Imagine you've raised a Series A. You have $10M in signed LOIs from enterprise customers. You shipped tens of thousands of units last year.
Your product is on retail shelves in two states. Someone says "you should apply for SBIR."
You start writing an NIH SBIR application and your reviewers do not care about any of it.
Different federal agencies score SBIR applications on different rubrics. Some weight publications and PI credentials heavily. Others weight letters of intent, customer pilots, and revenue.
The question of SBIR commercial traction vs publications is not academic -- it determines which agency will actually score your evidence as evidence. Founders searching for federal grants without publications, or SBIR programs for commercial-stage startups, are choosing between rubrics that reward very different things in practice.
This guide maps each major agency on a commercial-proof versus academic-proof matrix, so you can route your traction toward agencies that score it.
Short answer: SBIR commercial traction is weighted heavily at AFWERX (one of three scored criteria), first-class at NSF SBIR/STTR Pitch (Section III), translated to adoption pathway at ARPA-H, and largely ignored at NIH SBIR. Founders with revenue, LOIs, and pilots but no publications should default to AFWERX Open Topic, NSF Pitch, DIU CSOs, or prize competitions before NIH SBIR.
Why some agencies do not score your traction
Every federal grant program has a written or implicit rubric. Reviewers score against that rubric. That is all that matters.
A rubric mismatch is when your strongest evidence sits in a category the rubric does not weight. You can have $10M in LOIs and still score 6 out of 9 on NIH's "Investigator" criterion if you do not have publications.
The LOIs are real evidence. NIH reviewers just do not have a place to put them.
The cost is concrete. A federal SBIR Phase I application takes 40 to 80 hours of founder time.
If you apply into a rubric that does not weight your evidence, you lose those hours. The application is not technically deficient -- it just scores in the middle of the pack on the criteria that matter, which is the same as not getting funded.
Cada has seen this pattern in 5 of 17 commercial-stage roadmap engagements over the last 18 months. The fix is upstream of the application: choose an agency whose rubric scores what you have.
The agency rubric matrix: where commercial traction counts
Here is the matrix, agency by agency. Full methodology is in the agency rubric matrix v1 dataset (available on request).
| Agency | Scoring scale | Commercialization weight | Publications penalty | Best mechanism for commercial founders | Primary disqualifier |
|---|---|---|---|---|---|
| NIH SBIR Phase I | 1-9 (1 = best) | Low (1 sentence on Aims; penalized if dominant) | High (-15 in roadmap scoring) | Direct-to-Phase-II (D2P2) | No preliminary data |
| NSF SBIR/STTR Pitch | 1-9 (9 = best) | High (Section III: LOIs, pilots, MOUs are first-class) | Standard (-5) | Standard Pitch with R&D uncertainty | PI not US citizen/PR; engineering-only innovation |
| ARPA-H | 1-9 (9 = best) | Medium (15 percent of score; adoption pathway) | Standard (-5) | SSO Mission Office | Revenue framing crowding out health-impact framing |
| AFWERX (DoD SBIR) | 1-9 (9 = best) | High (~33 percent of score; primary purpose) | Standard (-5) | Open Topic | TRL 7+; no AF end-user; civilian SBIR language |
| DIU (CSO) | Per-program | High (TRL 4+ favored; commercial framing) | Standard (-5) | Commercial Solutions Opening | Below TRL 4; no defense angle |
| Prize competitions | Per-program | Often the only criterion | None | xTech, agency-specific prizes | Below purse threshold |
A few patterns worth naming. NIH penalizes "no publications" three times more heavily than every other agency in Cada's roadmap scoring.
Engineering-optimization technologies cap at score 4 (auto-decline) at NSF and ARPA-H but are welcomed at AFWERX, DIU, and most prize competitions. AFWERX is the only mechanism where customer discovery is the primary work product of Phase I, not a supplementary section.
The rest of this guide walks through each agency's rubric in detail.
NIH SBIR Phase I: $275K Phase I where publications carry the most weight
NIH SBIR Phase I is a $275K, 12-month award reviewed by a Center for Scientific Review (CSR) study section. The study section is a panel of 15 to 20 academic researchers who score your application from 1 to 9 (1 is the best score, opposite of every other agency on this list).
Five criteria are scored independently: Significance, Investigator(s), Innovation, Approach, and Environment. Reviewers also assign an Overall Impact score.
Fundable applications generally land at 1 to 3. Anything 6 or worse is not competitive.
Where commercial proof goes: one optional sentence on the Specific Aims page where you can mention market size. That is it.
The Cada NIH SBIR playbook explicitly flags "commercialization language dominating the scientific narrative" as a red flag that lowers scores. Reviewers are scientists. They want scientific premise, mechanism, preliminary data, and a sound experimental approach.
Where academic proof rules: preliminary data is technically "not required" but is de facto required. NIH study sections almost always penalize Phase I proposals lacking preliminary data.
The PI's publications appear in the Investigator section. Citations are required across four buckets: a mechanism or pathology landmark paper, a current standard-of-care paper, an outcomes or epidemiology paper, and a method-validation paper.
The honest caveat: NIH SBIR does not require the PI to be a U.S. citizen, but it does require the PI to be employed at least 51 percent by the small business during the award. This rules out "borrowing" an academic PI part-time.
If you want NIH SBIR and you do not have an in-house PI with publications, your most realistic path is Direct-to-Phase-II (D2P2). D2P2 lets companies skip Phase I if they have prior non-SBIR-funded preliminary data demonstrating feasibility.
Self-funded R&D from a commercial-stage company can qualify as that preliminary data.
If you have no publications, no preliminary data, and no in-house PI with academic credentials, NIH SBIR is the wrong rubric.
NSF SBIR/STTR Pitch: where LOIs are first-class evidence
NSF runs a two-stage process. The Project Pitch is a 1,750-character submission a Program Director reviews and either invites or declines.
Invitation rates fluctuate with topic and budget, but the pitch is the actual gate. Full proposals follow only after invitation.
NSF scores on four criteria: Innovation and Technical Risk, R&D versus Productization Clarity, Market Opportunity, and Team Execution Readiness. (NSF's official solicitation language differs slightly; these are Cada's internal labels for the same scoring dimensions.) The scale is 1 to 9 with 9 as best, decided by a single Program Director.
Where commercial proof counts: Market Opportunity is one of the four criteria. The Cada NSF Pitch playbook documents that customer discovery evidence including pilots, LOIs, MOUs, and signed design-partner agreements is "the strongest form of market validation at pitch stage."
This is first-class evidence. Specific named customer segments score better than TAM-only framings.
Where academic proof counts: Innovation classification is the gate. NSF sorts technology into three buckets: A) new scientific principle (score floor 7), B) novel application of known science to a new domain (score floor 5), or C) engineering optimization of existing approaches (score ceiling 4, auto-decline).
A founder building a better mousetrap with no genuine R&D uncertainty caps at 4 regardless of how strong the LOIs are.
The hard requirement: the PI must be a U.S. citizen or permanent resident. Founders sometimes swap in a co-founder as PI to clear this -- a real option but should be flagged early.
Sweet spot for commercial founder: technology with genuine R&D uncertainty, plus customer evidence (LOIs, pilots, design partners), plus a U.S.-citizen PI.
ARPA-H: 30 percent Innovation, 25 percent Health Impact, traction translated to adoption
ARPA-H uses a Solution Summary process reviewed by a single Program Manager (no peer panel). The scoring is 1 to 9 (9 best).
ARPA-H does not publish a precise weight breakdown; the weights below reflect Cada's internal calibration of how PMs score in practice: Innovation 30 percent, Health Impact and Scale 25 percent, Technical Feasibility 15 to 20 percent, Team and Execution Capability 15 percent, with adoption pathway integrated across the remaining criteria.
Three reviewer lenses operate inside one PM: Program Manager (mission fit), Domain Expert (technical merit), Commercialization/Adoption (can this team get the technology to patients).
Where commercial proof counts: Team criterion (15 percent) requires "three-pillar coverage" -- technical, clinical, and commercialization. A team without a commercialization lead is flagged as a gap.
Adoption pathway gets scored as part of Health Impact and Scale. Concrete user segments and delivery sites are required.
"FQHCs, IHS clinics, and community pharmacies in the 25 lowest-physician-density US counties" scores well. "We will partner with health systems" does not.
The honest catch: ARPA-H's compliance checklist enforces "patient-centered framing (health outcomes, not features or revenue)." Revenue evidence has to be re-framed as "evidence we can reach patients at scale" rather than "evidence the business will succeed."
Mission Office mismatch is a critical gap. ARPA-H has four MOs: HSF (Health Science Futures), PHO (Proactive Health Office), RSO (Resilient Systems Office), SSO (Scalable Solutions Office).
The PM has deep domain expertise in their mission area. Submitting to the wrong MO routes you to a PM without the technical context to score you fairly.
Sweet spot for commercial founder: novel mechanism with a 10x claim, plus a credible adoption-at-scale pathway, plus a team with commercialization expertise. SSO is the most commercial-founder-friendly MO -- its mission covers access, equity, last-mile delivery, and care coordination, where adoption pathway dominates scoring.
AFWERX Open Topic: customer discovery IS the work
AFWERX runs DoD SBIR Phase I as a 90-day, $75K (SBIR) or $110K (STTR) effort. The scoring is 1 to 9 (9 best) across three criteria: Technical Merit, Defense Need, and Commercialization Potential.
The defining quote from the Cada AFWERX playbook: "Open Topic Phase I is primarily about customer discovery (finding an AF end-user), not R&D feasibility. Technical work is secondary to customer engagement planning."
Where commercial proof counts: Commercialization Potential is one of three scored criteria, roughly 33 percent of the rubric.
The 8-9 anchor (high score) explicitly includes "dual-use is clearly articulated with separate commercial and defense market analysis; strong traction evidence (revenue, pilots, LOIs); credible transition pathway to AF program of record; realistic market sizing with credible sources." The 2-3 anchor (low score) is "no traction evidence; no transition pathway; market claims unsourced."
A founder with substantial signed LOIs and a real commercial product scores in the 8-9 range on this criterion before writing a single Technical Volume page.
Defense Need is the other criterion where commercial founders need to do specific work. The 8-9 anchor includes "credible customer discovery plan with named contacts."
This means an actual Air Force end-user (program office, base, or operational unit) who has expressed interest. Phase II eligibility requires a Customer Memorandum -- a document where the AF end-user commits to evaluating the technology.
Phase I is essentially scoped to securing this memorandum.
Hard caps: 90 days, $75K (SBIR) or $110K (STTR), TRL 3-6 ideal (TRL 7+ too mature, TRL 1-2 too early), subcontract less than 33 percent. PI must be employed at least 51 percent by the small business, U.S. ownership greater than 51 percent, no unresolved Foreign Ownership/Control/Influence (FOCI) issues.
Banned language: "broader impacts, hypothesis-driven, study section, unmet medical need." Reviewers (a Government Technical Advisor and a Commercialization reviewer) read this as a tell that the founder copy-pasted from an NIH or NSF application.
"Civilian SBIR language" is a flagged decline pattern.
Sweet spot: defense-adjacent technology, working product, revenue or LOIs, plus a real AF customer-discovery plan with named contacts. This is the strongest fit on this list for founders with traction and no publications.
DIU and prize competitions: when SBIR is the wrong vehicle
If you have a working prototype at TRL 4+ and a defense-relevant use case, DIU Commercial Solutions Openings (CSOs) are often a better fit than DoD SBIR. DIU uses Other Transaction (OT) authority -- no peer panel, no FAR contracting, decisions made by program offices.
Past performance and TRL evidence carry more weight than novelty. Engineering optimization (auto-decline at NSF) can win a DIU CSO if it solves a real defense problem.
For founders with no defense angle, prize competitions are the cleanest path. Cada's roadmap scoring notes that prizes skip the no-publications and no-patents penalties.
Some prize programs (xTech, Challenge.gov) have purses of $500K to $5M. Trade-off: prizes under $25K are flagged as low-ROI.
How to route your traction profile
Quick decision tree based on which evidence types you have:
- Health or biotech, plus academic publications, plus preliminary data -> NIH SBIR
- Genuine R&D uncertainty, plus LOIs or pilots, plus US-citizen PI -> NSF SBIR/STTR Pitch
- Health-impact-at-scale story, plus a team with commercialization expertise -> ARPA-H, SSO Mission Office
- Defense-adjacent product, plus revenue or pilots, plus dual-use story -> AFWERX Open Topic
- Working prototype TRL 4+, plus defense need -> DIU CSO
- Pure commercial product, no R&D angle, no defense angle -> Prize competitions or skip SBIR entirely
The decision tree is a starting heuristic. The actual routing depends on topic-technology fit, available solicitations, and your specific evidence mix.
The matrix is the calibration layer underneath.
Frequently asked questions
Can I apply to SBIR without any publications?
Yes for NSF, AFWERX, DIU, and prizes. Effectively no for NIH SBIR unless you have unpublished preliminary data that you can describe in the Approach section.
NSF and DoD SBIR programs apply a -5 score penalty in Cada's roadmap for "no publications" but it is recoverable. NIH applies a -15 penalty, which is rarely recoverable without a co-PI who has publications.
Do LOIs count as evidence in NIH SBIR?
Only as "letters of support" appended to the application. They are not first-class scoring evidence. The NIH playbook lists them as IMPORTANT but not CRITICAL.
Founders sometimes try to make LOIs the centerpiece of the Significance section, which triggers the "commercialization language dominates the scientific narrative" red flag and lowers scores.
What if my technology is engineering optimization rather than novel science?
NSF and ARPA-H will cap your score at 4 (auto-decline). Look at AFWERX Open Topic, DIU, or prize competitions instead.
AFWERX explicitly weights customer discovery and dual-use over scientific novelty. DIU CSOs are built for commercial-grade engineering. Most prizes care only about whether the product solves the problem.
How do I find an academic PI quickly if I want NIH SBIR?
You usually cannot. The NIH SBIR PI must be employed at least 51 percent by the small business during the award. Bringing in a part-time academic PI rarely clears this bar.
If you have prior preliminary data from self-funded R&D, Direct-to-Phase-II (D2P2) is a more realistic path because it accepts non-SBIR-funded preliminary data. D2P2 is a Phase II grant with no Phase I prerequisite.
Does revenue count as commercialization potential at AFWERX?
Yes. The 8-9 anchor on Commercialization Potential explicitly lists revenue, pilots, and LOIs as traction evidence.
Revenue plus dual-use articulation plus a Customer Memorandum path is the highest-scoring profile. The catch is that defense relevance still needs to be real -- a SaaS product with no plausible Air Force use case will fail Defense Need regardless of commercial strength.
Run your traction profile through the matrix
Knowing the rubric is the first step. The harder problem is mapping your specific evidence (LOIs, shipped units, retail footprint, prior pilots, beta-user counts) to the right agency, the right mechanism, and the right open solicitation.
Cada runs a 30-minute roadmap fit check that does exactly this. We take your traction profile, run it against the matrix, and tell you which 2-3 mechanisms to prioritize.
We tell you the ones to skip and why. No pitch.
If you want the underlying matrix in machine-readable form (YAML), email us and we will send the v1 dataset directly.