Small Business Grant Approval Rates: What Is Actually Published

The short answer: almost no small business grant program publishes an approval rate

Reviewed August 28, 2026 — every external figure below was opened at its source and re-checked on that date, and the source URL is printed next to it.

Approval rates that US funders actually publish

Fifteen small business funding rates are published by the funder itself, or computable from two counts the funder published for the same round. They cluster in one place: federal research grants, where the Paperwork Reduction Act, congressional reporting duties and long-standing statistical offices force agencies to count applications as well as awards. The SBIR and STTR programs account for nine of the fifteen. The full range runs from 8% (NIH SBIR Phase I and Fast Track, fiscal 2025) to 86% (CDFI Program Technical Assistance awards, fiscal 2024) — a spread that reflects how differently these programs are designed, not how differently applicants perform.

The published-rate table: fifteen figures with a named source and a date

Every row below names the funder, the exact program, the period the rate covers, and a link to the document it came from. Rows marked computed are a division of two counts the same funder published for the same round; the funder did not print a percentage.

NIH publishes the deepest small business grant success-rate series in the US government

The NIH Data Book carries an SBIR success-rate series running from fiscal 1998 to fiscal 2025, updated March 2026, with the application count and the award count printed beside every rate. It is the deepest series of its kind in US government data: nearly three decades of one small business grant program's odds moving year by year. NIH SBIR Phase I applications rose 48% between fiscal 2023 and fiscal 2025 (3,014 to 4,468) while awards fell 26% (472 to 349), and the success rate halved from 16% to 8%. Put another way, NIH received 6.4 SBIR Phase I applications for every award it made in fiscal 2023 and 12.8 for every award in fiscal 2025 — the price of an award, in competing applications, exactly doubled in two years. Read the two columns together: the odds moved because more companies applied into a shrinking pool of awards, not because the review got stricter.

NIH SBIR Phase I odds across 28 years: 30% in 2001, 8% in 2025

The NIH SBIR series begins in fiscal 1998 and is the longest published record of a US small business grant program's odds. Phase I peaked at 30% in fiscal 2001, fell to 20% by fiscal 2004, bottomed at 11% in fiscal 2011, recovered to 19% by fiscal 2019, and reached its published low of 8% in fiscal 2025. The whole-SBIR figure, combining Phase I, Phase II and Fast Track, tracked the same path from 34% to 10%. Anyone quoting a 20% SBIR success rate is quoting a real NIH number from fiscal 2004.

NIH SBIR Phase II and Fast Track fell alongside Phase I

NIH publishes separate series for SBIR Phase II and for Fast Track, the combined Phase I/II application route. Both moved in the same direction as Phase I between fiscal 2023 and fiscal 2025. Phase II remains the friendliest NIH small business route at 15% in fiscal 2025, because its applicant pool is largely restricted to companies that already completed a Phase I project. Fast Track collapsed hardest, from 18% to 8%, and now sits level with Phase I — it no longer carries the odds premium over Phase I that it held in fiscal 2023 and fiscal 2024. The SBIR and STTR guide covers how the phases and Fast Track differ.

The SBA's government-wide SBIR and STTR selection rates stop at fiscal 2022

The Small Business Administration is the only body that reports a selection rate across all eleven federal SBIR agencies at once, in its annual SBIR/STTR report to Congress. The most recent one published on sbir.gov as of August 28, 2026 covers fiscal 2022 — four years stale, and a fact worth knowing before you quote a “current” government-wide SBIR rate. In fiscal 2022, agencies selected 3,859 Phase I SBIR and STTR awards from 21,911 Phase I proposals, a 17.6% program-wide Phase I selection rate, and 39% of Phase I winners were first-time winners.

SBIR Phase I selection rates ranged from 6% to 22% across eleven agencies in fiscal 2022

The single most useful thing in the SBA report is that it breaks the government-wide rate apart by agency, which turns “what are my odds at SBIR” into a question with eleven different answers. In fiscal 2022 the Department of Homeland Security selected 22% of SBIR Phase I proposals and the Department of Education selected 6% — a 3.7× difference between two federal agencies running the same statutory program in the same year. Across the eleven agencies the median Phase I selection rate was 16% and the spread was 16 percentage points. Phase II inverts several of these: the Department of Transportation selected 89% of SBIR Phase II proposals in fiscal 2022 while the Department of Defense selected 30%.

NSF publishes a range for SBIR and a hard number for everything else

The National Science Foundation declines to publish a single SBIR figure. Its own applicant page says: “We can't fund all of the proposals we receive. Historical Phase I funding rates for NSF SBIR/STTR Phase I proposal have been between 10% and 20%.” That range is measured on invited proposals: NSF requires a Project Pitch first, and a full Phase I proposal may only be submitted “if you are invited.” NSF publishes no invitation rate for Project Pitches on its Seed Fund applicant pages, so the true rate from first contact to award is unknown and lower. For its whole portfolio NSF is precise: in fiscal 2023 it evaluated 38,340 competitive proposals and made 11,056 new awards, an overall funding rate of 29%. That was the highest rate in the ten fiscal years the digest tabulates, up from 23% in fiscal 2014, and 27% for research proposals specifically.

The CDFI Fund publishes application demand months before it announces awards

The Treasury's Community Development Financial Institutions Fund does something almost no other US funder does: it issues a press release stating how many organizations applied and how much they requested, then a second release naming the winners. For the fiscal 2024 CDFI Program and NACA Program round, 948 organizations requested roughly $1.55 billion — “more than three times the amount of available funding” — and the CDFI Fund called it “the largest number of organizations ever to apply to a single round.” The fiscal 2024 round ultimately awarded $408.2 million in Financial Assistance and $76.32 million in Technical Assistance, roughly $485 million against $1.55 billion requested — a 3.2× oversubscription in dollars. Because both counts are public and cover the same round, the rates below are arithmetic rather than estimate. CDFI awards flow to community lenders, who then lend to small businesses; the CDFI Program page explains that chain.

California published the largest small business grant denominator in US history

State programs occasionally publish both halves, usually when the program is large enough to be politically visible. The California Office of the Small Business Advocate reported that in the first two rounds of the California Small Business COVID-19 Relief Grant Program, “the Program selected just over 40,000 small businesses and non-profits to receive approximately $500 million in grant funding” while “more than 350,000 small businesses and non-profits applied in a competitive process, with funding requests totaling more than $4.5 billion.” That puts the round-1–2 award rate under 12%, and because the applicant figure is a floor (“more than 350,000”) the true rate is lower still. In dollar terms the round was oversubscribed 9×: more than $4.5 billion requested against approximately $500 million awarded in 2021. It remains the largest published applicant count for any US small business grant program we have found.

“Success rate,” “funding rate” and “selection rate” are three different denominators

The three US agencies that publish small business grant odds each use a different word for the measure, and each word means a different denominator. Comparing an NIH success rate to an SBA selection rate without adjusting is the most common technical error in grant-odds writing. NIH divides by reviewed applications and removes withdrawals. NSF divides by evaluated proposals and removes whole categories of submission. The SBA divides by proposals received and removes nothing. Applied to the same underlying pile of paper, the NIH and NSF definitions produce a higher rate than the SBA definition, because both shrink the denominator. Worked through: take 1,000 submissions of which 100 are withdrawn or returned before review and 90 are ultimately funded. NIH’s success rate is 10% (90 ÷ 900 reviewed). The SBA’s selection rate is 9% (90 ÷ 1,000 received). Same program, same year, same awards, and an 11% relative difference created entirely by the definition.

NIH counts reviewed applications and removes withdrawals from the denominator

The National Institutes of Health defines its measure in a standing methodology note from the Division of Statistical Analysis and Reporting: “Success rates are defined as the percentage of reviewed grant applications that receive funding (i.e., the likelihood of a project getting funded).” The denominator is the number of competing applications reviewed plus funded carryovers, and NIH states plainly what leaves the pile: “Excluded from the calculation of success rates are those applications that are withdrawn by an applicant prior to review, returned or administratively withdrawn by the NIH.” That exclusion matters for small businesses in particular, because a returned application — the usual outcome of a formatting or registration failure — never enters the denominator at all. NIH's published 8% for SBIR Phase I in fiscal 2025 is therefore an 8% chance given that your application was reviewed.

NSF counts evaluated proposals and states the arithmetic in its own report

The National Science Foundation defines its measure inside the Merit Review Digest itself: “Funding rate refers to the proportion of evaluated proposals that were awarded in a fiscal year. For example, if a directorate or office evaluated 8,000 proposals in the year, making 2,000 awards and declining the remaining 6,000, the funding rate for that directorate or office in that year would be 25%.” NSF explicitly removes preliminary proposals, contracts, Intergovernmental Personnel Act agreements, continuing grant increments, supplemental funding requests and Graduate Research Fellowship applications from the count. For SBIR applicants, the important consequence is that the Project Pitch stage sits outside this measure entirely, so NSF's 10–20% is not the odds of a company that starts the process — it is the odds of a company NSF already invited.

The SBA counts every proposal received, with no exclusions

The Small Business Administration uses the bluntest of the three definitions, printed in its SBIR/STTR annual report: “Proposal selection rates are the number of awards made divided by the total number of proposals received with the fiscal year.” No category of submission is removed. This is why the SBA's fiscal 2022 government-wide SBIR Phase I selection rate of 16% and the NIH fiscal 2022 SBIR Phase I success rate of 15% are close but not identical measures of overlapping populations. The SBA adds one honest caveat that most secondary write-ups drop: “the selection rate is an estimate. For FY22 awards, the proposals received were from both FY21 and FY22” — the numerator and denominator are not drawn from exactly the same cohort. In a year when proposal volume is rising, as fiscal 2022 was, that lag flatters the rate slightly.

Why you cannot find an approval rate for most small business grant programs

The absence is structural, not a gap in your research. A US funder has a strong legal and practical duty to publish who received money and a near-zero duty to publish how many people asked. Federal award data is pushed into USAspending.gov by statute; applicant counts have no such pipeline. Foundations report grants paid on IRS Form 990; the form has no line for applications received. Corporate contests treat entry volume as a marketing asset, published when it flatters and withheld when it does not. So the default state of a US grant program is: award count public, applicant count private, approval rate unknowable from outside.

The federal grant announcement format has no field for applications received

This is checkable in about a minute. Every federal funding opportunity on Grants.gov carries a structured synopsis record, and that record has fields for numberOfAwards (Expected Number of Awards), estimatedFunding, awardCeiling, awardFloor and closeDate. It has no field for expected applicants, and none for applications received. We pulled a live example on August 28, 2026 — the Department of Transportation's Small Business Transportation Technical Assistance Center opportunity, OST-OSDBU-2026-SBTTAC-REGION4-FY2026-1 — which publishes 1 expected award and $300,172 in estimated funding, and says nothing at all about how many organizations will compete for it. The denominator is missing from the schema, so it is missing from every downstream site that reads the schema.

State programs disclose applicant counts unpredictably, in legislative and annual reports

State small business grant programs are the most likely place to find a real published rate outside federal research, because state economic development agencies usually owe their legislature a report. The disclosure is inconsistent: some reports print applications received and grants awarded side by side, some print only dollars obligated, and the reporting year often lags two years behind the round. California's Office of the Small Business Advocate published both halves for its COVID-19 relief program because the program's scale made it a legislative priority. The practical rule for a state program is to search the administering agency's own publications page, and the state legislature's mandated-report library, for the program name plus the words “applications received.”

Corporate and contest grants almost never publish an applicant count

Private contests are where invented approval rates come from, because the winner count is loudly published and the entry count almost never is. WomensNet's Amber Grant is a fair illustration: the program awards three $10,000 grants every month and publishes its rules, its FAQ and its winners in detail, but neither its rules page nor its FAQ page states how many applications it receives — we checked both on August 28, 2026. Any percentage you see attached to a monthly contest like this one has a numerator from the sponsor and a denominator from somebody's guess. Treat a contest rate as unpublished unless the sponsor itself named the entry count for a specific, dated cycle.

How to find a published approval rate when one exists: five places to look

These five sources, worked in order, will find a real rate if a real rate exists. Nothing else reliably will.

1. The agency's statistical office. NIH RePORT and the NIH Data Book, and NSF's Merit Review Digest, are the two deep US series. Start here for any federal research program.
2. The program's report to Congress. The SBA's SBIR/STTR annual report is the only cross-agency source; state equivalents are reports to the legislature.
3. The funder's own award-announcement press releases. A minority — the CDFI Fund is the clearest example — publish application demand as a separate release before naming winners.
4. Federal Register notices under the Paperwork Reduction Act. These publish the agency's estimated number of respondents for an application form.
5. A Freedom of Information Act request. Applicant counts are ordinarily releasable aggregate statistics; applicant identities usually are not.

How to reason about your odds when the approval rate is not published

When a program publishes no rate you can still learn something real, provided you keep the two sides of the fraction honest. The supply side — how many awards exist and how often — is almost always knowable. The demand side — how many people compete for them — is usually not. The correct output of this exercise is a statement about award scarcity with its assumptions attached, not a percentage. Below is the method, applied to a hypothetical program that awards 25 grants twice a year, so the arithmetic is visible.

Step 1: compute the supply side from award count and cadence

Multiply the number of awards per round by the number of rounds per year to get annual award supply. A program awarding 25 grants in each of two rounds supplies 50 awards a year. Then narrow the supply to awards you could actually receive: subtract set-asides, geographic restrictions and categories you do not fit. If 10 of those 50 are reserved for a rural set-aside and you are urban, your real supply is 40 — a 20% cut before any competitor is considered, and at a $50,000 award ceiling, $2,000,000 of realistically reachable funding rather than $2,500,000. Every input here is published — award counts appear in the funder's own announcement, cadence appears in the program page, set-asides appear in the notice of funding opportunity — so this half of the fraction carries no guesswork at all.

Step 2: bound the demand side, or say plainly that you cannot

The applicant count is the number that does not exist. Three partial substitutes are legitimate, and each must be labeled as what it is. First, a comparable program that does publish: a state manufacturing grant with no published rate sits nearer the CDFI Fund's 54% than a national contest's odds, because eligibility is narrow. Second, an eligibility ceiling: if a program requires a certification held by 400 firms nationally, the applicant pool cannot exceed 400, so the 40 awards from Step 1 cannot be spread thinner than 10% of the eligible population — a floor on the award rate, not the award rate itself. Third, the funder's own oversubscription language — “more than three times the amount of available funding,” in the CDFI Fund's words — which bounds the ratio in dollars even when the headcount is unpublished. If none of the three applies, the honest answer is that the demand side is unknown.

Step 3: use the agency's Paperwork Reduction Act estimate as a denominator of last resort

Every federal agency that collects information from the public must publish an estimate of how many people will respond, in a Federal Register notice under the Paperwork Reduction Act of 1995. Those notices carry a line reading “Estimated Number of Respondents.” For example, USDA's Rural Utilities Service published a notice for the OneRD Guaranteed Loan Program on November 27, 2023 stating “Estimated Number of Respondents: 725.” This is a burden estimate produced for budgeting, not a count of applications, and it usually bundles several forms and several programs together. It gives you an order of magnitude for the denominator and nothing more precise than that. Search federalregister.gov for the program name plus “Estimated Number of Respondents.”

Never present a derived ceiling as an approval rate

Dividing 40 available awards by a 400-firm eligibility ceiling gives 10%, and that number is not an approval rate. It is an upper bound on award scarcity under the assumption that every eligible firm applies, which is never true; real application rates among eligible firms are typically a small fraction. Publishing that 10% as “the approval rate” would be the exact error this page exists to correct. State it as what it is: “at most 40 of the 400 eligible firms can be funded in a year, and we do not know how many of the 400 apply.” That sentence is useful, checkable and honest; the percentage is none of the three.

Four grant approval-rate claims that do not survive checking

These four claims recur constantly in small business grant content, and each one is either contradicted by a primary source or has no traceable source at all. Two of them are wrong in an interesting way: they are stale rather than invented, which is why they survive so long. A 20% SBIR success rate was accurate at NIH in fiscal 2004 and is wrong by a factor of 2.5 in fiscal 2025; the SBA has never awarded general small business grants, but the belief that it does is reinforced every time a site lists SBIR under “SBA grants” without saying that the eleven participating agencies fund and award them. For each claim below we give the primary document that settles it, with the date we read it. Where nothing settles it, we say that instead of substituting a guess.

Claim: “The SBA gives grants to small businesses”

The Small Business Administration says otherwise on its own grants page: “SBA does not provide grants for starting and expanding a business.” The same page states where SBA grant money does go: “SBA provides grants to nonprofits, Resource Partners, and educational organizations. These grants aim to support entrepreneurship through counseling and training programs.” The exceptions are research programs — SBIR and STTR — which SBA coordinates but which are funded and awarded by the eleven participating agencies, not by SBA itself. Any “SBA grant approval rate” for an ordinary operating business is a rate for a program that does not exist. Source: sba.gov/funding-programs/grants, read August 28, 2026.

Claim: “SBIR success rates are around 20%”

This was roughly true a decade ago and is badly stale now. NIH's own series shows SBIR Phase I at 20% in fiscal 2004 and 8% in fiscal 2025, with the whole SBIR portfolio at 10% in fiscal 2025. The nearest currently defensible statements are narrower and dated: NIH SBIR Phase I was 8% in fiscal 2025; the SBA's government-wide SBIR Phase I selection rate was 16% in fiscal 2022; NSF says its historical Phase I funding rates “have been between 10% and 20%.” A single blended “SBIR success rate” across eleven agencies, two phases and twenty years is not a number any agency publishes. Sources: NIH Data Book 115 and 96, read August 28, 2026.

Claim: a bare percentage with no named denominator

Figures of the shape “only X% of small business grant applications are approved” circulate without an agency, a program, a year or a source. We looked for a primary source behind the general claim and did not find one, and there is a structural reason none can exist: no US body collects applications to all small business grant programs, so no US body could compute a national approval rate even if it wanted to. Grants.gov aggregates federal opportunities, not applications. USAspending.gov records awards, not applications. A national small business grant approval rate is not a statistic anyone is in a position to publish, which is exactly why the number floating around has no author.

Claim: a real rate quoted with its year removed

The subtlest failure is a genuine, correctly-sourced figure quoted without the fiscal year it belongs to. NIH's SBIR Phase I success rate has been 28% (fiscal 1998), 11% (fiscal 2011), 16% (fiscal 2023) and 8% (fiscal 2025). Quote any of those without its year and you are three-quarters wrong most of the time. The rule we apply on this page, and the one worth applying to anything you read: a grant approval rate without a stated period and a linked source is not a fact, it is a rumour with a decimal point. Every figure in the table above carries both.

What GrantCompass measures instead: program competitiveness across 665 programs

Because published approval rates cover so few programs, GrantCompass rates every program in its catalog on a curator-assigned competitiveness score from 1 (very low) to 5 (very high), built from award-to-applicant scale, eligibility breadth, review structure and historical oversubscription. The figures in this section were computed on July 17, 2026 across the then-665-program US catalog, of which 661 carried a competitiveness score and 661 an estimated-hours figure. The catalog has since grown to 736 programs, 404 of them open, with a median published funding ceiling of $150,000 and a median application estimate of 12 hours as of August 2026. The US small business funding statistics page covers the landscape numbers; this section covers competitiveness only.

Private and foundation programs are the most competitive segment — not the federal government

Federal grants have a reputation as the hardest to win, and the GrantCompass catalog data says otherwise. Private and foundation programs median a competitiveness score of 4 of 5 across 154 scored programs — higher than federal's median of 3 of 5 across 185 programs. State programs are the least competitive segment overall, medianing 2 of 5 across 291 scored programs, and local and municipal programs sit at a median of 3 across 31 programs. The published federal data in the table above points the same way for the most selective end: an 8% NIH SBIR Phase I rate is competitive, but a national corporate contest that funds a few dozen businesses out of a self-selected national pool is more so.

Competitiveness across 661 scored programs skews toward the winnable end

Across the full GrantCompass catalog as computed on July 17, 2026, competitiveness scores distribute as follows: 116 programs (17.5%) score 1 (“very low”), 196 (29.7%) score 2 (“low”), 154 (23.3%) score 3 (“moderate”), 123 (18.6%) score 4 (“high”), and 72 (10.9%) score 5 (“very high”). That means 47.2% of scored programs sit in the two most winnable tiers, and only about one in nine sits in the most contested tier. The programs people write articles about are drawn almost entirely from that top tier, which is the main reason the public impression of grant odds is worse than the catalog suggests.

Federal programs take 8× longer to apply for than private ones

The median US small business grant application in the GrantCompass catalog took 14 hours to complete as measured on July 17, 2026 across 661 of 665 programs, and that single number hides an enormous split by level. Federal programs median 40 hours — unsurprising given SBIR-scale technical proposals — while private and foundation programs median just 5 hours. State programs sit at 15 hours and local and municipal programs at 8 hours. Hours are not odds: the 8% NIH SBIR Phase I rate and the 40-hour federal median are two independent costs of the same application, and a program can be cheap to apply for and still rarely funded.

The Program Hourly Rate: what each level pays per hour of applicant effort

Comparing each level's median hours against its median realistic award gives what GrantCompass calls the Program Hourly Rate: the median realistic-maximum award of a catalog slice divided by the median hours required to apply for it. It is a raw, unweighted catalog-level number, not a personalized estimate, and it deliberately ignores odds. Federal programs post the highest rate at $15,000 per hour despite requiring the most hours; local and municipal programs post the lowest at $2,500 per hour despite being the fastest to apply for, because their awards are small. See average small business grant amount for the award-size distribution behind these medians.

Why a competitiveness score is not an approval rate

A GrantCompass competitiveness score is an ordinal rating of how contested a program is, on a 1–5 scale. It is not a percentage, it does not convert to one, and no average of these scores produces an approval rate. Two programs scored 5 can have very different real odds, and a program scored 2 with a tight eligibility rule can be harder for a specific business than a program scored 4 with a broad one. GrantCompass researches approval-rate evidence program by program as part of a paid seven-field Win Brief, and does not publish those findings on public pages. What is published here is the competitiveness distribution, the hours data, and the agency figures above — each with its method stated.

Competitiveness profile by level and industry, all applicants

The tables below are the data behind the picker, printed in full so they are readable without JavaScript. Each row gives the competitiveness distribution across the five tiers in 1-2-3-4-5 order, the median competitiveness score, the median hours to apply, the median realistic-maximum award, and the number of catalog programs in that slice. All figures were computed on July 17, 2026 across the then-665-program catalog, of which 661 carried a competitiveness score. Read the distribution column first: a slice like federal technology, whose 98 scored programs distribute 19-13-21-28-17, is genuinely bimodal: 32 sit in the two least contested tiers and 45 in the two most contested. A median of 3 alone hides that. None of these columns is an approval rate, and none of them converts into one.

Federal and state programs, all applicants

Federal technology programs median 3 of 5 on competitiveness and 77.5 hours to apply, the heaviest slice in the catalog; state technology programs median 2 of 5 and 18 hours. The pattern holds across every industry: at the same industry, state programs are never more contested than federal ones and are far faster to apply for, while federal median awards run roughly two to seven times larger.

Local, municipal, private and foundation programs, all applicants

Local and municipal programs cluster tightly at a median competitiveness of 3 of 5 with small awards; private and foundation programs are the most contested slice in the catalog, with technology at a median of 4 of 5 and 13 of 45 programs scoring the maximum. Private and foundation programs are also the fastest tier overall, at a median of 5 to 6 hours by industry, which is precisely why they attract the largest self-selected applicant pools.

Competitiveness profile for first-time applicants

The same catalog, filtered to programs GrantCompass flags as workable for a business that has never applied for a grant before, computed July 17, 2026. The filter changes the picture sharply at the federal level: federal manufacturing drops from a median competitiveness of 3 of 5 across 113 programs to 1 of 5 across 28, and the median application estimate falls from 60 hours to 8. The all-industry federal slice shrinks from 188 programs to 62, a 67% reduction, and its median realistic award falls from $600,000 to $150,000. Restricting yourself to first-time-friendly programs therefore trades award size for accessibility: you give up three-quarters of the federal money ceiling and most of the opportunity set, and in return the paperwork drops by roughly 80% and what remains is markedly less contested.

Federal and state programs, first-time applicants

Federal first-time-friendly programs median 8 to 9.5 hours to apply across every industry, against 25 to 77.5 hours for federal programs generally. Median awards drop correspondingly, from $600,000 to $150,000 for the all-industry federal slice. State first-time-friendly programs move less, from a median of 15 hours to 12, because state programs were already the accessible tier.

Local, municipal, private and foundation programs, first-time applicants

Private and foundation programs barely move under the first-time filter: technology holds a median competitiveness of 4 of 5 across 37 programs against 45 unfiltered, because these programs are open to everyone by design and are contested for that reason. Local and municipal first-time-friendly programs are the fastest applications in the catalog at a median of 3 to 9 hours, with median awards from $10,000 to $32,500.

Key takeaways

• Only a handful of US funders publish an approval rate. NIH, NSF, the SBA's SBIR/STTR report and the Treasury's CDFI Fund account for nearly all of them.
• NIH SBIR Phase I was 8% in fiscal 2025 — 349 awards from 4,468 applications, the lowest in a series that runs back to fiscal 1998.
• Government-wide SBIR and STTR Phase I selection was 17.6% in fiscal 2022, and agency rates inside that ranged from 6% (Education) to 22% (Homeland Security).
• The three published measures use three different denominators. NIH's success rate excludes withdrawn applications; NSF's funding rate excludes preliminary proposals; the SBA's selection rate excludes nothing.
• Award counts are public and applicant counts are not. The Grants.gov opportunity schema has an Expected Number of Awards field and no applicant field at all.
• A derived ceiling is not an approval rate. Awards divided by an eligibility population assumes everyone eligible applies, which never happens.
• 47.2% of the 661 scored GrantCompass programs are low or very-low competition, so most programs are far more winnable than the famous ones.
• See which programs you qualify for in the explorable database of 736 programs or the free eligibility matcher.

How this page was researched

Every external figure on this page was opened at the funder's own document on August 28, 2026, and the URL is printed beside it. Nothing was taken from a secondary summary, a competitor page, or an AI-generated source. Where a funder published a percentage, that percentage is quoted as published. Where a funder published two counts but no percentage — the CDFI Fund and CalOSBA rows — the division is labeled computed and both counts are shown so you can check the arithmetic. Where no rate is published, this page says so instead of estimating one.

Written and maintained by Khalid Hamadeh, founder of GrantCompass. Corrections and source challenges are welcome — see about and contact. GrantCompass is an independent funding-discovery tool and is not affiliated with any government agency.