Two Companies Protested The Same 640 Million Dollar Award, Lost The Same Argument, And Only One Of Them Won

One solicitation, worth more than 640 million dollars, produced two decisions on September 10. The company that argued its technical rating was wrong got nothing. The company that argued the agency never wrote down why it was excluded got a sustain and its attorneys fees. Same procurement, same evaluators, same factor.

LOLSBA · Federal contracting desk · September 10, 2026

The Government Accountability Office headquarters at 441 G Street NW in Washington, D.C., which decided both RIVER RFQ protests. Photo: AgnosticPreachersKid, CC BY-SA 3.0, via Wikimedia Commons

On September 10 GAO published two bid protest decisions out of the same procurement. Same agency, same solicitation, same evaluation factor, same rating handed to both protesters. One decision came back denied in part and dismissed in part. The other came back sustained in part.

Reading them next to each other is the cleanest lesson in protest strategy anybody is going to get this month, because almost every variable is held constant.

The procurement

The solicitation is RIVER RFQ No. 70SBUR24Q00000106, issued by the Department of Homeland Security for United States Citizenship and Immigration Services. It is a small business set aside for IT professional services on the USCIS benefits portfolio, and across its task orders and option periods it is worth more than 640 million dollars.

USCIS awarded task orders to Peregrine Digital Services and BridgePhase, with a third possible award still open. Two of the unsuccessful vendors protested.

FusionEdge Solutions, LLCRed Cedar Harmonia, LLC
File numbersB-423829.2, B-423829.7B-423829.3, B-423829.5
HomeHerndon, VirginiaLeesburg, Virginia
Rating on the code challenge factorSome confidenceSome confidence
Attacked its own ratingYesYes
Result on that groundDeniedDenied
Other groundsScope of the corrective actionAwardee evaluations, price realism
Result on thoseSustainedDismissed
Costs recoveredYes, including attorneys feesNone

The argument both of them lost

Both companies were rated some confidence rather than high confidence under the code challenge factor, which is the part of this competition where vendors actually build something and get watched doing it. Both argued the rating was unreasonable. Neither got anywhere.

The specific findings that sank them are worth quoting because they are so ordinary. FusionEdge was marked down because its visual collaboration board "did not provide sufficient traceability" to the user stories, and because it "failed to deploy the major of their functionality to production" when the exercise called for it. Red Cedar Harmonia was marked down across seven separate areas, including a failure to engage the product owner during the demonstration, which GAO held was "consistent with the terms of the solicitation" because the performance work statement told contractors to coordinate with product owners on user centered design.

That is the whole of it. Two vendors, two different sets of technical criticisms, one identical outcome: GAO looked at the evaluators' documented findings, found them supported by the record and consistent with the solicitation, and declined to substitute its own judgment.

The argument one of them won

FusionEdge's second ground had nothing to do with its own score.

USCIS had taken corrective action and announced that when it went back to consider a third task order award, it would look only at LightFeather IO. FusionEdge would not be considered at all. FusionEdge protested the scope of that corrective action, and GAO sustained it.

The reasoning is about paperwork rather than merit. The bases for limiting the corrective action were undocumented. The original source selection decision contained no finding that a third task order would not go to FusionEdge. The RFQ obliged the agency to "fairly consider all Quotes." And there was no documented comparison anywhere in the record between the LightFeather and FusionEdge quotations.

GAO's holding: "This is plainly unreasonable and fails to meet the agency's obligation to conduct and document a proper best-value tradeoff analysis."

The remedy is that when USCIS makes a source selection decision for the third task order, it has to be reasonable and consistent with the solicitation, which in practice means FusionEdge has to be in the room. GAO also recommended that FusionEdge recover the costs of filing and pursuing the supplemental protest, including reasonable attorneys fees, under 4 C.F.R. section 21.8(d)(1), with a certified claim due inside 60 days.

The pattern this fits

Nothing in either decision is novel law. What makes the pair useful is that they isolate the variable.

An attack on a technical rating asks GAO to disagree with evaluators who watched a live demonstration and wrote down what they saw. GAO does not do that, and it says so in almost identical language in both decisions. The record either supports the rating or it does not, and a documented record almost always supports it.

An attack on documentation asks a completely different question. It does not ask whether the agency reached the right answer. It asks whether the agency can show its work. When the answer is no, as it was on the scope of this corrective action, the protester wins without ever establishing that its own quotation was better.

Red Cedar Harmonia's other grounds were dismissed rather than denied, which is a distinction worth understanding. Its challenges to the awardees' evaluations and its price realism claim were thrown out for lack of a sufficient factual basis, meaning GAO never reached the substance. On price realism GAO restated the standing rule: "absent a solicitation provision advising offerors that the agency intends to conduct a price realism analysis, agencies are neither required nor permitted to perform such analysis." The RFQ did not have one, so the claim had nowhere to go.

What to take from it

One. A some confidence rating on a hands on evaluation factor is close to unprotestable if the evaluators wrote down what they observed. Both of these companies tried and both were denied on the record as it stood.

Two. Corrective action is a live protest ground, not a surrender. An agency that narrows the field during corrective action without documenting why has created a fresh, winnable issue, and this decision is the template for it.

Three. Price realism is not available on request. It exists only if the solicitation says it does, and GAO will not read one into a document that lacks it.

Four. The costs recommendation matters more than it looks. Reasonable attorneys fees on a supplemental protest of a 640 million dollar procurement is not a symbolic award, and it went to the protester who found the documentation gap rather than the one who argued about its own score.

Two companies, one solicitation, the same rating and the same losing argument. The difference between them was which second ground they picked.