Once a year GAO tells Congress how the bid protest system did, and once a year the coverage picks out the sustain rate and calls it the headline. This year that number went down, from 16 percent to 14, and the easy story writes itself: harder to win, agencies doing better, protests less worth filing.
The easy story is wrong, and the report contains the number that proves it.
What the report says
The decision document is GAO-26-900695, the bid protest annual report to Congress for fiscal year 2025. The figures:
| Measure | FY2025 | FY2024 |
|---|---|---|
| Cases filed | 1,688 | 1,803 |
| Protests | 1,617 | |
| Cost claims | 24 | |
| Requests for reconsideration | 47 | |
| Cases closed | 1,737 | 1,706 |
| Merit decisions | 380 | 386 |
| Protests sustained | 53 | 61 |
| Sustain rate | 14 percent | 16 percent |
| Effectiveness rate | 52 percent | 52 percent |
| Hearings held | 3 cases, 0.5 percent | 1 case, 0.2 percent |
| ADR cases | 53, 91 percent successful | 76, 92 percent successful |
The two rates measure completely different things
This is the part that gets flattened in most write ups, so it is worth being precise.
The sustain rate is sustains divided by merit decisions. It only counts cases GAO actually decided on the merits. Fifty three out of 380.
The effectiveness rate is the share of closed protests in which the protester obtained some form of relief. That includes a sustain, and it also includes every case where the agency looked at the protest and voluntarily took corrective action rather than defend it.
So the sustain rate answers "if GAO rules, how often does the protester win." The effectiveness rate answers "if I file, how often do I get something."
Do the arithmetic
Fifty two percent of 1,737 closed cases is roughly 900 protesters who got relief. Fifty three of them got it from a GAO sustain. The other 847 or so, about sixteen times as many, got it because the agency moved first.
Round it however you like and the shape holds: somewhere around 94 percent of all successful protests in fiscal 2025 never reached a GAO decision at all. The agency folded first.
That is not a rounding artifact and it is not new. It is what the system has looked like for years. What is worth noticing this year is that the two rates moved differently. Merit wins got harder, from 61 to 53 and from 16 percent to 14. The overall chance of getting relief did not move a point.
What that combination means
If the sustain rate falls while the effectiveness rate holds, the cases that stopped being winnable at GAO were replaced, one for one, by cases that resolved before GAO had to look at them.
There are two readings and both are worth stating.
The generous reading. Agencies are getting faster at recognising a good protest and fixing the procurement without spending four months defending it. Fewer merit decisions is what a maturing system looks like, and the protester still gets a re-evaluation.
The uncomfortable reading. A voluntary corrective action produces no published decision, no reasoning, and no precedent. When 900 protests resolve and only 380 generate any written analysis, the body of law that everybody in this field navigates by is being built from a shrinking fraction of the actual disputes. The other side of every corrective action is an awardee who loses the work with nothing on the record explaining why.
The hearings number is the other tell
GAO held hearings in 3 cases, which is 0.5 percent of activity. It was 1 case and 0.2 percent the year before, so it tripled, and it is still statistically nothing.
Nearly every federal bid protest in the United States is decided on a paper record, without anybody being questioned. That is by design and it is defensible on speed grounds, because GAO works to a 100 day clock. It is also worth remembering the next time a protest turns on what somebody at an evaluation board actually thought.
Where the sustains came from
The most common grounds for a sustain, in order:
One. Unreasonable technical evaluation. Two. Unreasonable cost or price evaluation. Three. Unreasonable rejection of proposal.
All three are the same complaint wearing different clothes: the agency reached a conclusion the record does not support. None of them is about a rule being broken. They are about work not being shown.
That matters for anyone drafting a protest, because it says the winning ground is almost never "the agency was not allowed to do that" and almost always "the agency did not explain why it did that."
The filing number
Cases filed fell 6 percent, from 1,803 to 1,688. Closed cases went up slightly, from 1,706 to 1,737, which is what you would expect from a docket working through a backlog rather than a system anybody has walked away from.
A 6 percent drop in a single year is not a trend. It is one year.
What to take from it
If you are deciding whether to protest, the sustain rate is the wrong number to look at. Fourteen percent describes the subset of protests that survive to a decision, and most protests do not go that far in either direction. Fifty two percent is the number that describes filing.
If you are an awardee, the same figures read differently. Roughly half of protests against an award produce some form of relief for the protester, and the overwhelming majority of that happens without any tribunal finding that anything was wrong.
If you are the agency, the three sustain grounds are a checklist. Every one of them is fixed at the evaluation stage by writing down the reasoning, not at the protest stage by defending it.
The headline this year was that it got harder to win a protest. It did not. It got harder to win one in the way that generates a public decision, which is not the same thing at all.