The SBIR Awardee Read: Venture Track, Grant Mill, or Too Early

Use this guide when you have a list of companies pulled from SBIR, STTR, or other federal research award records and need to decide which ones get a meeting. It's written for early-stage investors, fund platform and talent partners, and angels who source deep tech from federal grants. It gives you six scored evidence dimensions, three labels (Venture Track, Grant Mill, Too Early), and a 20-minute procedure that ties every score to a public record you can cite.

What separates a venture-track awardee from a grant mill?

A venture-track awardee uses a few federal awards to de-risk one technology and then raises private equity. A grant mill is built around winning award after award, and when it grows, it grows through government procurement rather than venture capital.

Terms used throughout:

• SBIR (Small Business Innovation Research) and STTR (Small Business Technology Transfer) are the federal programs that fund research at small companies. An STTR award names a partner research institution.
• Phase I is the smaller feasibility award. Phase II is the larger development award that can follow it.
• Principal investigator (PI) is the person named as research lead on an award. The PI doesn't have to be an owner.
• Grant mill means a firm whose business is the award pipeline itself. There is no statutory definition. CSIS compared firms with 50 or more awards against firms with 3 to 9.

The research says awards are a good sign, but only the first few. The National Academies review of NSF's program found that funded firms "were more than five times as likely to attract angel investments or venture capital" as a control group. Sabrina Howell's study of DOE awards found that an early-stage award "approximately doubles the probability that a firm receives subsequent venture capital". After that the effect reverses. Lerner found that "multiple awards did not increase performance", and a DOE EERE study cited by ITIF found that firms with previous SBIR awards had half the probability of later venture investment. The likely mechanism is that the firm reorganises around writing proposals.

CSIS joined SBIR.gov award histories, open patent data, and SEC Form D filings to see where heavy winners end up over five years:

Mills are not stagnant. They win procurement contracts and patents at higher rates than smaller awardees. They just rarely sell equity. A Harvard Business School working paper found that between FY2000 and FY2021, prior SBIR awardees received an average of $2.7 in non-SBIR contracts for every $1 awarded. An independent audit of Arizona awardees found 10 long-tenure firms with 5 or more Phase II awards, holding $122M between them, and no venture record for any of them.

A Grant Mill label is a judgement about capital fit, not about the quality of the science.

Which award record fields carry the read?

Most of the read comes from the firm name, agency, phase, dates, amount, the identifiers (UEI and DUNS), and the PI name. The employee count and research-institution fields help when they're filled in, but they often aren't.

The SBIR.gov award API returns these fields, grouped by what you use them for:

• Identity: firm, uei, duns. UEI is the Unique Entity Identifier assigned by SAM.gov, 12 characters long.
• Award shape: agency, branch, phase, program, topic_code, solicitation_number, solicitation_year, award_year, award_amount.
• Timing: proposal_award_date, contract_end_date.
• People: pi_name, pi_email, ri_name (the research institution).
• Size: number_employees.

Two of these fields often let you down. One third-party scraper reports that ri_name is present on STTR awards only. A Harvard Business School paper found the employee count is reported for only about 60% of business-year observations. One practitioner audit matched companies "by name, UEI, and DUNS", which means no single key is enough. Nobody has published how often identifiers go missing, change, or duplicate for the same company, so check your own export. Pick ten firms you already know and see whether each one comes back as exactly one entity. The SBIR.gov API page has also carried a maintenance notice, so save your raw export with its pull date before you score anything.

Records differ in how much weight they can carry. Rely on the top layers and treat the bottom ones as leads to check.

The six dimensions: what each proves and how to score it

Score each dimension 0, 1, or 2, where 2 points toward a venture-scale business, and attach the record behind the score. A score you can't cite is a guess, and it counts as 0 until you find the record.

1. Award pattern and agency spread

This dimension shows whether the awards follow one technology or follow the solicitations. Count Phase I and Phase II awards per agency per year, and read the topic codes.

• 2: One technology line, typically a Phase I that led to a Phase II, at one or two agencies.
• 1: Around 3 to 9 awards with a clear topic thread. That's CSIS's comparison group, of which about 1 in 10 raised.
• 0: High volume across unrelated topics and agencies, close to or above the benchmark trigger of more than 20 Phase I awards in 5 fiscal years.

NSF Phase I funding rates have historically run between 10% and 20%. A firm that wins over and over at those rates is very good at writing proposals, and that skill is exactly what the evidence links to lower venture odds.

2. Phase I to Phase II conversion

The conversion ratio is Phase II awards divided by Phase I awards. The SBA sets a floor for heavy users, and the floor rises with volume.

Sources: sbir.org, CRS R43695, Vertx Partners. The increased standards took effect April 1, 2023.

• 2: The Phase II continues the Phase I topic. On any Phase I count large enough to mean something, the ratio is at or above 0.50.
• 1: Only Phase I so far, or a ratio between 0.25 and 0.50.
• 0: A ratio below 0.25 on a large Phase I count.

At small counts the ratio is noise. One Phase I and one Phase II gives a ratio of 1.0, and that only means the firm converted. Read it as converted or not yet. A high ratio is not a venture signal by itself either. Phase II grants make up 80% of total SBIR funding. Phase I de-risks the technology before investors look, while Phase II can replace the private capital a venture-track firm would otherwise raise. A firm that keeps converting and never raises is choosing the award path.

3. Commercialization and follow-on contracts

This dimension asks whether anything was sold or invested beyond the awards, and what kind of thing it was. For firms large enough to be assessed, check the outcome tables on the SBA performance benchmarks page. For everyone else, look for non-SBIR contracts and product sales.

• 2: Evidence of private investment or non-government sales: a commercialization pass that rests on the sales or investment figures, or named commercial customers.
• 1: Non-SBIR government contracts. This is a real business on the procurement path, and that path is where mills tend to go.
• 0: Nothing beyond awards, or a pass that rests on patents or on a self-report you can't date.

The type of pass matters. Under the standard tier, patents equal to at least 15% of Phase II awards can substitute for sales. Under the increased standards, patents can't be used to pass. For firms with 50 or more Phase II awards the bar is $250K in sales or investment per award, and at 101 or more it is $450K. A firm that fails an increased standard is limited to 20 new Phase I and Direct-to-Phase II awards at each participating agency the following year. That limit will show up later in its award pattern.

4. Private-capital trace

This dimension shows whether anyone outside government has priced the equity. Search SEC Form D filings under the firm's current name and every former name.

• 2: A Form D showing a venture-style raise, under the current or a former name.
• 1: Indirect evidence of non-SBIR capital, such as a Strategic Breakthrough award, which requires substantial matching funds from non-SBIR sources.
• 0: None found after searching every name. Record the names searched and the date.

Read a silent Form D history against what the awards can pay for. An award ceiling divided by its term gives the monthly spend that award can carry:

Sources: NSF 26-510 coverage, azcunadynamics, fundinglandscape. The SBA adjusts its Phase II ceiling. Check the current figure before relying on it.

A firm whose headcount fits inside a Phase II budget has no need for your money. NIH also allows a fee of up to 7% of total costs per phase, so a firm can stay viable on awards alone. In the CSIS data, private capital and follow-on procurement are separate outcomes, not ranked against each other. Score them separately and don't let a strong contract record make up for a missing raise.

5. Founder versus hired principal investigator

This dimension shows whether the person whose science won the award owns the result. Match pi_name on every award against the founders on the company site, the SAM.gov points of contact, and any university named in ri_name.

• 2: The same PI recurs on the core awards and appears among the founders or points of contact.
• 1: The PI is a faculty co-founder on an STTR award with an unclear time commitment, or a founder is PI on some awards and a hire is PI on others.
• 0: The PI isn't among the owners, or PIs rotate across unrelated topics.

A SAM.gov point of contact is a business contact, not proof of ownership. Treat it as supporting evidence next to the company site and any filing that names officers.

Keyword search doesn't help here. In Refolk's index of professional profiles, only 1 US profile combines a "Principal Investigator" title with "SBIR" in the headline, and that person's actual title is Founder. Founder-type profiles mention SBIR far more often.

That 57 is not clean either. In a 25-profile sample, one was a "Founder & Proposal Architect" and one was off-topic, so even founder headlines include grant consultants. The reliable approach is to take pi_name from the award and look up that specific person. For university spinout deal sourcing, that means searching for the named people and their roles, which is what Refolk is built for: describe the person you need in plain English and get back profiles with their current employer and history.

6. Commercial hiring

This dimension shows whether the firm is spending money on selling. Look for named sales, business development, and product hires at the firm.

• 2: Two or more named sales, BD, or product hires, each confirmed at this employer.
• 1: One such hire, or a founder publicly carrying the commercial role.
• 0: Only research and proposal roles visible.

Keyword matches are mostly noise here. Refolk's index shows how much:

A commercial hire at an SBIR firm usually doesn't put "SBIR" in their headline, so search by company and role. In Refolk you can ask for the head of sales or business development hired at a named firm and check the employer on each result. That replaces scrolling through keyword hits that are mostly misspellings. The number_employees field on the award record can show whether headcount grew, but it's filled in for only about 60% of business-year observations.

How do the scores become Venture Track, Grant Mill, or Too Early?

Run five gates in order and stop at the first one that fires. The total score matters only when no earlier gate fires. The gates exist so that an established award shop and a firm with one award don't both land at a low total and get treated the same way.

1. Mill gate: Award pattern is 0 and private capital is 0. Label: Grant Mill.
2. Too Early gate: No Phase II yet, or one Phase II whose contract_end_date hasn't passed, and private capital is 0. Label: Too Early. Set a re-check for the contract end date.
3. Capital gate: Private capital is 2 and founder versus PI is at least 1. Label: Venture Track.
4. Total gate: 8 or more out of 12. Label: Venture Track.
5. Default: Grant Mill for this fund at this time. Re-read the firm when a Form D appears or a contract ends.

These cut lines are my calibration, not a published standard. No study has validated a threshold for separating these groups. Before you trust the 8-of-12 line, run the scorecard on ten past meetings whose outcomes you already know and adjust it.

Copy this scorecard once per company:

Company: [firm name]  |  UEI: [12 characters]  |  Prior DUNS: [if any]
Pulled: [date]  |  Awards in deduped row: [count]  |  Lapse window flagged: [yes/no]

1. Award pattern and agency spread: [0/1/2]
   Record: [awards per agency per year, topic thread]
2. Phase I to II conversion: [0/1/2]
   Ratio: [x.xx] vs 0.25 / 0.50  |  Record: [award rows]
3. Commercialization and follow-on: [0/1/2]
   Record: [benchmark table and date, or named contract]  |  Rests on: [sales / investment / contracts / patents]
4. Private-capital trace: [0/1/2]
   Record: [Form D date and type, or "none found: names searched, date"]
5. Founder versus PI: [0/1/2]
   Record: [pi_name per award vs SAM.gov contacts vs company site]
6. Commercial hiring: [0/1/2]
   Record: [named hire, title, employer confirmed]

Total: [x/12]
Gate fired: [Mill / Too Early / Capital / Total / Default]
Label: [Venture Track / Grant Mill / Too Early]
Re-check on: [contract_end_date, next benchmark table, or next Form D search]

The 20-minute procedure, step by step

Do these eight steps in order for each company: dedupe, score the six dimensions, then label. The time budget is about 20 minutes per company, and the commercialization check takes the longest. CSIS treats private capital and follow-on procurement as separate outcomes and doesn't rank one over the other. No published source sets an order for these checks, so the order below is the one that reuses records most efficiently.

How this read goes wrong: failure modes and false positives

Most mistakes come from a record that looks current or complete when it isn't. Each failure mode below has a check, and each check takes a minute or less.

1. A stale self-report reads as current. DoD treats commercialization reports as voluntary after five years, so an old pass can sit in the record and look fresh. Check: the date on the benchmark assessment table, and whether any Form D dates support the claimed investment.
2. A hired PI is presented as a founder. Pitch materials can call the lead scientist a co-founder while pi_name changes from award to award, or never matches the people who own the firm. Check: SAM.gov points of contact against pi_name across every award, not just the latest one.
3. The lapse gap reads as decline. The programs expired on September 30, 2025, which paused new awards until the reauthorization was signed on April 13, 2026. Zero awards in that window says nothing about the firm. Nobody has published award-count gaps by agency for this period. Check: compare the firm's award count with peers at the same agency over the same window.
4. A DUNS to UEI split makes one firm look like two. Each half has half the awards, so a mill can score as a modest awardee. Check: dedupe on name plus address, not just identifiers.
5. Headline keyword sourcing pulls in the wrong people. "SBIR" in a headline matches grant consultants, proposal writers, and misspellings. Three of the four sales, BD, or commercialization matches in Refolk's index were spelling-variant false hits. Check: confirm that each person's employer is the firm with that UEI.
6. Reading "mill" as "no growth". CSIS shows heavy winners moving into procurement at 71% against 44% for the 3-to-9 group. If you write a mill off as stagnant, you'll misjudge it when it turns up as a supplier or acquisition target. Check: keep the label to capital fit and record the procurement path in the notes.
7. Patents pass the benchmark in place of sales. Under the standard tier, a patent count can stand in for commercial results, so a firm can pass with no revenue at all. Check: read the sales and investment figures specifically and score dimension 3 on those.

A false positive in the other direction deserves equal attention. A firm with a Form D that raised mainly to cover the match required for a large award can look like Venture Track while it behaves like an award shop. If the capital gate fires but commercial hiring is 0, take the meeting to ask what the money was for, and say in the notes that the label is provisional.

What the reauthorization changes in the read

The reauthorization keeps the programs running through September 30, 2031. It changes two things this framework reads: it caps how many proposals a firm can submit, and it adds a large award tier that requires private matching money. Both make the private-capital dimension more informative and the raw award count less so.

S. 3971 was signed on April 13, 2026. The mechanisms that matter for scoring:

• Per-agency proposal caps. From fiscal year 2027, each agency must cap how many proposals one company can submit in a fiscal year and publish that cap at least 90 days before the fiscal year starts. The statute leaves the number to each agency. One consultancy cites a "9-proposal cap", but that figure is unverified. Check each agency's published notice.
• Strategic Breakthrough tier. Up to $30 million per project over up to 48 months. Applicants need at least one prior Phase II award and substantial matching funds from non-SBIR sources. The tier is reported to be available only at agencies with extramural R&D budgets above $100M.
• No lifetime cap. A proposed $75M lifetime funding cap per company was dropped, so heavy winners are limited by the submission caps rather than by a total.
• Carryover. Unspent FY2026 funds can be used in FY2027, so award timing around that boundary may look uneven.

How to adjust the scoring:

• Award pattern. Once caps apply, a heavy user can't build volume as fast. Compare velocity against the agency's published cap, not against the pre-cap history.
• Private capital. A Strategic Breakthrough winner has by definition brought matching non-SBIR money, which earns at least a 1 on dimension 4. Look for the Form D that funded the match.
• Benchmarks. The SBA posts assessment outcome tables on its performance benchmarks page. I couldn't verify a fixed publication date, so note the date of the table you used every time.

Before you book or decline the meeting

Before the label goes into your pipeline, confirm that each score has a citation, each known trap has been checked, and a re-check date is set. Labels go stale as contracts end, filings appear, and benchmark tables update.

To keep the read current:

• Too Early firms: re-run the scorecard on the Phase II contract_end_date. That's when the firm either raises, wins another award, or goes quiet.
• Grant Mill firms: set a Form D search to repeat each quarter. One venture-style filing is enough reason to re-score.
• Venture Track firms: re-check commercial hiring before the second meeting. New sales and product hires are the earliest public sign that the award money is building a business.
• The framework itself: after every ten labelled firms whose outcomes you later learn, check whether the 8-of-12 total gate is sorting them correctly, and move it if it isn't.