Most stalled SBIR drafts are not stalled because the science is wrong. They are stalled because the Specific Aims do not survive the way reviewers actually read. An experienced reviewer at hour two of a 30-proposal stack is not absorbing your prose linearly; they are running a fast, structured scan against a small number of failure modes they already know how to spot. If the Aims trip any of those traps, the draft loses score points the reviewer never has to consciously justify.
This is a five-part pressure test, not a scoring formula. The applicable solicitation and agency criteria control; these checks help surface places where an otherwise strong technical case may be difficult to evaluate.
Diagnostic 1: The novelty claim does not appear before the third paragraph
Heilmeier Question 3 (“What’s new in your approach, and why do you think it will be successful?”) has a specific load-bearing role in how reviewers form an early impression. If the answer to it is not visible by the bottom of paragraph two of Specific Aims, the reviewer’s working hypothesis for the rest of the read becomes “this is incremental work on an established direction.” Once that hypothesis is set, every later piece of novelty in the draft gets discounted because it has to fight the prior the reviewer already formed.
The most common failure pattern: authors who have a genuinely novel approach lead with three paragraphs of literature framing because that is how academic writing is taught. The novelty claim ends up under the heading “Innovation” two pages later, where it lands in a reviewer who has already mentally categorized the work.
The fix is structural, not stylistic. The novelty claim belongs in paragraph one: explicit, one sentence, comparative. Not “our approach uses X.” That is a description. The claim is the explicit statement of what is different from the closest prior work and why that difference matters for the gating problem. Most drafts have the sentence; almost none have it in the right structural position.
A test: hand the Aims to someone who has not seen the draft. Ask them, after thirty seconds, what is new here. If they cannot answer in one sentence, the novelty claim is not where it needs to be, regardless of whether it is in the document somewhere.
Diagnostic 2: The Aims read against each other, not in sequence
Specific Aims are supposed to compose. Aim 1 produces the result that feeds Aim 2, which produces the result that feeds Aim 3. In practice many drafts have three Aims that read as three parallel sub-projects, each with its own internal logic, that the reviewer has to mentally stitch together.
This shows up most often when the Aims were authored across multiple revisions by different team members. Aim 2 is the cleanest section because the metabolic-engineering co-author owned it; Aim 3 is the densest because the ML co-author owned it; Aim 1 is the framing section because the founder wrote it last after seeing the others. Each Aim is locally strong; the composition is invisible.
The reviewer-side cost is concrete. A reviewer who has to assemble the composition during the read spends working memory on stitching that they cannot spend on technical evaluation. The score they assign is based on the cumulative cognitive cost of the read, not on the absolute strength of any one section.
The structural fix is a one-sentence transition between Aims that explicitly names the dependency: “Aim 2 takes the output of Aim 1 and… and that constraint sets the success criterion for Aim 3, below.” This sounds clunky in academic writing. In a Specific Aims section it is the difference between a reviewer who can hold all three Aims in working memory and one who can hold one Aim at a time.
Here is a quick check: read the three Aims in reverse order, starting from Aim 3. If the dependencies become more legible in reverse order than in forward order, the forward composition is broken.
Diagnostic 3: The success criterion for each Aim is qualitative
“Demonstrate the feasibility of,” “characterize the behavior of,” “establish a framework for.” These verbs can hide the acceptance criterion. By themselves, they do not tell a reviewer what observable outcome will show that an Aim succeeded.
The cost shows up in two ways. The first is in the milestone review. A program officer running a mid-cycle check who reads “demonstrate the feasibility” has no contractual hook to either accept or reject the work shown. The PI can always argue feasibility was demonstrated by what was produced; the PO can always argue it was not.
The second cost is in the initial review. A reviewer scoring the draft against the Heilmeier checkpoints question (“how will progress be measured?”) sees a qualitative criterion and infers that the team has not yet decided what success looks like. The inference is sometimes wrong; the inference is always damaging.
The fix is to make every Aim’s success criterion a number, a binary, or a comparison against a named reference. “Achieve a TRL-4 demonstration on the [DOMAIN] substrate, defined as ≥80% recovery of the bench-scale yield.” “Produce ≥20 reference test cases against which the harness produces the same output on three consecutive runs.” “Match the published 2024 [REFERENCE] benchmark within 10% on the [DATASET] cohort.”
A test: highlight every verb in the Aims section. If the highlighted verbs are dominated by “demonstrate / characterize / establish / explore,” the success criteria are qualitative and need rewriting.
Diagnostic 4: The risk register is missing or generic
Heilmeier Question 5 (“What are the risks and the payoffs?”) is the section where most drafts visibly thin out. The pattern: a one-paragraph risk section near the bottom, with three generic items (“technical complexity,” “regulatory uncertainty,” “team capacity”). No specifics. No mitigation that names a contingency milestone. No acknowledgment that the program could kill itself in particular ways.
This reads to a reviewer as one of two things: either the team has not seriously considered failure modes, or the team has considered them and elected not to disclose them. Both readings cost score points. The first costs more than the second; reviewers can interpret silence as either incompetence or sandbagging, and both are penalized.
The strong pattern is a risk register with three to five named items, each with a specific mitigation, each with a documented variance trigger. The format that lands:
R[n]: [Specific failure mode named in concrete language]. Mitigation: [the concrete action that will be taken, named at the milestone level]. Trigger: [the specific observable that will cause the mitigation to fire]. Cost of trigger: [the variance the program would absorb].
Three to five rows of this format does more for the reviewer’s confidence than three pages of risk-adjacent prose. The format is also what a program officer needs to run a mid-cycle conversation, the kind that starts with “I see you triggered R2; walk me through where you are.” That conversation will happen eventually anyway; the risk register is what makes it productive when it does.
Try this one cold: read the risk section to someone who has not seen the draft and ask them to name the most likely way the program fails. If they cannot answer in one sentence, the risk section is not specific enough.
Diagnostic 5: The team section does not answer Heilmeier Question 4
“Who cares? If you are successful, what difference will it make?” The Heilmeier formulation of this question is about downstream beneficiaries, not about the team. But in practice the team section is where reviewers form their durable impression of whether this team in particular can ship the proposed work.
The failure mode is the team section that reads as a list of resumes. Two paragraphs each on the PI, the co-investigator, the consultant. Degrees, prior positions, publications. The reviewer learns who these people are; they do not learn why these people are the right people for this work.
The pattern that lands: each team member’s paragraph leads with the specific outcome they have already shipped that maps to one of the Aims. “Dr. [NAME] led the [PRIOR PROGRAM] that produced the [SPECIFIC ARTIFACT]; the methodology used there is the direct antecedent of Aim 2.” The reviewer can then map each Aim to a person who has shipped its closest analog. The team becomes a function of the proposed work, not a list of biographies.
A test: highlight the first sentence of each team paragraph. If those sentences read as resume openers (“Dr. X is the Principal Investigator and has 15 years of experience in…”), the team section is not doing the work it needs to do. The strongest opening sentence names a shipped outcome that maps to a numbered Aim.
What this teardown adds up to
These five checks do not predict a score or an award. They make it easier to see whether the draft states its innovation, dependencies, success measures, risks, and team fit in terms a reviewer can evaluate against the actual criteria.
The catch is that authors of strong drafts almost never see these diagnostics in their own work. The reason is structural: the author wrote the draft and therefore knows what they meant. Knowing what you meant is the exact disqualifying condition for spotting where a reviewer at hour two of a 30-proposal stack will misread you.
The Federal Funding Readiness Sprint uses these checks alongside the applicable solicitation and agency review criteria. The output is a Heilmeier Gap Map that anchors material findings to proposal locations, separates blockers from competitiveness risks, and supports a revision roadmap and proceed, revise, redirect, or stop decision. Scope, optional funder-facing artifacts, and delivery schedule are set after intake.
If you have a draft that has already missed one cycle and you are wondering whether the resubmit will land, that is what the Sprint is built for. Not for SBIR mills that bill percent-of-award. For deep-tech founders and small businesses leaving non-dilutive money on the table because the draft cannot survive a red-team read.
Book a scoping call →. Thirty minutes is enough to tell whether your draft is at the stage where a Sprint is the right shape.