under which to analyze. Tr. at 86:11–20 (“THE COURT: . . . I think we spoke about this briefly, about how the directive for Task 4, unlike for other tasks, utilized the permissive ‘may include.’ . . . What was the reason for that? [GOVERNMENT]: I think it goes back to the degree of similarity, that here the agency . . . listed a lot of subpoints, but they’re allowing for a wider field of experiences with the ‘may include’ rather than the ‘shall include.’”).

Here, RREP 1 does not discuss experience with RPA work; rather, it discusses the same AI-based experience from Unissant’s RREP for Task 3. Compare AR at 8402–8403 (TAB 45b²) (Unissant’s Proposal) with AR at 8390–91 (TAB 45b²) (Unissant’s Proposal). Unissant conceded this at oral argument. Tr. at 114:23–115:10 (“THE COURT: . . . [T]he Government notes that your RREP submitted is the same for Task 3 and Task 4, but these were different tasks, correct? [UNISSANT]: Yes. THE COURT: So how does that justify the same exact RREP? . . . [UNISSANT:] The same RREP can accomplish multiple things. . . . THE COURT: So your argument is that it’s broad enough? [UNISSANT]: Yes.”). AI-based experience is not the basis of Task 4, and CBP need not accept AI-based experience for a subtask focused on RPA work. AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). Further, CBP’s evaluation of Task 4 only entails an assessment of the broader Task 4 requirement to have RPA experience, not any additional subtasks: “No – There is no evidence of any RPA work being performed on the contract.” AR at 10090 (TAB 71) (Track 2 TET Consolidated Review). As such, CBP did not fault Unissant for not demonstrating experience with the Task 4 sub-bullets in its assessment, so Unissant’s argument CBP treated the bullets in Task 4 as “requirements” fails. Id. As such, CBP rationally found Unissant’s RREP 1 failed on this subtask because the RREP contains a different experience than the one requested by the subtask. Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Accordingly, CBP rationally evaluated Unissant’s RREP 1 for Task 4 in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 1 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 19.2) (Solicitation); Garufi, 238 F.3d at 1332.

f. Unissant’s RREP 1, Task 5

Unissant argues CBP used an unstated evaluation criterion for Task 5 – Security and Privacy Support, see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. Unissant MJAR at 16–17. Unissant explains its RREP 1 “describes the work performed by Unissant for [XXX XXXXXX XX XXXXX XXXXXXXXXXX XXXXXXX XX XXX XXXXXXXXXXXXXX XXX XXXXXXXX XXX XXXXXXXXXX] Id. at 17. The relevant language of Unissant’s RREP 1 for Task 5, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at 8414–8515 (TAB 45b²) (Unissant’s Proposal).

With CBP’s Evaluation. AR at 10091 (TAB 71) (Track 2 TET Consolidated Review).

Task Area 5: Unissant remediates vulnerabilities [XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXX] and we ensure least privilege is practiced in active directory environment. We support the [XXXXXXXX XXXXXXXXXX XXXXXXX] to conduct security incident management operations such as intrusion detection, system monitoring for indications of compromise or attack, vulnerability assessments, identity management, audit logs reviews for anomalies or events indicating malicious activity, network forensics, and insider threat indicator analysis. Our [XXXXXXXX XXXXXXXXXX XXXXXXX] provides 24X7 support to ensure that cyber security risks and threats are mitigated quickly. Unissant has developed and implemented an Incident Response Plan documenting incident response procedures per program requirements. We also provide [XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXX] We support Federal audit requirements associated with the periodic FISMA reviews. Unissant provides support [XXXXXXXXXXXXXXXXXXXXXXXXXXX RREP 1 Task 5: No – The

vendor did not address the following requirements in the RFP scope for Task 5:

o Coordinate with stakeholders to support . . . CBP security / privacy process and work to reconcile issues and blockers impeding approval of PTA, ATT, or ATO.

o Communicate with project owners on status of PTA, ATT, or ATO and escalate issues to appropriate Government POCs.

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX]

Compare AR at 8414–15 (TAB 45b²) (Unissant’s Proposal) with AR at 10091 (TAB 71) (Track 2 TET Consolidated Review).

CBP found Unissant’s RREP 1 not similar in scope and complexity because it did not address “[c]oordinat[ing] with stakeholders to support CBP security / privacy process and work[ing] to reconcile issues and blockers impeding approval of PTA, ATT, or ATO” or “[c]ommunicat[ing] with project owners on [the] status of PTA, ATT, or ATO and escalat[ing] issues to appropriate Government POCs.” AR at 10091 (TAB 71) (Track 2 TET Consolidated Review). For this subtask, Unissant argues its “proposal focused on the security and privacy support it provided under this contract, including cybersecurity incident management and ad-hoc reporting as required.” Unissant MJAR at 17 (citing AR at 8414–8515 (TAB 45b²) (Unissant’s Proposal)). Unissant asserts it described “its governance activities, including its Incident Response Plan, audit compliance, vulnerability remediation.” Id. Unissant notes it “did not call out communication or coordination because these activities were engrained within every step of the processes that Unissant did describe.” Id. (emphasis added). The government responds: “Although Task 5 solicits security and privacy support, Unissant chose for its RREP 1 a project that primarily provided security operations support. These are different things. . . .” Gov’t’s Cross-MJAR and Resp. for Group 1 at 69 (emphasis in original) (internal citations omitted). Here, Unissant’s RREP 1 does not directly address “[c]oordinat[ing] with stakeholders to support . . . CBP security / privacy process and work to reconcile issues and blockers impeding approval of PTA, ATT, or ATO” nor “[c]ommunicat[ing] with project owners on status of PTA, ATT, or ATO and escalat[ing] issues to appropriate Government POCs.” See AR at 8414–515 (TAB 45b²) (Unissant’s Proposal); AR at 10091 (TAB 71) (Track 2 TET Consolidated Review). Unissant concedes its RREP “did not call out communication or coordination.” Unissant MJAR at 17 (citing AR at 8414–15 (TAB 45b²) (Unissant’s Proposal)). At oral argument, Unissant asserted coordinating with stakeholders to support client security and privacy is “inherent” in the work described in its RREP 1 for Task 5. Tr. at 117:8–24 (“[THE COURT]: Can you identify where in the RREP it clearly addresses examples of ‘coordinating with stakeholders to support client security privacy’? . . . [UNISSANT:] [O]ur proposal focused on cyber incident management, governance activities such as creating an incident response plan, audit compliance, and vulnerability remediation . . . [and] if you’re doing all of those things without talking and coordinating with the customer, it would just fail. It’s . . . inherent in those things.”) (emphasis added). Also at oral argument, counsel for Unissant used much of his “own knowledge” to explain why the RREP 1 for Task 5 was similar. Tr. at 119:17–19 (“THE COURT: Where does it say that [Unissant performed security and privacy support] in the RREP? [UNISSANT]: That’s based upon my own . . . knowledge. And so FISMA, the Federal Information Security . . . Modernization Act, since 2014, is where that requirement comes from. We’re talking about FISMA compliance here. FISMA includes the information security and privacy requirements.”). CBP was not required to determine whether Unissant’s RREP “inherently” addressed coordination and collaboration because offerors are responsible for submitting well-written, detailed proposals. See supra; Structural Assocs., 89 Fed. Cl. at 744. As such, CBP rationally found Unissant’s RREP 1 failed on this subtask because it does not discuss the experience required by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Accordingly, for each of the subtasks CBP determined Unissant was missing for Task 5, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated Unissant for this RREP 1 in accordance with the terms of the Solicitation given the lack of similarity. AR at 689 (TAB 19.2) (Solicitation); Garufi, 238 F.3d at 1332.

3. Attainx’s RREPs

a. Attainx’s RREP 1, Task 2

AttainX argues CBP used an unstated evaluation criterion for Task 2 – Operational Maintenance Support Service, see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. AttainX MJAR at 16–17. The relevant language of AttainX’s RREP 1 Task 2, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at 1708 (TAB 24b²) (AttainX Proposal).

With CBP’s Evalution.

AR at 8907 (Tab 50)

(CBP’s Review of AttainX’s RREPs).

We maintained [XXXX XXXX XXXXXXXX] software applications including documenting, triaging, analyzing and providing short term and long-term fixes for the defects reported by the end users and the IT stakeholders on the IT applications and support systems that are in production environment. Relevant tasks include Operational Support, IT Service Management, Software Maintenance and Upgrades, and Training. AttainX resolved [XXXX] technical issues in one year through [XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXX], performed testing [XXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXX] leading to continued 24/7 availability of system. We provided help desk support for solutions in Production, supporting [XXXX] tickets per month and [XXX] end users. AttainX utilized [XXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXX]

•RREP 1 Task 2: The

vendor did not address

24X7x365 requirement.

Although the vendor mentioned there is a

“leads to continued 24/7

availability of system”,

this does not mean that they actually preform

24X7X365 O&M of the system. The system being available and monitoring of the system are two separate things.

Compare AR at 1708 (TAB 24b²) (AttainX Proposal) with AR at 8907 (TAB 50) (CBP’s Review of AttainX’s RREPs).

CBP found AttainX’s RREP 1 not similar in scope and complexity because it “did not address [the] 24X7[X]365 requirement.” AR at 8907 (TAB 50) (CBP’s Review of AttainX’s RREPs). For this subtask, AttainX argues the following from its RREP demonstrates a similar experience: “We maintained [XXXX XXXX XXXXXXXX] software applications including documenting, triaging, analyzing, and providing short term and long-term fixes for the defects reported by the end users and the IT stakeholders on the IT applications and support systems that are in [the] production environment.” AttainX MJAR at 17–18 (citing AR at 1708 (TAB 24b²) (AttainX Proposal)). AttainX notes it “explained that it ‘resolved [XXXX] technical issues in one year through [XXXXXXXXXXXXXXXXXXXXX XXXXXXXXX]’; and it ‘provided help desk support for solutions in Production, supporting [XXXX] tickets per month and [XXX] end users.’” Id. Further, AttainX highlights [XXXXX XXXXXXXXXXX XXXXXXXX XX XXX XXXXXXXXX XXXXXXXXXX XXX XXXXXXXX XXXXXX XX XXXXXXX XX XX XXXXXXX] Id. The government argues “CBP’s concern was that neither of the RREPs that AttainX supplied for Task 2 demonstrated 24x7x365 O&M support services.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 33–34 (emphasis in original). The government reasons RREP 1 “did not address the more complex and sophisticated O&M support required to identify and resolve problems with newly developed and enhanced applications, before they impact users, much less experience providing such support 24x7x365.” Id. at 34. The government explains “AttainX’s RREP 1 described support that was purely driven by user experience, ticket generation and resolution of issues as reported by users, of the simpler kind usually found in helpdesk support.” Id. At oral argument, the government noted when addressing AttainX’s system the determination of whether an offeror has a 24x7x365 service is necessarily “binary.” Tr. at 157:19 (“[GOVERNMENT:] It’s binary. It’s an up or down.”).

In University Research Co., LLC, University Research Co. (URC) alleged “the agency failed to conduct a proper past performance evaluation when [it] failed to distinguish between degrees of relevance within each offeror’s past performance.” Univ. Rsch. Co., LLC v. United States, 65 Fed. Cl. 500, 504 (2005). In reviewing the agency’s action, this court emphasized “all the agency asked for was information on contracts ‘for similar products or services,’ not identical ones,” meaning the agency’s “obligation was met through [its] determination that the past performance . . . met a threshold level of relevance.” Id. at 508 (emphasis in original). This court acknowledged there are multiple ways to review past performance, but “[a]ll that is required is the binary choice of yes/no regarding relevance.” Id. Here, CBP engaged in a similar “binary choice of yes/no regarding relevance” in assessing whether 24x7x365 services were present in the RREP. Id. The Court agrees with the government’s argument—an offeror either does or does not have 24x7x365 services.

Here, AttainX’s RREP describes a support system able to quickly resolve technical issues, but it does not describe the advanced O&M support capable of identifying and resolving problems before they occur. AR at 1708 (TAB 24b²) (AttainX Proposal); AR at 8907 (Tab 50) (CBP’s Review of AttainX’s RREPs). Attainx’s RREP 1 thus does not “address [the] 24X7x365 [O&M] requirement” necessitated by Task 2, a requirement that is inherently binary. Id.; Univ. Rsch. Co., LLC, 65 Fed. Cl. at 508 (“All that is required is the binary choice of yes/no regarding relevance.”). As such, CBP rationally found AttainX’s RREP 1 failed on this subtask because the RREP has a support system dissimilar to the 24x7x365 O&M system requested by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

Accordingly, CBP rationally determined AttainX was missing this subtask for Task 2 and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated AttainX for this RREP 1. Garufi, 238 F.3d at 1332.

b. AttainX’s RREP 2, Task 2

AttainX argues CBP used an unstated evaluation criterion for Task 2 – Operational Maintenance Support Service, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. AttainX MJAR at 16–17. The relevant language of AttainX’s RREP 2 Task 2, and CBP’s corresponding evaluation, is as follows:

Compare RREP 2. AR at 1709–10 (TAB 24b²) (AttainX’s Proposal).

With CBP’s Evalution. AR at 8907 (TAB 50) (CBP’s Review of AttainX).

AttainX is providing Tier 1, Tier 2 and Tier 3 Help Desk support [XXXXXXXXXXXXXXXXXXXX XXXXXXX]. These applications support [XXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXX.

These systems generate over [XXXXXXXXXXXXXXX

XXXXX] Due to a high number of users and requests for certificates, we have a large call volume at the help desk for customer support in the following areas to include

[XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXX] are currently using [XXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXX]. Users report the service requests [XXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX] The system [XXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXX

•RREP 2 Task 2: The vendor did not address 24X7x365 requirement or automated notification of issues requirement. Although the vendor

mentioned that the systems and applications are “mandated to be operational and available 24/7”, this does not mean that they actually preform 24X7X365 O&M of the system. The system being available and monitoring of the system are two separate things. In addition, the vendor mentioned automated notification, but it not notification related to issues requiring corrections, but rather notifications for users experience.

Compare AR at 1709–10 (TAB 24b²) (AttainX’s Proposal) with AR at 8907 (TAB 50) (CBP’s Review of AttainX).

First, CBP found AttainX’s RREP 2 not similar in scope and complexity because it “did not address [the] 24X7[X]365 requirement.” AR at 8907 (TAB 50) (CBP’s Review of AttainX). AttainX argues its RREP shows experience “‘providing Tier 1, Tier 2 and Tier 3 Help Desk support for [various] applications’ . . . [XXXXXX XXXXXXXXX XXXXXX XXXXXXXX XXXXXX XXX XXXXX XXXXXXXXXX XXXXX XXXXXX XXXXXXXX XXX XXXXXX XX XXXXXXX XX XXXXXX XXXX XXXX XXXXXXX XXX XXXXXX] AttainX MJAR at 18 (citing AR at 1709 (TAB 24b²) (AttainX Proposal)). AttainX reasons “[t]hese services . . . result in allowing government systems to remain available 24x7x365.” Id. (citing AR at 1710 (TAB 24b²) (AttainX Proposal)). The government responds: “Task 2 required offerors to demonstrate experience automating notification of issues requiring correction in the context of providing [O&M] support services for newly developed or enhanced applications . . . [which] are different from the helpdesk improvements that AttainX described in RREP 2.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 26. The government then explained, “O&M support for newly developed and enhanced applications is a complex and specialized task that involves identifying and resolving potential problems before they impact users.” Id. Here, AttainX’s RREP 2 does not directly address having a 24X7X365 O&M service; it only addresses “Help Desk support” with the ability to serve thousands of industry users and resolve help desk tickets. See AR at 1709–10 (TAB 24b²) (AttainX’s Proposal). Nowhere in AttainX’s RREP 2 does it describe having the constant, on-demand services as required by a 24X7X365 O&M system. Id. AttainX’s RREP only describes the ability to resolve help desk XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX]

The

legacy process was inefficient and contributed to a significant lag in the response from the help desk team.

AttainX successfully

[XXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXXXXXXX

XZXXXXXXXXXXXXXXXXXXXXXXXXXX]

tickets in large quantities, which is distinct from the advanced technical support for newly developed and enhanced applications required by Task 2’s O&M services. Id.; AR at 377–78 (TAB 12b⁴) (Solicitation – Statement of Work). At oral argument, AttainX conceded its help desk does not provide O&M services while arguing CBP should have been more specific in its references to O&M services in its Task 2 description. Tr. at 136:31–19 (“[ATTAINX:] [Task 2] does not call for operations and maintenance support. It calls for on-call remote support to correct the issues. . . . [I]f [CBP] wanted operational and maintenance support, they could have asked for it. They asked for 24/7/365 on-call remote support, and that’s precisely what a help desk is, [] on-call remote support to help with issues requiring correction and [] the reading into the language of operational maintenance is not [correct]. Yes, it’s the header, yes, it’s the first sentence of a distinct paragraph, but if the Government wanted only operational and maintenance support, then we could have fixed that.”) (emphasis added). Task 2 however specifically calls for “Operational Maintenance Support Service,” stating, “[t]he Contractor shall perform operational maintenance support activities for the newly developed and enhanced applications developed by the Contractor.” AR at 377–378 (TAB 12b⁴) (Solicitation – Statement of Work). Given AttainX’s RREP does not discuss 24X7x365 O&M services, it is not similar to Task 2. Id. As such, CBP rationally found AttainX’s RREP 2 failed on this subtask because the RREP has a support system dissimilar to the one requested by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Second, CBP found AttainX’s RREP 2 not similar in scope and complexity because it “did not address . . . [the] automated notification of issues requirement.” AR at 8907 (TAB 50) (CBP’s Review of AttainX). AttainX argues RREP 2’s help desk system “demonstrated experience with automated notification of issues requiring correction” because its [XXXXXXXX XXXXXXXXXXXXX XXXXXXX X XXXXXX XX XXXX XX XX XXXXXXXX XX XXXXX XX XXXXXXXXX XXXX XXX XXXXXX XXXXXXX] AttainX MJAR at 10–11 (citing AR at 1709–10 (TAB 24b²) (AttainX’s Proposal)). The government responds: “Task 2 called for automated notifications for issues covered by O&M support services [so] AttainX’s quote, which described in RREP 2 an upgrade [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] . . . is not responsive to that requirement.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 28. The government notes: “Task 2 . . . called for more than a helpdesk; it called for performing not just ‘defect resolution, [but also] corrective maintenance.’” Id. (citing AR at 248–49 (TAB 10b⁵) (Task 2, Solicitation)).

Here, AttainX’s RREP contains a form of “automated notification” because its help desk system “automatically creates . . . ticket[s].” AR at 8907 (TAB 50) (CBP’s Review of AttainX); AR at 1709–1710 (TAB 24b²) (AttainX’s Proposal). Task 2, however, calls for the “develop[ment] [of] mechanisms to automate notification of issues requiring correction; any issue that impacts the ability of a user to interact with applications must be resolved immediately.” AR at 377–78 (TAB 12b⁴) (Solicitation – Statement of Work). AttainX’s RREP discusses the automatic [XXXXXXXXXX], but this automatic feature is different from the notification of (1) issues requiring immediate correction and (2) completed corrections. Id.; see AR at 1709–10 (TAB 24b²) (AttainX’s Proposal). AttainX’s help desk does not go further than automatic notification and misses the later steps of the subtask requirement: “automate notification of issues requiring correction; any issue that impacts the ability of a user to interact with applications must be resolved immediately. The Contractor shall notify the Government when Corrective issues are corrected.” AR at 377–78 (TAB 12b⁴) (Solicitation – Statement of Work) (emphasis added). Given AttainX’s RREP does not provide experience related to automated notification of issues requiring correction and resolution of such issues, it is not similar to Task 2. Id. As such, CBP rationally found AttainX’s RREP 2 failed on this subtask because the RREP has a support system dissimilar to the notification and resolution system requested by the subtask. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Accordingly, for each of the subtasks CBP determined AttainX was missing for Task 2, CBP rationally evaluated each in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated AttainX for RREP 2. Garufi, 238 F.3d at 1332.

4. CMCI’s RREPs

a. CMCI’s RREP 1, Task 3

CMCI argues CBP used an unstated evaluation criterion for Task 3 – Artificial Intelligence/Machine Learning (AI/ML), see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. CMCI MJAR at 2, 7. The relevant language of CMCI’s RREP 1 for Task 3, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at 3018 (TAB 28b³) (CMCI Proposal).

With CBP’s Evaluation. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review).

[XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

CMCI: 2nd look

•RREP 1 – Task 3: No – The vendor did not address the following requirement in the RFP scope for Task 3:

o Developing an approach to labeling, annotation, and ontology development to support application of AI at CBP.

o Applying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.

o Establishing and managing a centralized library of annotated data for use across the organization for AI model training.

o Managing and execution of security and accreditation and Authority to Test (ATT) processes and Authority

Compare AR at 3018–19 (TAB 28b³) (CMCI Proposal) with AR at 10088 (TAB 71) (Track 2 TET Consolidated Review).

First, CBP stated CMCI’s RREP 1 did not address “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI argues: “[t]he evaluators essentially disqualified CMCI for using the more practical word ‘collation’ rather than the more esoteric term ‘ontology’ . . . [as] the actual experience described was the organizing of massive amounts of data into a form that could be used by CBP, which is exactly what it said it wanted.” CMCI MJAR at 8. The government responds collation and ontology have different meanings: “‘[c]ollation’ is the process of organizing and arranging data or information in a specific order or sequence, while ‘ontology’ refers to the formal representation of knowledge or concepts and their relationships in a specific domain.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 49. The government at oral argument did not, however, have a clear response as to the difference between “ontology” and “collation.” Tr. at 155:16–24 (“[GOVERNMENT:] [C]ollation is, you know, kind of stacking -- putting things together, but ontology -- and I think this is consistent with the definition you read -- is really kind of looking at the relationships between these things. There’s different ways that this data can be managed, and they want the AI then to be, you know, looking at this and looking for different relationships between things and how they should be structured, where collation is not going at that same level.”).

At the outset, the Court notes CMCI argued at oral argument none of CBP’s evaluations of the offerors’ RREPs mention the word “similar” in its assessment of whether or not an offerors quote was similar. Tr. at 160:12–16 (“[CMCI:] [I]n all of the evaluation documents for everyone, never once [did CBP] use the term ‘similar.’ That does not appear in the document anywhere, nor does ‘dissimilar.’ The term ‘complexity’ never appears in the evaluation document anywhere.”).

CMCI conceded at oral argument its RREP lacks four of the seven subtasks described by Task 3. Tr. at 151:4-8 (“THE COURT: . . . CBP found your Task 3 RREP was lacking for four of the seven items in the bulleted task description list. Do you agree that four of the seven were missing? [CMCI]: I do, Your Honor.”). This concession alone is enough to eliminate CMCI from award given it acknowledged, in lacking experience with four subtasks, its RREP 1 is dissimilar to the Solicitation’s description of Task 3. See AR at 689 (TAB 18.2) (Solicitation). For the first subtask involving the development of an approach to labeling, annotation, and ontology development, CMCI conceded its RREP 1 for Task 3 does not use the term ontology but rather “collation.” Tr. at 152:15–21 (“[CMCI:] Now, admittedly, the client did not use the term ‘ontology,’ but the client did use the term ‘collation’ and provides various examples of how they have successfully, for CBP, harnessed realtime aggregate rich data and used that data through artific[i]al intelligence to provide the screening and other types of artificial intelligence tasks that CB[P] requires.”). CMCI noted it wishes it used the word ontology in its RREP submission for Task 3, rather than collation, to describe its experience. Tr. at 154:11–13 (“[CMCI]: Yeah, I’m certain the client sitting here now wishes they had [said ontology], but they did not.”). Further, CMCI also agreed with the Court’s IEEE definition of ontology. See supra Section VI.A.1.; Tr. at 151:17–21 (“THE COURT: . . . [D]o you agree with the Court’s IEEE definition of ‘ontology’? [CMCI]: Because I lack any other more specialized knowledge, I will accept that definition.”). Per the IEEE definition, ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. CMCI’s experience “organizing [] massive amounts of data into a form that could be used by CBP” is not akin to developing a “knowledge structure,” as the latter is a more advanced way of establishing a web of information, including how the information relates. Id.; see CMCI MJAR at 8. As such, CBP did not apply an unstated criterion in assessing CMCI’s RREP 1 for this subtask; it distinguished RREP 1 as not addressing the subtask given its experience with “collation” rather than “ontology.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332. Second, CBP stated CMCI’s RREP did not address “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI argues this subtask is satisfied by “CMCI’s experience using [various commercial AI and data] tools” “to process data and for standardizing and harmonizing data.” CMCI MJAR at 8 (citing AR at 3018–19 (TAB 28b³) (CMCI Proposal)). CMCI also argues: “one can reasonably infer that to use data robots one would fundamentally require the data to be annotated.” Id. at 9. The government reasons, “[s]tandardizing and harmonizing is not data labeling or ontology . . . [because] [l]abeling and annotation is the process of assigning tags or categories of data, making it easier to identify and classify for machine learning algorithms [and] . . . [o]ntology is formal representation of knowledge or concepts and their relationships in a specific domain.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 51.

As noted by the government, “standardizing and harmonizing data,” AR at 3018 (TAB 28b³) (CMCI Proposal), is different than “applying labeling ontology and annotating data” because the former is a general class encompassing a multitude of processes while the latter is a specific, sophisticated process for organizing and labeling data. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI at best describes a general approach to working with data in RREP 1, which does not meet the specifications of applying a labeling ontology. Id. Further, CBP need not “infer” the RREP includes an approach to annotating data given its use of data robots when the RREP does not describe such an approach. See supra; Structural Assocs., 89 Fed. Cl. at 744. As such, CBP did not apply an unstated criterion in assessing CMCI’s RREP 1 for this subtask; it distinguished RREP 1 from the subtask because the RREP does not describe an approach to “applying labeling ontology and annotating data.” AR at 10088 (TAB 71) (Track Third, CBP stated CMCI’s RREP 1 did not address “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI argues “CMCI’s proposal discusses its experience utilizing ‘statistical and predictive models, using home-grown and advanced AI/ML tools that allowed CBP to harness real-time, aggregate rich data and gain agility for immediate action.’” CMCI MJAR at 9 (citing AR at 3019 (TAB 2b³) (CMCI Proposal)). CMCI reasons “[a]n AI/ML practitioner would know that all (no exception) statistical models are trained using large sets of annotated data which are typically stored and updated in a data store or library.” Id. The government responds: “the fact that data and models are ‘typically stored and updated in a data store or library,’ . . . says nothing about where CMCI’s data and models were actually stored in the work performed for this RREP.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 52 (quoting CMCI MJAR at 9). The government reasons, “[i]f CMCI’s data and models were indeed stored in a centralized library, it should have said so in its RREP.” Id. at 53.

2TET Consolidated Review); see Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

For this subtask, CMCI’s RREP 1 provides: [XXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.”

XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] AR at 3019 (TAB 28b³) (CMCI Proposal) (emphasis added). At oral argument, CMCI explained it uses a “library” to manage annotated data: “the RREP references using tools like Google Libraries, which I believe is a centralized library.” Tr. at 164:11–13 (CMCI). While CMCI concedes its RREP does not say centralized, CMCI’s RREP describes a library—the Google Library—like the “centralized library” detailed in Task 3. AR at 376–80 (TAB 12b⁴) (Solicitation – Statement of Work); Tr. at 164:18–20 (“[CMCI:]: The word ‘centralized’ doesn't appear in the RREP, admittedly, but the description of the tasking includes that.”). Contrary to the government’s argument CMCI “should have said [it had a centralized library] . . . in its RREP” if it did, CMCI’s RREP explicitly demonstrates it has a centralized library. Gov’t’s Cross-MJAR and Resp. to Group 1 at 53; AR at 3019 (TAB 28b³) (CMCI Proposal [XXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] CBP’s decision to eliminate CMCI for this specific subtask was arbitrary and capricious, as “the procurement official’s decision lacked a rational basis,” Garufi, 238 F.3d at 1332. CBP stated CMCI did not discuss having a centralized library when RREP 1 plainly references expertise using such. Per the Solicitation, a “quotation will be ineligible for award and will not be further evaluated” once CBP deemed a primary RREP not similar in scope and complexity. AR at 689 (TAB 18.2) (Solicitation). Thus, in terms of prejudice, CBP properly eliminated CMCI for failing to address (1) “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI at CBP”; (2) “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology”; and (3) “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CBP’s arbitrary and capricious decision on the subtask regarding centralized libraries therefore does not increase CMCI’s change of award because CMCI’s RREP fails for lack of discussion of the other subtasks, discussed supra and infra. Alfa Laval Separation, Inc. v. United States, 175 F.3d 1365, 1367 (Fed. Cir. 1999) (citing Statistica, Inc. v. Christopher, 102 F.3d 1577, 1581 (Fed. Cir. 1996)) (“To prevail in a bid protest, a protestor must show a significant, prejudicial error in the procurement process. . . . To establish prejudice, a protestor . . . must show ‘that there was a substantial chance it would have received the contract award but for that error.’”). As such, CMCI cannot establish prejudice with regard to this subtask. Id. Fourth, CBP stated CMCI’s RREP did not address “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI argues its RREP meets this subtask: “CMCI’s work on all phases of Enterprise Life Cycle would encompass promoting applications to live production, which would entail adherence to ATT, ATO processes and mitigations of items collected as part of POAMs.” CMCI MJAR at 9. CMCI reasons, “[t]he SOW words are not parroted, but the experience described should have been understood by someone with industry knowledge to have necessarily subsumed the potential tasking, especially so because CMCI’s experience was for CBP.” Id. at 9–10. The government responds Task 3 requires “a quoter with experience actually developing and creating the appropriate documentations needed to manage or execute ATT/ATO processes . . . [and] CMCI’s prior adherence to various protocols fails to speak to its ability to ‘[m]anag[e] and execut[e] security and accreditation’ of those processes.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 53. Here, RREP 1 does not explicitly address “managing and execution of security and accreditation and ATT processes and ATO processes for AI/ML technologies,” as there is no clear description in the RREP of such processes. AR at 3018 (TAB 28b³) (CMCI Proposal). While CMCI argues this experience was “subsumed in” its RREP’s discussion of Enterprise Life Cycles, there is no express reference to this subtask in the RREP and CBP need not infer the subtask experience was “subsumed in” CMCI’s discussion of Enterprise Life Cycles in its assessment of RREPs. Structural Assocs., 89 Fed. Cl. at 744; see Tr. at 166:5–17 (“[THE COURT:] Does your Task 3 RREP explicitly discuss experience actually developing and creating the appropriate documentations needed to manage or execute ATT or ATO processes? [CMCI]: It does in the reference . . . that we support all phases of enterprise life cycle activity, which includes getting the authorizations to use. That’s part of the enterprise life cycle activity. So, again, it was not spun out to tracking the exact words, but to people with technical expertise, that would be understood to be a subsumed task as part of enterprise life cycle activities.”) (emphasis added). “[A]dherence to ATT, ATO processes” is not akin to “managing and execution of security and accreditation and ATT processes and ATO processes for AI/ML technologies,” as the latter is a step above mere compliance. Compare AR at 3018 (TAB 28b³) (CMCI Proposal) with AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CBP’s decision to eliminate CMCI for this specific subtask was not arbitrary and capricious because the RREP is dissimilar. Garufi, 238 F.3d at 1332. CBP’s review of CMCI’s RREP 1 Task 3 was not arbitrary and capricious as CBP properly eliminated CMCI from competition for the first, second, and fourth subtasks in CBP’s review. CMCI was not prejudiced by CBP’s review of the third subtask because CMCI did not have a substantial chance of award due to the other subtask failures. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); Alfa Laval Separation, 175 F.3d at 1367 (citing Statistica, Inc., 102 F.3d at 1581) (“To prevail in a bid protest, a protestor must show a significant, prejudicial error in the procurement process. . . . To establish prejudice, a protestor . . . must show ‘that there was a substantial chance it would have received the contract award but for that error.’”). b. CMCI’s RREP 2, Task 5 CMCI argues CBP used an unstated evaluation criterion for Task 5 – Security and Privacy Support, see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. CMCI MJAR at 2, 5. The relevant language of CMCI’s RREP 2 for Task 5, and CBP’s corresponding evaluation, is as follows:

Compare RREP 2. AR at 3047-3048 (TAB 28b⁵) (CMCI Proposal).

With CBP’s Evaluation. AR at 10088

(TAB 71) (Track 2 TET Consolidated Review).

[XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

RREP 2 – Task 5: No – The vendor does software application security like code scans, but no indication of supporting federal security guidelines per the SOW.

o The Contractor shall provide security and privacy support to CBP’s portfolio of projects to attain the appropriate authorities and support security /

privacy activities. The Contractor shall support the development of documentation including Privacy

Threshold Assessments (PTAs),

Authorizations to Test (ATTs), and

Authorizations to Operate (ATOs) for the CBP’s portfolio of projects.

o Coordinate with stakeholders to support the CBP security / privacy process and work to reconcile issues and blockers impeding approval of

PTA, ATT, or ATO.

o Communicate with project owners on status of PTA, ATT, or ATO and escalate issues to appropriate

Government POCs.

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX]

Compare AR at 3047–48 (TAB 28b⁵) (CMCI Proposal) with AR at 10088 (TAB 71) (Track 2 TET Consolidated Review).

At the outset, the Court notes the parties did not articulate in a straightforward manner what CMCI’s working relationship is with [XXXXXXXX]—the contractor described in CMCI’s RREP 2 for Task 5. Tr. at 170:18–23 (“THE COURT: So [XXXXXXXX] supports AAG’s utilization of the EMASS for control, compliance, supporting artifacts, and then on to establishing strict process control mechanisms for obtaining authorization decisions . . . ? [CMCI]: Yes.”); Tr. at 174:1–3 (“[GOVERNMENT:] Just from the description, the Government could not assume that [XXXXXXXX] developed the documents needed for the authorization.”). From the RREP language, [XXXXXXXXXXXXXXXXX]. See AR at 3047– 48 (TAB 28b⁵) (CMCI Proposal).

First, CBP stated CMCI’s RREP 2 failed because “[t]he vendor does software application security like code scans, but [there is] no indication of supporting federal security guidelines per the SOW.” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CMCI argues: First, the RFQ and SOW did not state that “supporting federal security guidelines” was required. Second, CMCI’s proposal did, in fact, specifically address federal security guidelines, discussing its experience with network and Cloud security (in addition to other parts of the write-up pertaining to application security).

CMCI MJAR at 5. CMCI specifically points to its experience with the Enterprise Mission Assurance Support Service (eMass) system and compliance with federal standards, such as [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] Id. (internal quotations omitted). The government responds: “the SOW seeks deeper experience” than “complying with federal system security requirements (i.e., running vulnerability and virus scans and patch management)” because the Task calls for “security and privacy support.” Gov’t’s Cross-MJAR and Resp. to Group 1 at 54.

CMCI first protests in its MJAR “the RFQ and SOW did not state that ‘supporting federal security guidelines’ was required.” CMCI MJAR at 5. Task 5 is titled “Security and Privacy Support” and states “[c]ontractor[s] shall provide security and privacy support to CBP’s portfolio of projects to attain the appropriate authorities and support security / privacy activities.” AR at 379 (TAB 12b⁴) (Solicitation – Statement of Work). At oral argument, CMCI conceded Task 5 relates to federal security guidelines. Tr. at 169:8–18 (“THE COURT: . . . CBP stated your RREP 2 did not demonstrate past experience ‘supporting federal security guidelines.’ [] Task 5 is called, ‘Security and Privacy Support,’ and it seems to explicitly contemplate the awardee performing security and privacy functions. Do you agree with that and do you agree that Task 5 relates to conducting PTAs, ATTs, ATOs related to cyber and privacy guidelines? [CMCI]: Yes.”). Given this concession, the first part of CMCI’s argument fails—Task 5 does relate to supporting federal security guidelines. Further, CMCI’s RREP does not explicitly discuss “supporting federal security guidelines”; it merely discusses past compliance with certain federal guidelines. AR at 3047–48 (TAB 28b⁵) (CMCI Proposal). Mere compliance with a handful of federal guidelines as performed by CMCI is dissimilar from the specific security and privacy support experience expected by CBP. See AR at 379 (TAB 12b⁴) (Solicitation – Statement of Work) (noting Task 5 requires experience with numerous Federal Information Security Modernization Act of 2014 subparts, National Institute of Standards and Technology protocols, CBP policy documents, and other guidelines). As such, CMCI’s RREP 2 for Task 5 was properly eliminated under this subtask. Compare AR at 3047–48 (TAB 28b⁵) (CMCI Proposal) with AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

In its review of other Task 5 subtasks, CBP did not include its boilerplate language describing why RREP 2 failed to be similar; CBP only included the relevant subtask language: o The Contractor shall provide security and privacy support to CBP’s portfolio of projects to attain the appropriate authorities and support security / privacy activities. The Contractor shall support the development of documentation including Privacy Threshold Assessments (PTAs), Authorizations to Test (ATTs), and Authorizations to Operate (ATOs) for the CBP’s [portfolio] of projects.

o Coordinate with stakeholders to support the CBP security / privacy process and work to reconcile issues and blockers impeding approval of PTA, ATT, or ATO.

o Communicate with project owners on status of PTA, ATT, or ATO and escalate issues to appropriate Government POCs.

See AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). In other words, while CBP listed the subtasks it seemingly believed CMCI’s RREP 2 lacked, it failed to provide analysis and merely listed three subtasks without context or explanation. As CMCI puts it, CBP “simply [] regurgitated [three] additional RFQ Task 5 subtasks without stating that CMCI’s proposal had failed to show experience with these subtasks.” CMCI MJAR at 6. Thus, according to CMCI, the “Court should not consider reasons not supported by the Agency’s contemporaneous documentation.” Id. CMCI reasons “[e]ven if one were to infer that the Agency’s reiteration of subtasks implied that the Agency had concluded that CMCI’s proposal failed to show experience with the listed potential subtasks, such a conclusion is not supported by the evidence.” Id. The government responds:

CMCI invites the Court to ignore the bulleted list of Task 5 subtasks for which CBP found that CMCI had not exhibited relevant experience. CMCI [MJAR] 6 (suggesting that the agency’s failure to include certain form language it had typically used with other vendors means it failed to show why it made its decision).

The Court should decline the invitation because here the agency’s rationale and intent is apparent from the administrative record. See, e.g., Veterans Contracting Grp., Inc. v. United States, 920 F.3d 801, 806 (Fed. Cir. 2019) (refusing to find that the “contracting officer lacked any rational basis for cancelling [a] solicitation [where] the record discloses a reasonable motivation for cancellation”); cf. [Fla.

Power & Light Co. v. Lorion, 470 U.S. 729, 744 (1985)] (concluding that where “the reviewing court simply cannot evaluate the challenged agency action on the basis of the record before it, the proper course, except in rare circumstances, is to remand to the agency for additional . . . explanation”).

Gov’t’s Cross-MJAR and Resp. to Group 1 at 54 n.11. The government conceded at oral argument, however, the necessary explanatory language—i.e., “the vendor does not address the following subtasks”—is missing from CBP’s review. Tr. at 187:22–25 (“THE COURT: [ I]t doesn’t even say these bullet points from the tasks are not similar. [GOVERNMENT:] [I]t doesn’t say these bullet points from the tasks are not similar.”) (emphasis added). Given the lack of explanatory language, at oral argument, the government resorted to inferring from what was present in CBP’s review to demonstrate CBP’s rationale. Tr. at 175:18 (“[GOVERNMENT]: I’m just saying what’s not said.”); Tr. at 175:22–23 (“[GOVERNMENT]: [I am] inferring from what is here.”). Further, the government cited to Veterans Contracting in its MJAR and at oral argument to show the missing rationale was not arbitrary and capricious. See Gov’t’s Cross-MJAR and Resp. to Group 1 at 54 n.11; Tr. at 190:1–10 (“THE COURT: . . . I don’t think Veterans Contracting stands for [the proposition] a rationale can be missing. [I]t’s still necessary that the agency explain its reasoning. [GOVERNMENT]: The rationale has to be there. THE COURT: Yes. [GOVERNMENT]: Whether the explanation is so apparent that the agency doesn’t have to spend pages and pages explaining it, I think that’s what Veterans Contracting stands for.”). In Veterans Contracting, the Federal Circuit concluded the plaintiff failed to show “that the contracting officer lacked any rational basis for cancelling the . . . solicitation [because] the record disclose[d] a reasonable motivation for cancellation.” Veterans Contracting, 920 F.3d at 806. This language, which recognizes the agency provided a “reasonable motivation for” its decision, id., does not suggest there is a rule permitting an agency to not explain its reasoning in making a decision, contrary to what the government stated at oral argument. Tr. at 190:1–10.

In the instant case, CBP neglected to include the necessary explanatory language—i.e., “the vendor does not address the following”—in its review; it merely pasted three subtasks. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). The Court notes, given CBP’s review of other RREPs, it was perhaps logical for CMCI to assume the boilerplate language listed for the first bullet point (but missing for the remaining bullet points) applied to the remaining points. See AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). Ultimately, however, there is no record of the government analyzing these remaining Task 5 RREP 2 bullet points. Id. The government conceded this lack of analysis at oral argument. Tr. at 174:23–175:7 (“THE COURT: [W]here does the evaluation discuss [the missing requirements]? . . . [GOVERNMENT:] [T]he analysis doesn’t go much deeper”). As such, CBP’s evaluation of CMCI’s RREP 2, Task 5 was arbitrary and capricious because it lacked the key language explaining why the RREP was not similar to the subtasks provided, meaning “the procurement official’s decision lacked a rational basis” as the agency did not “provide[] a coherent and reasonable explanation.” Garufi, 238 F.3d at 1332 (citation omitted).

As noted, supra, the Solicitation stated a “quotation will be ineligible for award and will not be further evaluated” once CBP deemed the RREP not similar in scope and complexity. AR at 689 (TAB 18.2) (Solicitation). Thus, despite the agency’s arbitrary and capricious actions related to RREP 2, Task 5, CBP properly eliminated CMCI for failing to address supporting federal security guidelines. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). CBP’s lack of reasoning for the remaining subtasks, though arbitrary and capricious, does not increase CMCI’s chance of award because CMCI’s RREP 2 fails for not discussing supporting federal security guidelines, and its RREP 1 for Task 3 likewise failed for lack of similarity, see supra. Alfa Laval Separation, 175 F.3d at 1367 (citing Statistica, Inc., 102 F.3d at 1581) (“To prevail in a bid protest, a protestor must show a significant, prejudicial error in the procurement process. . . . To establish prejudice, a protestor . . . must show ‘that there was a substantial chance it would have received the contract award but for that error.’”). As such, CMCI cannot establish prejudice with regard to this subtask being arbitrary and capricious. Id.

5. AOI’s RREPs

Related to its MJAR arguments, AOI filed a Motion to Strike the Duneja Declaration attached to the government’s Cross-MJAR to Group 2 plaintiffs from the record for providing “extra-record explanations . . . to supplant CBP’s superficial findings.” AOI Mot. to Strike at 1– 2; see Duneja Decl. AOI argues: “The declaration does not limit itself to explaining arcane or technical concepts, it takes the next step of providing rationale for why AOI’s RREPs were not the same or sufficiently similar.” AOI Mot. to Strike Reply at 2. The government responds the Declaration was “included to assist the Court in understanding technical matters and background allegations at issue in this protest.” Gov’t’s Reply to Mot. to Strike at 1–2. The government explains, “the Court can resolve the cross-motion for judgment on the administrative record without reviewing or relying on the Duneja declaration . . . [so] the Court need not consider it to resolve the protest.” Id. at n.2.

The Duneja Declaration provides technical background on the technology described in the Track 2 Task areas. See Duneja Decl. For example, for Task 2 – Operational Maintenance Support Service, the Declaration provides the following technical background: The role of a system administrator entails user account management, server management, software updates and Patches and not fixing application defects. . . .

Incident management refers to the process of identifying, analyzing, and resolving incidents or disruptions in systems or services. It involves promptly addressing and resolving issues to restore normal operations and minimize any negative impact on users. Defect resolution, on the other hand, is a part of software development or quality assurance processes. It involves identifying and fixing defects or bugs in software or applications. This can include debugging, troubleshooting, and make code changes to eliminate defects. Systems administrator is responsible for managing and maintaining computer systems, networks, and servers. . . . Building dashboards does not equate to automatically setting up automatic notifications of issues requiring corrections. Notifications of downtime to reduce response time is system monitoring and not defect notification and resolution.

Duneja Decl. at Appx10. At oral argument, the government agreed with the Court the Declaration is primarily a “glossary of computer science background.” Tr. at 282:22–283:1 (“THE COURT: . . . [I]t’s the Government’s position that for the substantive paragraphs that are given, they are effectively a glossary of computer science background in order to understand the technology. [GOVERNMENT]: Yes, Your Honor.”). The information contained in the Declaration is largely technical in nature and does not contain “rationale for why AOI’s RREPs were not the same or sufficiently similar” to the Tasks. Mot. to Strike Reply at 2. Given the Declaration merely provides additional technical background, the Court need not and does not rely on it to rule on the parties’ MJARs and Cross-MJARs.¹⁷ Accordingly, the Court finds as moot AOI’s Motion to Strike.

a. AOI’s RREP 1, Task 3

AOI argues CBP used an unstated evaluation criterion for Task 3 – Artificial Intelligence/Machine Learning (AI/ML), see supra Section I.A, in assessing its RREP 1 and determining scope and complexity. See AOI Cross-MJAR at 15. The relevant language of CMCI’s RREP 1 for Task 3, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at 808 (TAB 21b²) (AOI Proposal).

With CBP’s Evaluation. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

[XXX] has a large and diverse user base located primarily in the United States but is also distributed across the world. To provide better services to its members, Alpha Omega recommended [XXX] to

•RREP 1 Task 3: No – The vendor did not address the following requirement in the RFP scope for Task 3:

o Gathering business and/or technical requirements relative to use cases Compare AR at 808 (TAB 21b²) (AOI Proposal) with AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

17The government attached the Tran Declaration and Gartner Article, Magic Quadrant for IT Service Management, to its Cross-MJAR to Group 1 plaintiffs. See Appx. to Gov’t’s Cross-MJAR and Resp. to Group 1, ECF No. 93-1. The Tran Declaration and Gartner Article, like the Duneja Declaration, provide technical background. See id. As with the Duneja Declaration, the Court need not, and does not, consider these attachments to rule on the parties’ MJARs and Cross-MJARs.

First, CBP stated AOI did not address in its RREP: “[g]athering business and/or technical requirements relative to use cases identified by the Government for the evaluation, or trial, of potential AI/ML capabilities.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues its experience “provid[ing] recommendations on leveraging the use of AI and ML technologies to” commercial clients meets the requirements of this subtask. AOI MJAR at 17. AOI reasons, “in the context of this commercial effort, AOI demonstrated a similar approach to recommending AI/ML solutions based on the customer’s requirements and business needs.” Id. The government responds: “there is no discussion of [] gathering business or technical requirements” in AOI’s RREP. Gov’t’s Cross-MJAR and Resp. to Group 2 at 66. At oral argument, the government explained: “[AOI] recommended AI to its client [XXXXXXXX]. Even if that could be construed as somehow gathering business requirements from the client, which it sounds like it’s actually the opposite, it’s [AOI] telling their client what to do rather than listening to the client, gathering information and requirements from the client, . . . they . . . simply don’t explain how they fit into this particular requirement.” Tr. at 197:21–198:4. AOI’s RREP does not discuss experience with “[g]athering business and/or technical requirements relative to use cases.” See AR at 808 (TAB 21b²) (AOI Proposal). At oral argument, AOI could not succinctly describe why its RREP was similar to the Task 3 subtask, and instead countered AOI merely failed to use “buzz words” in its RREP description. Tr. at 199:15–23 (“[AOI]: [T]his is a question of buzz words . . . . For the Government to contend that it was not similar because the word ‘case’ did not follow the word ‘use’ is really form over substance.”). “[P]rovid[ing] recommendations” to commercial clients, however, is not akin to “[g]athering business and/or technical requirements relative to use cases,” as these are different skills. Compare AR at 808 (TAB 21b²) (AOI Proposal) with AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). Given AOI RREP 1 fails to address this subtask, and the experience provided by AOI only involves “providing recommendations,” CBP properly eliminated AOI for this RREP subtask. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Second, CBP stated AOI did not address in its RREP: “[d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues it “discussed its approach to ‘data preparation, cleansing, and enrichment to integrate data from multiple sources and eliminate duplicates, empty fields, outliers, and data integrity issues prior to conducting any analysis using the data . . . [which] ensures that data in the Data Lake is of high quality.’” AOI MJAR at 18–19 (citing AR at 808 (TAB 21b²) (AOI Proposal)). The government responded at oral argument AOI’s RREP is only similar “in that it[ deals with] data,” but otherwise emphasized presenting experience with data preparation for a task involving ontology development is trying to “fit a square peg into a round hole.” Tr. at 202:8–18.

Data preparation is not the same as “[d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP.” AR at 10088 (TAB 71) (Track Third, CBP stated AOI did not address in its RREP “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As with the above subtask, CBP argues it “discussed its approach to ‘data preparation, cleansing, and enrichment to integrate data

2TET Consolidated Review). As discussed, supra, an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. The process of “‘data preparation, cleansing, and enrichment,’” AOI MJAR at 18–19 (citing AR at 808 (TAB 21b2) (AOI Proposal)), is different from creating a knowledge structure weaving data together. See Santosa et al. at 161815. At oral argument, AOI conceded the experience presented in its RREP was not “necessarily . . . the same” as the ontology-related work contemplated by this Task 3 subtask. Tr. at 201:10–12 (“[AOI:] [W]e don’t necessarily contend that it was the same. We’re simply contending that it was similar.”). RREP 1 does not address this subtask, so AOI was properly eliminated. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

from multiple sources and eliminate duplicates, empty fields, outliers, and data integrity issues prior to conducting any analysis using the data . . . [which] ensures that data in the Data Lake is of high quality.’” AOI MJAR at 20 (citing AR at 808 (TAB 21b²) (AOI Proposal)). AOI reasons its “experience in data preparation and cleansing is ‘similar’ to applying labeling ontology and annotating data of various types—both have the same function and result of ensuring the data is of ‘high quality’ for AI/ML purposes.” Id. The government responds CBP “could not determine, based on RREP 1, that AOI could meet the capabilities of Task 3 related to data labeling, annotation, and ontology development based on the [XXX] recommendation engine.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 68 (citing AR at 10088 (TAB 71) (Track 2 TET Consolidated Review)).

At oral argument, AOI conceded its RREP did not specifically address applying a labeling ontology or annotating data according to an ontology. Tr. at 206:15–18 (“THE COURT: Is there any discussion of labeling, annotation, or ontology? [AOI]: Again, Your Honor, I think not in those specific words.”). As above, AOI’s experience with “data preparation, cleansing, and enrichment,” AOI MJAR at 18–19 (citing AR at 808 (TAB 21b²) (AOI Proposal)), is unrelated to the field of ontology, specifically “[a]pplying labeling ontology . . . and annotat[ing] data.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). For the same reasons discussed, supra, RREP 1 fails. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Fourth, CBP stated AOI did not address in its RREP the subtask “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues its RREP 1 is similar because it “described . . . ingesting and normalizing . . . data from multiple sources, [and] maintaining a data lake of more than [XXXX] .” AOI MJAR at 22 (emphasis added) (citing AR at 808 (TAB 21b²) (AOI Proposal)). The government responds, “a data lake and a centralized library are not necessarily the same . . . [as a] data lake . . . can store ‘large amounts of structured, semistructured, and unstructured data’ . . . [as opposed to a] centralized library of annotated data for use . . . for AI model training.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 72 (emphasis added) (first citing AOI MJAR at 22; and then citing AR at 249 (TAB 10b⁵) (Solicitation)). At oral argument, the government argued a data lake is not necessarily the same as a centralized library of annotated data. Tr. at 208:14–209:1 (“[GOVERNMENT:] [F]or [a] data lake, . . . it stores data, sure, and the centralized library of annotated data also stores data. . . . [But t]here is no indication that the data here was annotated . . . or being used for AI model training.”).

While data “lakes” and “libraries” are similar in that both hold data, Task 3 requests experience with maintaining a library of annotated data. See AR at 10089 (TAB 71) (Track 2 TET Consolidated Review) (faulting AOI for lacking experience “managing a centralized library of annotated data”). RREP 1 does not describe experience managing annotated data stored in a centralized library. See AR at 808 (TAB 21b²) (AOI Proposal). AOI’s data lake is thus not equivalent to the experience required by this subtask. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). As such, RREP 1 fails on this subtask. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332. The Court notes it defers to CBP’s expertise in determining whether the RREP met the Solicitation’s technical requirements given the highly technical nature of this subtask. See E.W. Bliss, 77 F.3d at 449 (holding, “a court will not second guess [agency discretionary determinations]” when a protestor’s challenges “deal with the minutiae of the procurement process in such matters as technical ratings”).

Fifth, CBP stated AOI did not address in its RREP: “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues RREP 1 “described a ‘stringent authorization process, similar to an [ATO] process’ that it executed for [an] AI/ML product.” AOI MJAR at 24 (citing AR at 808 (TAB 21b²) (AOI Proposal)). The government responds AOI failed to address this subtask in its RREP. Gov’t’s Cross-MJAR and Resp. to Group 2 at 65.

Here, RREP 1 solely describes AOI’s “stringent authorization process, similar to an Authority to Operate (ATO) process, prior to being deployed into production.” AR at 808 (TAB 21b²) (AOI Proposal). The RREP lacks description of any other components of the subtask, such as “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes,” “coordination with . . . security teams,” or “identification of any Plan of Action and Milestones (POA&Ms).” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). This RREP is therefore not similar to the subtask. See id. As such, RREP 1 fails on this subtask. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Accordingly, CBP rationally evaluated each subtask in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated AOI for RREP 1. Garufi, 238 F.3d at 1332.

b. AOI’s RREP 2, Task 3

AOI argues CBP used an unstated evaluation criterion for Task 3 – Artificial Intelligence/Machine Learning (AI/ML), see supra Section I.A, in assessing its RREP 2 and determining scope and complexity. See AOI Cross-MJAR at 15. The relevant language of CMCI’s RREP 2 for Task 3, and CBP’s corresponding evaluation, is as follows:

Compare RREP 2. AR at 809–10 (TAB 21b²) (AOI Proposal).

With CBP’s Evaluation. AR at 10089 (Tab 71) (Track 2 TET Consolidated Review).

[XXX] is supporting [XXXXXX] with Artificial Intelligence/Machine Learning (AI/ML) Support Services and other emerging and advanced technologies such as data and predicative analytics.

[XXXXXXXXX] is a Global Leader in developing [XXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXX

RREP 2 Task 3: No – The vendor did not address the following requirement in the RFP scope for Task 3:

o Developing an approach to labeling,

identification of any Plan of Action and

Milestones (POA&Ms), and support to address them, as needed.

through data quality checkpoints built into our Agile story cards.

Compare AR at 809–10 (TAB 21b²) (AOI Proposal) with AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

First, CBP stated AOI did not address in its RREP 2: “[d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues RREP 2 discussed experience “‘gather[ing] and validat[ing] information’ and ‘design[ing] and develop[ing] various [data] filtering mechanisms.’” AOI MJAR at 19. The government reasons this RREP involved “metadata tagging,” which is not what Task 3 requested. Gov’t’s Cross-MJAR and Resp. to Group 2 at 68. The government argues AOI’s tagging is distinct from creating a tagging and categorization systems to “empower ML algorithms to . . . extract patterns from data.” Id. The RREP states:

[Contractor] [XXX] defined business cases, along with Key Performance Indicators (KPIs), gathered and validated information, and implemented a plan of action based on proven design and execution patterns and standards, tailored as needed for each data story (and prepared for using AI/ML). . . . [XXX] designed and developed various filtering mechanisms and analytic wrangling options to facilitate honing broad data into focused results and provided the ability to store search criteria and result snapshots for future use.

AR at 809 (TAB 21b²) (AOI Proposal). The RREP, while describing experience working and filtering data, does not discuss “[d]eveloping an approach to labeling, annotation, and ontology development to support [the] application of AI.” See AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As stated supra, an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Filtering data is distinct from creating a knowledge structure weaving data together. See id. RREP 2 does not address this subtask, so AOI was properly eliminated. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Second, CBP stated AOI did not address in its RREP: “[a]pplying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues its “RREP described experience in transferring data into ‘a centralized location’ and applying ‘advanced analytic algorithms’ to ‘identify and forecast trends and outcomes,’ which included designing and developing ‘various filtering mechanisms and analytic wrangling options’ to turn ‘broad data into focused results.’” AOI MJAR at 20. AOI reasons its “experience in this regard is ‘similar’ to applying labeling ontology and annotating data of various types because they serve the same function of ensuring the data is of ‘high quality’ for AI/ML purposes.” Id. at 20–21.

While AOI argues its experience and Task 3 have the same “end goal” of high quality data, AOI’s RREP seeks to achieve this goal differently than the approach requested by Task 3. Compare AR at 809 (TAB 21b²) (AOI Proposal) with AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI’s RREP describes experience using algorithms to narrow broad sets of data. AR at 809 (TAB 21b²) (AOI Proposal). The subtask at issue calls for labeling and annotating data according to a pre-existing ontology. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). As an ontology involves a sophisticated “knowledge structure,” AOI’s RREP “applying ‘advanced analytic algorithms’ to ‘identify and forecast trends and outcomes’” does not fit the subtask’s goal. AOI MJAR at 20; AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). RREP 2 does not address this subtask, so AOI was properly eliminated. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Third, CBP stated AOI did not address in its RREP: “[e]stablishing and managing a centralized library of annotated data for use across the organization for AI model training.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). AOI argues its RREP 2 is similar because it “demonstrated experience in transferring a large amount of data into a ‘centralized location.’” AOI MJAR at 22.

Here, simply having a “centralized location” is not the same as “[e]stablishing and managing a centralized library of annotated data for use across [an] organization for AI model training.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); AR at 809 (TAB 21b²) (AOI Proposal). The subtask requires the establishment and management of a library of annotated data. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review). AOI’s RREP, although mentioning centralized data storage, does not explain AOI’s experience with building and managing storage of annotated data for the purpose of training an AI system. AR at 809 (TAB 21b²) (AOI Proposal). This subtask therefore fails, meaning AOI was properly eliminated. Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332. Fourth, CBP stated AOI did not address in its RREP experience “[m]anaging and execut[ing] [] security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies, including coordination with CBP OIT security teams, identification of any Plan of Action and Milestones (POA&Ms), and support to address them, as needed.” AR at 10080 (TAB 71) (Track 2 TET Consolidated Review). AOI argues RREP 2 “explained the company’s implementation of a plan of action tailored to [individual] data stor[ies], and highlighted the regular support of cybersecurity needs through its AI/ML support services.” AOI MJAR at 24.

Here, the RREP lacks any overlap with the subtask language, which required “[m]anaging and execution of security and accreditation and Authority to Test (ATT) processes and Authority to Operate (ATO) processes for AI/ML technologies.” AR at 10089 (Tab 71) (Track 2 TET Consolidated Review). AOI fails to expand upon why this RREP experience is similar in its MJAR—it merely quotes the “company’s implementation of a plan of action tailored to [individual] data stor[ies], and highlight[s] the regular support of cybersecurity needs through its AI/ML support services.” See AOI MJAR at 24. Given the RREP’s lack of alignment with the subtask, RREP 2 fails, and AOI was properly eliminated for this subtask. AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345; Garufi, 238 F.3d at 1332.

Accordingly, CBP rationally evaluated each subtask in accordance with the Solicitation and did not apply “unstated criterion.” Banknote, 56 Fed. Cl. at 387, aff’d, 365 F.3d 1345. CBP rationally eliminated AOI for RREP 2. Garufi, 238 F.3d at 1332.

B. Whether CBP Failed to Document Its Reasoning or Ignored the Contents of Offerors’ Proposals

Ekagra and Constellation also argue CBP failed to adequately document its reasoning or ignored the contents of the offerors’ proposals. See Ekagra MJAR at 9; Constellation MJAR at 4–6. As discussed supra Section III.B, Unissant, AttainX, CMCI, and AOI agreed their arguments related to CBP’s lack of documentation are the same as their arguments related to unstated criterion, so the Court will not re-address those arguments here. See Tr. at 245:17–22; Tr. at 246:11–17; Tr. at 267:23–268:3; Tr. at 269:14–18.

At oral argument, the government asserted CBP’s “analysis [and explanation] can be much shorter if [an offeror is] way outside the expected range” of similarity in scope and complexity. Tr. at 54:10–18 (“THE COURT: [I]s the Government saying that if you’re way off the target, then less analysis is required? [GOVERNMENT]: Well, I don’t want to say less analysis, because the analysis is still looking as to how close you are, but you don’t have to state what’s obvious in your documentation of that analysis. So the analysis can be much shorter if you’re way outside the expected range.”). The Court and parties engaged with the car example introduced by Judge Hardiman in Obasi Invest. Ltd., see supra Section VI.A., to tease out why less agency explanation is permissible in circumstances where an offer is “way outside the expected range.” Tr. at 54:10–18; see Obasi Invest. Ltd. v. Tibet Pharms., 831 F.3d 179, 186 (3rd Cir. 2019). In Obasi, the Third Circuit recognized “[w]e might describe a ‘sedan’ as similar to a ‘truck’—both are vehicles . . . [b]ut an ordinary English speaker would not say a sedan is similar to a ‘light-duty pickup truck.’ The use of a narrowing term of art that distinguishes one class of trucks from others connotes a likeness . . . beyond basics like personal transportation.” Obasi Invest. Ltd., 831 F.3d at 186. In the instant case, analogizing RREPs to Obasi’s example, the government argued the agency need not extensively explain why, for example, a “sedan” or “vespa” offer would be inappropriate for a Solicitation seeking “trucks.” Tr. at 39:16–21 (“[GOVERNMENT]: So if you provided an experience that was deemed to be sufficiently dissimilar, CBP noted that. CBP is not required to provide the detailed analysis explaining why [it is not accepting your offer when] . . . you submitted a sedan.”). The agency could therefore provide a rational basis as to why an offer is not similar in scope and complexity even without a fulsome analysis, particularly where the quote submitted was not close to similar in scope and complexity with the relevant Task. Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”). The Court notes the agency’s explanation must still be rational. Id. at 1339.

1. Ekagra’s RREP 2, Task 1

Ekagra challenges CBP’s evaluation of its RREP 2 for Task 1 – Digital Transformation and Software Development Support, see supra Section I.A, by arguing CBP ignored Ekagra’s discussion of code development. Ekagra MJAR at 9. The relevant language of Ekagra’s RREP 2 for Task 1, and CBP’s corresponding evaluation, is as follows:

Compare RREP 2. AR at 4706–07 (TAB 34b¹) (Ekagra Proposal).

With CBP’s Evaluation. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

[XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

RREP 2 Task 1 - No - Although Ekagra does have some digital transformation experience through the usage of their scanning technologies for code as well as migration to the cloud.

However, it does not appear that they have proven experience with software development (writing code, developing algorithms, designing UI, etc[.]). A scanning tool with “eyes on code” capability does not equate to actual code development.

Scanning tool to examine the code for mistakes is different than actual coding.

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX

XXXXXXXXXXXXXXXXXXXXXX]

Compare AR at 4706–07 (TAB 34b¹) (Ekagra Proposal) with AR at 10089 (TAB 71) (Track 2 TET Consolidated Review).

CBP stated Ekagra did not address in its RREP “proven experience with software development (writing code, developing algorithms, designing UI, etc[.]).” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). Ekagra argues CBP “ignor[ed] all of Ekagra’s discussion of code development,” including its explanation of [XXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] Ekagra MJAR at 10–11 (citing AR at 4706 (TAB 34b¹) (Ekagra Proposal)). Ekagra explains [XXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX] Id. at 11 (citing AR at 4706 (TAB 34b¹) (Ekagra Proposal)). Ekagra also notes it [XXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXX] Id. Ekagra, finally, reasons its RREP described the following related to writing code, algorithms, and designing UI: [XXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXX] Id. (citing AR at 4706–07 (TAB 34b¹) (Ekagra Proposal)). The government responds, “RREP 2 does not describe an experience in which Ekagra developed software . . . [but r]ather, it describes Ekagra’s work [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXX] Gov’t’s Cross-MJAR and Resp. to Group 1 at 37 (citing AR at 4706-07 (TAB 34b¹) (Ekagra Proposal)). The government argues Ekagra’s example of [XXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXX] is not demonstrative of software development given it describes a [XXXXXXXX]. Id. at 38 (citing AR at 4706 (TAB 34b¹) (Ekagra Proposal)). The government also argues Ekagra’s example of [XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXX] reflects a [XXXXXX] not akin to developing code. Id. The government reasons: “[n]owhere in the RREP does Ekagra state that it had proven experience with software development (writing code, developing algorithms, designing UI, etc.).” Id.

Here, Ekagra categorizes its RREP 2 experience “[XXXXXXXXXXXXXXXXXXXX XXXXXX] which describes Ekagra’s [XXXXXXXXXXXX] throughout:

[XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX]

AR at 4706–07 (TAB 34b¹) (Ekagra Proposal) (emphasis added). This RREP does not specifically describe “proven experience with software development (writing code, developing algorithms, designing UI, etc[.]).” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). At oral argument, Ekagra conceded its RREP does not specifically discuss software development and argued CBP’s technical experts should infer the RREP encompasses such. See Tr. at 237:24–238:2 (“[EKAGRA:] Your Honor, if the Government is an expert in this field, as the technical experts are supposed to be, then they know what these terms mean, and these terms are all software development terms.”). In its review, CBP acknowledged Ekagra has [XXXXXXX XXXXXXXXXXXXXXX] through the usage of their [XXXXXXXXXX] for [XXXX] as well as [XXXXXXXXXXXXXXX] but clarified the RREP largely describes “[a] [XXXXXX] with [XXXXXXXXXXX] which “does not equate to actual code development.” AR at 10089 (TAB 71) (Track 2 TET Consolidated Review). CBP did not ignore the contents of Ekagra’s proposal; rather, CBP recognized the proposal to largely offer different experience than what Task 1 requested. See id. Given Ekagra’s RREP largely describes [XXXXXXXX], CBP provided a rational basis as to why Ekagra’s RREP is not similar in scope and complexity—it does not discuss writing code, developing algorithms, or designing user interfaces. AR at 10089 (TAB 71) (Track 2 TET Consolidated Review); Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

2. Constellation’s RREPs

a. Constellation’s RREP 1, Task 2

Constellation challenges¹⁸ CBP’s evaluation of its RREP 1 under Task 2 – Operational Maintenance Support Service, see supra Section I.A. Constellation MJAR at 4–5. The relevant language of Constellation’s RREP 1, for Task 2, and CBP’s corresponding evaluation is as follows:

18In addition to generally challenging CBP’s review of its RREPs, Constellation briefly makes an unstated criterion argument in its MJAR, see Constellation MJAR at 5 (“To the extent that CBP based its decision on something else,

Compare RREP 1. AR at 3471-72 (TAB 30b²) (Constellation’s Proposal).

With CBP’s Explanation. AR at 10569 (TAB 88e³) (CBP Evaluation of Constellation).

Constellation has operated and maintained the various software applications within CBP HCD IMSE for a period of several years and has been highly successful in maintaining smooth operations. . . . For RPA issues, Constellation responded to all user inquiries and requests within 2-3 hours and provided live support to troubleshoot and resolve information on issues. . . . The basis for Constellation’s operational maintenance (O&M) support to IMSE and HC capabilities is to facilitate smooth program coordination, control, and communications between the MSD and individual program offices. The project plan ensures effective and efficient resource utilization by monitoring current operations identifying and ensuring early identification and resolution of maintenance issues and obstacles; and completion of the program/project milestones, deliverables and outcomes is on-time and on-budget. For IMSE, our team ensured:

•Integrate O&M cost, schedule, and work accomplished and establish the basis for planning, budgeting, authorizing work, and measuring performance.

•Communicate how the O&M tasking will be performed to those participating in the project.

•Serve as the primary management and control tool for the project teams and MSD leadership.

•Allocate effort and attention appropriately to assure progress toward O&M project objectives, while remaining flexible to permit interpretation and revisions necessary to overcome potential obstacles.

•Ensure all milestones, schedules, and deliverables are consistent and provide direct correlation to the expectations of CBP within each project

•Ensure accurate milestone tracking and progress measurement.

•Facilitate risk mitigation and corrective action planning. Constellation realizes that O&M is integral to continuing RREP 1 Task 2: The RREP does not show an understanding of the O&M requirements in the scope of work.

then it was using an undisclosed evaluation criteria.”). To the extent Constellation makes an unstated criterion argument, the Court addresses these issues as part of Section VI.B.2.

to deliver mission needs ensuring access and use while reducing downtime and security risk. For IMSE, our approach was to also enhance efficiency and capability by considering improvements and innovations in processes, methodology, or tool sets employed to meet mission goals. Our support approach has been refined over the years in collaboration with OFO and brings together stakeholders to collectively develop effective enhancements and updates while maintaining critical systems and their access by stakeholders and users. This was done with little to no impact to current operations and reinforcing critical internal customer relationships.

Constellation considered the operations maintenance thresholds, deliverables, requirements, and mission objectives for each IMSE application to ensure that the operations maintenance plan is tailored specifically to the size, scope, complexity, risk, and security categorization of the effort. Our approach enabled MSD to build trust and credibility with OFO stakeholders who rely on technology-related solutions.

Compare AR at 3471–72 (TAB 30b²) (Constellation’s Proposal) with AR at 10569 (TAB 88e³) (CBP Evaluation of Constellation).

CBP stated Constellation’s RREP 1 does “not show an understanding of the O&M requirements in the scope of work.” AR at 10569 (TAB 88e³) (CBP Evaluation of Constellation). Constellation responds: “Constellation provided a detailed description of both its O&M methodology and tied that description to various previous projects.” Constellation MJAR at 5 (citing AR at 3471–83 (TAB 30b²) (Constellation’s Proposal)). The government responds “the focus of Task 2 was a vendor’s ability to provide O&M support services . . . [which] requires on-call remote support, and anticipatory problem detection and resolution[] for issues with newly developed software before they impact users,” and it is “not clear from Constellation’s RREP 1 that it has experience with incident management and defect resolution in this O&M context.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 41–42.

Constellation addresses O&M work throughout its RREP: The basis for Constellation’s operational maintenance (O&M) support to IMSE and HC capabilities is to facilitate smooth program coordination, control, and communications between the MSD and individual program offices. . . . For IMSE, our team ensured:

•Integrate O&M cost, schedule, and work accomplished . . .

•Communicate how the O&M tasking will be performed . . .

•Allocate effort and attention appropriately to assure progress toward O&M project objectives, . . . .

•Facilitate risk mitigation and corrective action planning. Constellation realizes that O&M is integral to continuing to deliver mission needs ensuring access and use while reducing downtime and security risk.

AR at 3471–72 (TAB 30b²) (Constellation’s Proposal) (emphasis added). At oral argument, the government agreed the RREP discusses O&M support: “I don’t know if [CBP was] confused during the evaluation, but I can’t deny that the text for O&M is here. It does certainly seem to say that it will communicate O&M tasking.” Tr. at 255:3–6. Though counsel for the government caveated his statement by noting his lack of technical expertise, the government conceded RREP 1 ultimately shows experience with the O&M requirements of Task 2. Id.; Tr. at 256:23–25 (“[GOVERNMENT]: I’m not a technical expert, so I can’t make that verification. It certainly does address some of the bulleted items.”). As such, CBP’s evaluation of Constellation’s RREP 1, Task 2 was arbitrary and capricious because CBP overlooked Constellation’s discussion of O&M tasking requirements in its RREP. Garufi, 238 F.3d at 1332; compare AR at 10569 (TAB 88e³) (CBP Evaluation of Constellation) (“The RREP does not show an understanding of the O&M requirements in the scope of work.”) with Tr. at 255:3–6. (“[GOVERNMENT:] I can’t deny that the text for O&M is here.”). CBP’s decision to eliminate Constellation for this specific subtask was arbitrary and capricious as “the procurement official’s decision lacked a rational basis” in failing to acknowledge the O&M tasking language provided in Constellation’s RREP. Garufi, 238 F.3d at 1332.

Per the Solicitation, a “quotation will be ineligible for award and will not be further evaluated” once CBP deemed the RREP not similar in scope and complexity. AR at 689 (TAB 18.2) (Solicitation). In terms of prejudice, CBP’s faulty reasoning for RREP 1, though arbitrary and capricious, therefore does not increase Constellation’s chance of award because Constellation’s quote fails for its RREPs under Task 3, see infra. Alfa Laval Separation, 175 F.3d at 1367 (citing Statistica, 102 F.3d at 1581) (“To prevail in a bid protest, a protestor must show a significant, prejudicial error in the procurement process. . . . To establish prejudice, a protestor . . . must show ‘that there was a substantial chance it would have received the contract award but for that error.’”). The government, in conceding Constellation’s RREP describes O&M tasking, made this argument at oral argument. Tr. at 257:2–5 (“[GOVERNMENT:] Constellation also had problems with their RREPs for Task 3. Losing on either would be enough to disqualify their quote.”). As such, Constellation cannot establish prejudice for its quote, despite CBP’s reasoning for RREP 1, Task 2 being arbitrary and capricious. Alfa Laval Separation, 175 F.3d at 1367 (citing Statistica, Inc., 102 F.3d at 1581).

b. RREPs 1 and 2, Task 3

Constellation also challenges CBP’s evaluation of its RREPs 1 and 2 under Task 3 – Artificial Intelligence/Machine Learning (AI/ML) Support Services, see supra Section I.A. Constellation MJAR at 5. When asked at oral argument why CBP’s evaluation of Constellation’s RREPs 1 and 2 for Task 3 overlapped, the government noted “the Agency . . . combined them because there’s overlap between what was missing from RREP 1 and RREP 2 for the machine learning/AI task . . . [but] [t]hat doesn’t suggest that the Agency somehow viewed these as one [RREP]. . . . [CBP] had over 100 quotes here, so I think . . . they were trying to be efficient with documentation.” Tr. at 258:8–17. The relevant language of Constellation’s RREPs 1 and 2 for Task 3, and CBP’s corresponding evaluation, is as follows:

Compare RREP 1. AR at

3485-86 (TAB 30b²)

(Constellation Proposal).

Compare RREP 2. AR at

3487-88 (TAB 30b²)

(Constellation Proposal).

With CBP’s Explanation.

AR at 10569-70 (TAB 88e³) (CBP Evaluation of Constellation’s Proposal).

Artificial

Intelligence/Machine

Learning (AI/ML) Support Services – Constellation personnel have strong experience with implementing AI/ML

models within CBP,

particularly for use in advanced data analytics and visualization.

Constellation has operated and maintained the various AI/ML-related software applications within CBP

HCD IMSE for a period of several

years and has been very successful in maintaining smooth operations. The

software applications we currently maintain include Microsoft

PowerApps/Power

Automate, SharePoint

(including customized features and configurations), and UiPath Robotic Process

Automation. We

coordinate with system owners for applications like COSS, SAP, and

SharePoint on a proactive basis to maintain awareness of changes. . . .

Gathering business and/or technical requirements relative to use cases identified by the Government for the evaluation, or trial, of potential AI/ML capabilities - IGS has established a workflow-based delivery approach that helps manage, monitor, and systematically fulfill work requests based on HHS needs and the critical underlying BIIS data. We use Kanban for BIIS to support optimizing work request flow and management.

We also use the Data

Management Maturity

Framework to help us meet, and sometimes exceed, task order requirements. This model helps us to continue to collaborate with stakeholders, to continually mature all work products, and to continue to establish best practices within each of the data maturity categories. We also provide technical project support by tracking performance of assigned development activities and managing workload and workload capacity. In addition, we provide data and user access to the operational data systems managed by OAPS including maintaining user accounts and defining and maintaining role-based access controls. We use tracking tools such as

RREP 1 & 2 Task 3: The

RREP does not describe any AI/ML related work. They provide examples of

(CSP), to host data and software for the purpose of developing and evaluating potential

AI/ML capabilities for the Government.

•Supporting ‘sandbox’

development, testing, and evaluation of software,

and to maintain in the innovation lab and cloud services platform.

Compare AR at 3485–86 (TAB 30b²) (Constellation Proposal) and AR at 3487–88 (TAB 30b²) (Constellation Proposal) with AR at 10569–70 (TAB 88e³) (CBP Evaluation of Constellation’s Proposal).

CBP stated Constellation did not address the following in its RREPs:

•The . . . develop[ment] [of] any prototype in AI/ML;

•Gathering business and/or technical requirements relative to use cases identified by the Government for the evaluation, or trial, of potential AI/ML capabilities;

•Developing an approach to labeling, annotation, and ontology development to support application of AI;

•Applying labeling ontology for CBP and annotat[ing] data of varying types from various sources according to the ontology;

•Establishing and managing a centralized library of annotated data for use across the organization for AI model training;

•Using various services and platforms, such as cloud service provider (CSP), to host data and software for the purpose of developing and evaluating potential AI/ML capabilities for the Government; and

•Supporting ‘sandbox’ development, testing, and evaluation of software, and to maintain in the innovation lab and cloud services platform.

AR at 10569–70 (TAB 88e³) (CBP Evaluation of Constellation’s Proposal).

For each of these seven missing subtasks for both RREPs 1 and 2, Constellation only argues it has a “detailed and comprehensive discussion of its AI/ML experience and expertise.” Constellation MJAR at 5. It does not point to specific language in each RREP to support the missing subtasks identified by CBP in its evaluation. Id. The government responds: “Neither RREP demonstrates that Constellation has the capability to develop or prototype any AI/ML model.” Gov’t’s Cross-MJAR and Resp. to Group 2 at 42. The government reasons “Constellation’s brief does not address, nor even acknowledge these omissions, stating only (without citation or elaboration) that its RREPs had addressed AI/ML.” Id. at 44.

At oral argument, the parties discussed Constellation’s RREPs in depth, given Constellation’s briefing lacked detail as to which portions of RREPs 1 and 2 met the Task requirements. For RREPs 1 and 2, the parties discussed RREP 2’s headings containing the relevant Task 3 “buzz words” but failing to substantively align with the Task 3 requirements. See Tr. at 255:25–266:11. For example, RREP 2 enumerates the following subtasks in bold— “[d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP” and “[a]pplying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.” See AR at 3487–88 (TAB 30b²) (Constellation Proposal). RREP 2 then states:

IGS works very closely with the Data and ETL teams to work on data mismatch issues that occur between the on-prem and SAP Cloud environments. Our team does the initial investigation by determining which tables and views are impacted prior to sending this data to the data and ETL teams for remediation. Our process has reduced the overall turn-around time from the time the issue was identified to resolution. The general cause of these data mismatches is usually related to the data loading process and can usually be resolved quickly once the issue has been identified and the root of the issue has been diagnosed.

AR at 3487–88 (TAB 30b²) (Constellation Proposal). The government argues Constellation listed the correct subtask “heading” in bold, but then describes irrelevant skills thereafter. For example, at oral argument, the government acknowledged the headings in Constellation’s RREPs match the task requirements but described the content of Constellation’s RREPs as proving the ability to make “scrambled eggs” when the Solicitation requested “pickup truck services.” See Tr. at 255:25–266:11; Tr. at 264:25–266:11 (“[GOVERNMENT]: But the specifics, Your Honor, are completely divorced from what the bold statement is. This is nothing about ontology development or data annotation or data labeling. They said we provide pickup truck services in bold, but then as explaining what that experience is, they talk about making scrambled eggs. It’s completely divorced from what that heading is. So this is my point about they say, ‘well, we have the right terminology.’ But they didn’t. They just put in a couple of buzzwords and then their experience is completely so far afield from what the Agency was looking for.”).

Looking at the specific language of both RREPs 1 and 2, neither RREP describes experience resembling the seven subtasks CBP identifies as missing. Compare AR at 3485–86 (TAB 30b²) (Constellation Proposal) and AR at 3487–88 (TAB 30b²) (Constellation Proposal) with AR at 10569–70 (TAB 88e³) (CBP Evaluation of Constellation’s Proposal). While both RREPs include the Task 3 subtasks as bolded titles, neither RREP provides actual experience with those subtasks. Id. Taking the above example, Constellation provides “work on data mismatch issues” in the data loading context in response to the subtasks of “[d]eveloping an approach to labeling, annotation, and ontology development to support application of AI at CBP” and “[a]pplying labeling ontology for CBP and annotate data of varying types from various sources according to the ontology.” Compare AR at 3487–88 (TAB 30b²) (Constellation Proposal) with AR at 10569–70 (TAB 88e³) (CBP Evaluation of Constellation’s Proposal). As stated supra, an ontology requires a sophisticated “knowledge structure,” consisting of a web of “terminologies or topics, their definitions and relational information within one or multiple domains.” Santosa et al. at 161815. Resolving “data mismatches” related to data loading is distinct from, and less sophisticated than, creating a knowledge structure by weaving data together. See id. Given Constellation’s RREPs describe experience outside the scope of Task 3’s requirements, CBP provided a rational basis as to why the RREPs are not similar in scope and complexity—RREPs 1 and 2 do not describe any AI/ML related work with any specify or accuracy. Garufi, 238 F.3d at 1332 (“[T]he procurement official’s decision [did not] lack[] a rational basis.”).

C. Whether CBP Engaged in Disparate Treatment

Plaintiffs AttainX, Unissant, and AOI bring disparate treatment claims regarding CBP’s evaluation of their RREPs. Specifically, AttainX alleges in evaluating RREPs for awardees NiyamIT and Tarkik, “the agency provided an exacting standard to AttainX, [but] . . . provided a less exacting standard for other entities.”¹⁹ AttainX MJAR at 19; see AttainX Reply at 8 (“AttainX [has] also demonstrated that the agency treated offerors disparately.”). Plaintiff Unissant likewise argues CBP’s “evaluation demonstrates disparate treatment because the Agency gave the awardees credit for capabilities simply because they used key words [in their RREPs], without evaluating the substance of what the awardees were proposing . . . [but] when the Agency evaluated Unissant’s proposal it determined that Unissant did not meet requirements because it did not . . . use [the same key words] even though it proposed the same capabilities.” Unissant MJAR at 24–25. Finally, plaintiff AOI asserts, with respect to at least four subtasks of Task 3, CBP “used different standards to evaluate [different] quotations” and thereby created a record “replete with stark examples of disparate treatment where the evaluators deployed two different grading scales.” AOI MJAR at 15–16. In discussing Task 3, which required offerors exhibit experience managing a “centralized data library,” AR at 249 (Tab 10b⁵) (Solicitation), AOI argues it experienced disparate treatment when its RREP, which referenced a “[d]ata [l]ake,” was deemed insufficient even though awardee Novilo’s Task 3 RREP, which discussed a “centralized library of annotated data,” was found to be sufficient by CBP. AOI MJAR at 20–21 (citing AR at 808 (Tab 21b²) (AOI’s Proposal); and AR at 7358 (Tab 41b²) (Novilo’s Proposal)).

19At oral argument, AttainX confirmed it withdrew the portion of its disparate treatment argument related to NiyamIT’s “proposal with respect to automating notification of issues requiring correction under Task 2.” Tr. at 294:6–10; AttainX Reply at 10 n.5. AttainX also withdrew its disparate treatment claims related to Novilo. See Tr. at 302:6–12.

In response to AttainX and Unissant, the government states, “[plaintiffs] cannot prevail on a claim of disparate treatment . . . because [they] cannot show that CBP ‘unreasonably downgraded [their] proposal[s] for deficiencies that were “substantively indistinguishable” or nearly identical from those contained in other proposals.’” Gov’t’s Group 1 Reply at 12, 27–28 (quoting Office Design Grp. v. United States, 951 F.3d 1366, 1373 (Fed. Cir. 2020)). Similarly, in responding to AOI, the government contends “AOI’s disparate treatment claim is really just a disagreement over the agency’s technical evaluation and conclusion[s].” Gov’t’s Cross-MJAR and Resp. to Group 2 at 72 (citations omitted). According to the government, “AOI misunderstands the quotes that the awardees submitted” and cannot satisfy the test set forth in Office Design because the awardees’ quotes are substantively different from AOI’s. Id. at 70. FAR 1.102-2 requires government agencies to treat “[a]ll contractors and prospective contractors . . . fairly and impartially,” but emphasizes they “need not be treated the same.” FAR 1.102-2(c)(3) (emphasis added); see also Frawner Corp. v. United States, 161 Fed. Cl. 420, 452 (2022) (noting under the FAR and Federal Circuit precedent, “a plaintiff must establish more than just varying treatment to meet its burden of showing disparate treatment.”). Thus, to succeed on a disparate treatment claim, see supra Section V.A.4, a “protestor must show that the agency unreasonably downgraded its proposal for deficiencies that were ‘substantively indistinguishable’ or nearly identical from those contained in other proposals” or “that the agency inconsistently applied objective solicitation requirements between it and other offerors, such as proposal page limits . . . or submission deadlines.” Office Design Grp., 951 F.3d at 1372 (quoting Enhanced Vets. Sols, Inc. v. United States, 131 Fed. Cl. 565, 588 (2017)). This test ensures reviewing courts do not have “free reign to second-guess the agency’s discretionary determinations,” id. at 1373, by limiting judicial review to situations in which a challenger contends an agency disparately treated “nearly identical provisions in the proposals under consideration.” Enhanced Veterans Sols., 131 Fed. Cl. at 588. Here, all three protestors argue CBP employed a different standard when evaluating their RREPs than when evaluating those of the awardees. See AttainX MJAR at 19; Unissant MJAR at 24–25; AOI MJAR at 15–16. Citing the rule from Office Design, the government replies, no plaintiff has overcome the “high burden, designed to separate the Court’s role from the agency’s,” which requires a showing CBP “unreasonably downgraded” plaintiffs’ proposals “for deficiencies that were ‘substantively indistinguishable’ or nearly identical from those contained in other proposals.” See Gov’t’s Cross-MJAR and Resp. to Group 2 at 70 (quoting Office Design Grp., 951 F.3d at 1373).

Plaintiffs acknowledged at oral argument “trying to apply the Office Design standard to this set of facts is difficult,” Tr. at 303:3–17 (Unissant), because each bidder’s proposal contains unique RREPs different in substance and kind than those presented by other offerors. The government similarly stated at oral argument each offeror’s RREPs are “so far afield” from those of other offerors application of the Office Design test is impractical. See Tr. at 309:14–310:16. Both in briefing and at oral argument, plaintiffs acknowledged their proposals differ in meaningful ways from those they allege are “substantively indistinguishable” or nearly identical. Office Design Grp., 951 F.3d at 1372 (internal quotations omitted). For instance, despite arguing CBP engaged in disparate treatment when it “stated Unissant did not address the . . . [Task 2] requirement” of performing 24x7x365 on-call remote support because “Unissant did not mention 24x7x365 support” in its Task 2 RREPs, Unissant recognizes the “awardees[’] [offers] . . . [explicitly] mentioned ‘24x7x365 support’ [in their Task 2 RREPs].” Unissant MJAR at 24–25. AttainX makes a similar concession. Indeed, although contending at oral argument “[t]he focus on 24/7/365 . . . is misplaced,” Tr. at 298:11–13, AttainX acknowledges such support experience is explicitly discussed in the awardees’ proposals but “is not in [AttainX’s] proposal.” Tr. at 298:16–299:2. AOI likewise alleges it experienced disparate treatment without adequately showing its RREPs were “substantively indistinguishable” from those of the awardees discussed in its MJAR. Office Design Grp., 951 F.3d at 1372. To the contrary, AOI acknowledges its RREPs in issue were substantively different from those of the awardees. Tr. at 309:7–12 (“[AOI:] So we are not necessarily looking to whether there’s substantively indistinguishable components between these experiences, which by their nature have to be different because they are different experiences, but whether the Agency conducted the evaluation fairly and evenhandedly.”); see AOI MJAR at 15–27. Thus, although Unissant, AttainX, and AOI largely contend CBP disparately treated “substantively indistinguishable” proposals, Office Design Grp., 951 F.3d at 1372 (internal quotations omitted), they admit their RREPs were substantively different than those of the awardees discussed in their MJARs.

As expressly stated by the Federal Circuit, a court reviewing an agency’s discretionary determinations does not have “free reign to second guess” agency decisions, including when protestors allege disparate treatment. Office Design Grp., 951 F.3d at 1373. Rather, the Court is permitted to intervene only when a plaintiff succeeds in showing “the agency unreasonably downgraded its proposal for deficiencies that were ‘substantively indistinguishable’ or nearly identical from those contained in other proposals.” Id. at 1372. As explained, supra, all three plaintiffs argue their proposals were treated disparately from those of awardees, but in doing so, they expressly recognize differences between their RREPs and those they must show are “nearly identical” to theirs in order to prevail on their claims. Id. The Court will not “interfere with the agency’s broad discretion,” id. at 1373, by substituting its judgment regarding whether individual RREPs meet the solicitation’s requirements for CBP’s. See supra. Rather, because plaintiffs are unable to show their RREPs at issue are “substantively indistinguishable” from those of the awardees, “the [C]ourt . . . dismiss[es] the[se] claim[s].” Office Design Grp., 951 F.3d at 1372– 73; see also FAR 1.102-2(c)(3).

D. Whether CBP’s Decision Prejudiced CMCI and Constellation

As noted, supra Sections VI.A.4 and VI.B.2.a, CBP acted arbitrarily and capriciously in: (1) finding CMCI’s RREP 1 for Task 3 did not address the subtask related to “managing a centralized library of annotated data,” AR at 10088 (TAB 71) (Track 2 TET Consolidated Review); (2) failing to adequately explain its reason for deeming CMCI’s Task 5 RREP 2 insufficient, id.; and (3) finding Constellation’s RREP 1, Task 3 did “not show an understanding of the O&M requirements in the scope of work,” AR at 10569 (TAB 88e³) (CBP Evaluation of Constellation). The Solicitation stated “[i]f the Government determines any of the required two (2) RREPs are not similar in scope and complexity, the quotation will be ineligible for award and will not be further evaluated,” see supra Section I.A. AR at 689 (TAB 18.2) (Solicitation); see supra Tr. at 228:15–22 (“[THE COURT:] [W]hen an offeror submitted an RREP that is not similar in scope and complexity, CBP would eliminate the offeror from award. Is it from award for the entire contract or just that a specific task? . . . [THE GOVERNMENT]: From a full track. . . .”); Tr. at 19:8–14 (THE COURT: . . . [W]hat do you think it means, “The quotation will be ineligible for award and will not be further evaluated?” That allows the Government still some discretionary wiggle room? [PLAINTIFF SPOKESPERSON UNISSANT]: The words mean what they say. They will be ineligible for award.”). CBP’s reasoning for CMCI’s and Constellation’s other RREP submissions was not arbitrary and capricious, see supra Sections VI.A.4.a and VI.B.2, and CBP therefore properly eliminated them.

CMCI and Constellation do not have a substantial chance of award because of their respective dissimilar RREPs under Task 3 and Task 5, see supra Sections VI.A.4.a and VI.B.2. CMCI and Constellation have therefore failed to “show a significant, prejudicial error in the procurement process.” Alfa Laval Separation, 175 F.3d at 1367 (“To prevail in a bid protest, a protestor must show a significant, prejudicial error in the procurement process. . . . To establish prejudice, a protestor . . . must show ‘that there was a substantial chance it would have received the contract award but for that error.’”); Bannum, 404 F.3d at 1353.

E. Whether the Court Should Grant An Injunction²⁰

Track 1 and Track 2 plaintiffs both argue they are entitled to injunctive relief. See, e.g., LSI MJAR at 8–10; Constellation MJAR at 6–7; Arch MJAR at 33–40; GTS MJAR at 32– 33; AttainX MJAR at 10–28; Unissant Reply at 15–18; Ekagra MJAR at 21–23; AOI MJAR at 29–32; CMCI MJAR at 10. Track 1 plaintiffs explicitly allege: (1) they “demonstrated actual success on the merits,” see, e.g., Arch MJAR at 36; (2) the “significant irreparable injury in the form of lost profits, lost critical contracting opportunities, and the loss of . . . rights to a fair competition” tip the balance of hardships in their favor, id. at 38; (3) and the “strong public interest in maintaining the integrity of the procurement process” entitles them to injunctive relief, id. at 38–39 (citations omitted). Track 2 plaintiffs similarly allege they are “entitled to injunctive relief,” AttainX MJAR at 10–28. See also Unissant Reply at 15–18 (“[T]he injunctive factors weigh heavily in favor of granting a permanent injunction.”); Ekagra MJAR at 21–23; AOI MJAR at 29–32; CMCI MJAR at 10.

20At oral argument, the parties addressed how an injunction would impact the contract roll-out timeline and stated the CBP must award the contract all at once, treating the awardees as a “group,” and could not break up or stagger awards in the case of injunction. See Tr. at 341:21–23 (“THE COURT: So they would all start out as a fixed group? [GOVERNMENT]: I think for the two tracks, yes.”).

The Court considers the following factors when determining whether to issue a permanent injunction: “(1) whether . . . the plaintiff has succeeded on the merits of the case; (2) whether the plaintiff will suffer irreparable harm if the court withholds injunctive relief; (3) whether the balance of hardships to the respective parties favors the grant of injunctive relief; and (4) whether it is in the public interest to grant injunctive relief.” PGBA, 389 F.3d at 1228– 29. Plaintiffs here are not entitled to injunctive relief because they do not prevail on the merits. See supra Sections V–VI. The Court therefore need not consider the remaining three permanent injunction factors. Info. Tech., 51 Fed. Cl. at 357 n.32 (“Absent success on the merits, the other factors are irrelevant.”), aff’d, 316 F.3d 1312.

For the foregoing reasons, the Court now (1) DENIES Ekagra’s Motion for Judgment on the Administrative Record, ECF No. 74; (2) FINDS AS MOOT Ekagra’s Motion for Preliminary Injunction, see supra FN 1, ECF No. 2; (3) DENIES plaintiff Unissant’s Motion for Judgment on the Administrative Record, ECF No. 77; (4) DENIES AttainX’s Motion for Judgment on the Administrative Record, ECF No. 68; (5) DENIES LSI’s Motion for Judgment on the Administrative Record, ECF No. 80; (6) DENIES Arch’s Motion for Judgment on the Administrative Record, ECF No. 84; (7) DENIES AOI’s Motion for Judgment on the Administrative Record, ECF No. 81; (8) FINDS AS MOOT AOI’s Motion to Strike, ECF No. 112, see supra Section VI.A.5.; (9) DENIES GTS’ Motion for Judgment on the Administrative Record, ECF No. 82; (10) DENIES Constellation’s Motion for Judgment on the Administrative Record, ECF No. 83; (11) DENIES CMCI’s Motion for Judgment on the Administrative Record, ECF No. 75; (12) GRANTS the government’s Cross-Motions for Judgment on the Administrative Record, ECF Nos. 93 & 106; (13) GRANTS Novilo’s Cross-Motions for Judgment on the Administrative Record, ECF Nos. 92 & 104; (14) GRANTS CAN Softtech’s Cross-Motion for Judgment on the Administrative Record, ECF No. 91; (15) GRANTS NiyamIT’s Cross-Motions for Judgment on the Administrative Record, ECF Nos. 90 & 101; (16) GRANTS Catalina’s Cross-motion for Judgment on the Administrative Record, ECF No. 105; (17) GRANTS Chevo’s Cross-Motion for Judgment on the Administrative Record, ECF No. 103; (18) FINDS AS MOOT Chevo’s Motion to Dismiss, ECF No. 103, supra n.6; and (19) DENIES Catalina’s Motion for Leave to File Supplemental Briefing, ECF No. 165, supra n.14. The Clerk is directed to enter judgment accordingly.

IT IS SO ORDERED.

s/ Ryan T. Holte

RYAN T. HOLTE

Judge