{/* Schema recommendation: BlogPosting + FAQPage + Table.
- BlogPosting: author Nalin Vahil, datePublished and dateModified 2026-08-21. Emitted by the post template from frontmatter.
- FAQPage: emitted from the frontmatter faq block. Do not duplicate the FAQ in the body; the template renders it below the article.
- Table candidates for rich extraction: the cross-agency ratio table in "The ratio table: SBIR budget percentages by category." Internal linking: first SBIR mention links to /insights/sbir-guide-for-startups; companion pieces linked inline (/post/sbir-budget-justification-strategy, /post/how-to-structure-your-sbir-budget-a-guide-to-allowable-costs-rates); CTA links to /roadmap-intake. */}
Reviewers do not start with your budget justification. They start with your ratios: your SBIR budget percentages, from direct labor down to indirect costs, get pattern-matched before anyone reads a sentence of narrative.
A defensible federal R&D budget is one whose category ratios sit inside the bands reviewers consider normal, with a named justification for any line that cannot. Here is the shape of those bands: direct labor runs 40-60% of total cost, equipment 5-20%, travel 2-5%, subcontracts 10-30%, and indirect costs 10-30%. Ratios outside those bands draw reviewer scrutiny, and a few of them, like the SBIR 33% subcontract cap, are hard compliance limits rather than judgment calls.
Cada's published budget justification guide covers the narrative: how to write the story behind each line. This piece publishes the numbers. These are the reasonableness bands we use when reviewing budgets before submission, built from writing hundreds of proposals across 30+ agencies.
Why reviewers check your ratios before your narrative
Cost reasonableness review is pattern-matching. A reviewer who has scored 40 budgets knows what a working R&D budget looks like, and they scan yours against that template in about 30 seconds.
A budget that is 70% subcontracts raises one question instantly: who is actually doing the work? A $95,000 request on a program that typically funds $1.5M projects signals you have not scoped the problem. Neither impression gets corrected by a well-written justification, because the justification only gets a careful read if the shape looks sane.
That is the practical order of operations. Ratios first, narrative second. Get the percentages into normal bands and your justification gets read with trust instead of suspicion.
The ratio table: SBIR budget percentages by category
This is the cross-agency reasonableness table. To be clear about sourcing: no agency publishes these bands. They are practitioner norms, the ranges Cada uses to screen budgets before submission, and the red-flag thresholds are where we have seen reviewer scrutiny reliably begin.
| Category | Typical range (% of total) | Red flag | What the flag says to a reviewer |
|---|---|---|---|
| Direct labor | 40-60% | Below 30% or above 70% | Below 30%: who is doing the work? Above 70%: why no materials or equipment? |
| Fringe benefits | 15-25% of direct labor | Above 35% of labor | Rate needs documentation (payroll taxes, health insurance, retirement) |
| Equipment | 5-20% | Above 30% | Is this R&D or an equipment purchase? |
| Materials and supplies | 5-15% | Vague line items | "Lab supplies, $40,000" gets questioned; itemized categories do not |
| Travel | 2-5% | Above 10% | What are you flying to, and why does the research need it? |
| Subcontracts and consultants | 10-30% | Above 33% (SBIR hard cap) | Are you the prime or a pass-through? Over 33% is non-compliant, not just suspect |
| Indirect costs (F&A) | 10-30% | Above 40% | Why is overhead eating the research budget? |
| Fee or profit | 5-7% | Above 7% | Above 7% invites negotiation at NSF and DoD; NIH caps fee at 7% of total costs |
A budget where every category sits inside these grant budget reasonableness ratios will almost never draw a cost-related reviewer comment. One category outside a band is fine if the justification addresses it head-on. Two or more, and the budget itself becomes the story reviewers tell about your application.
Hard caps by agency: where the ratios become rules
The bands above are norms. Each agency also has hard limits where a wrong number means non-compliance, not scrutiny. Here are the four programs we see first-time applicants target most. (For the mechanics of which costs are allowable in the first place, see the guide to structuring an SBIR budget and its allowable costs.)
NIH SBIR Phase I
NIH's Phase I budget guideline is $314,363, the SBA-adjusted amount effective FY2025, and some topics carry approved waivers allowing higher requests (source: seed.nih.gov). Most Phase I applications use the modular format, which applies at up to $250,000 in direct costs per year.
Three structural rules matter most:
- Subcontract cap: the small business must perform at least two-thirds of the Phase I research. Subcontracts plus consultants cannot exceed 33% of the total (source: SBA SBIR/STTR Policy Directive).
- STTR split: if you apply STTR instead, the small business must perform at least 40% and the partner research institution at least 30%.
- Fee: NIH allows a fee of up to 7% of total costs. Take it as a fee line, not buried in indirect.
On the SBIR indirect rate for startups: if you have no federally negotiated rate, NIH lets you request up to 40% of modified total direct costs without submitting justification (source: seed.nih.gov). Modified total direct costs excludes equipment and the portion of each subaward above $25,000, so the effective ceiling is lower than 40% of your grand total.
One correction worth stating plainly, because outdated answers still circulate: NIH announced a 15% cap on indirect cost rates in February 2025. Federal courts permanently blocked that policy, the injunction was upheld on appeal in January 2026, and no cap is in effect. The 40% SBIR provision stands.
NSF SBIR Phase I
NSF runs an all-in cap: $305,000 total, covering direct costs, indirect costs, and fee combined (source: NSF solicitation 24-579, Section V.C).
- Fee: up to 7% of the budget's Line J (direct plus indirect), entered on Line K, usable for any purpose.
- Outsourcing: subawards plus consultants cannot exceed one-third of the total, roughly $101,667 at the cap.
- PI effort: at least one calendar month per six months of project duration.
- Indirect: reference your negotiated rate agreement if you have one. If not, the government-wide de minimis rate under 2 CFR 200 is 15% of modified total direct costs, or you can propose a justified estimate.
AFWERX SBIR (Open Topic)
AFWERX is the outlier: $75,000 total for an SBIR Open Topic Phase I and $110,000 for STTR, over a 90-day period of performance. Ceilings change by cycle, so verify against the current solicitation on DSIP before budgeting.
The cost volume is entered as a web form, category by category. The rate structure we see clear review without questions, based on Cada's pre-submission reviews of AFWERX budgets:
- Fringe: 15-25% of direct labor for a typical startup.
- Overhead: 0-50% of labor plus fringe; rates under 20% draw less scrutiny for companies without an established rate.
- G&A: 5-20% applied to the subtotal.
- Profit/fee: 5-7% is typical for DoD contracts. Above 7% is uncommon and invites negotiation.
Two things founders miss. The PI must spend more than 50% of their working time on the project during those 90 days. And AFWERX contracts carry DCAA audit obligations from award, meaning daily timesheets and segregated cost tracking, not reconstructed records at audit time.
ARPA-H
ARPA-H expects the budget inside the proposal itself: the Basis of Estimate section of the 6-page Solution Summary, with a cost table showing each category as a percentage of total. The ARPA-H budget basis of estimate is where the ratio table above came from; it is the one place Cada's norms are applied as an explicit pre-submission screen.
Scale is the biggest adjustment for SBIR-experienced founders. Typical base periods run $1M-$5M. In our review experience, below roughly $500,000 reads as too small to matter for ARPA-H's mission, and above $10M reads as too large for an initial award.
Frankly, ARPA-H is still a young agency and its norms are still settling. Treat these bands as the current pattern, not settled policy.
The five federal grant budget red flags that trigger scrutiny
Across agencies, five ratio flags account for most cost-related reviewer pushback. Each one is really an unspoken question.
- Direct labor under 30%. Who is actually doing the work? Usually this means too much outsourcing or an equipment-heavy budget. Fix: move scope in-house or explain exactly which people execute which aims.
- Equipment over 30%. Is this R&D or a procurement request? Agencies fund research, not capital equipment. Fix: lease, use shared facilities, or justify why the instrument is inseparable from the research.
- Subcontracts approaching 33%. Are you the prime or a pass-through? At 32% you are compliant but will be asked. Fix: keep a visible margin below the cap, and name what your own team uniquely does.
- Indirect above 40%. Why is overhead eating the research budget? Above NIH's no-questions threshold, expect to defend the rate. Fix: negotiate a rate, or restructure costs so more of them are direct.
- Travel above 10%. What are you flying to? Fix: tie every trip to a program requirement (kickoff, testing site, program review) with named destinations and GSA-rate math.
For any flag you cannot restructure away, write the justification sentence before the reviewer asks the question. A named reason beats a normal-looking ratio you cannot defend.
How to sanity-check your own budget in 20 minutes
This is the same screen we run on every budget before submission. Four steps:
- Compute the percentages. Divide each category by the grand total. Spreadsheet, ten minutes.
- Compare against the ratio table. Mark every category outside its typical band, and separately mark anything past a red-flag threshold.
- Restructure or justify. For each marked category, either move money until it sits in band, or write one sentence naming the specific project reason it cannot.
- Check the hard rules as pass/fail. Award ceiling, subcontract cap, PI effort, fee limit. These are binary, no judgment involved.
Here is a fictional example. Meridian Assay Labs (an illustrative company, not a client) budgets a 12-month NIH SBIR Phase I at $300,000: direct labor $120,000 (40%), fringe $24,000 (20% of labor), equipment $70,000 (23%), materials $20,000 (7%), travel $6,000 (2%), one consultant $10,000 (3%), indirect $50,000 (17%).
The screen catches one item: equipment at 23% sits above the typical 5-20% band, though under the 30% red flag. The fix is not restructuring, it is one sentence: "Aim 2 requires a dedicated hypoxia workstation ($52,000) at a sampling cadence the regional core facility cannot provide, and no lease option exists for this configuration." Flag anticipated, question answered, budget defensible.
Run your numbers before a reviewer does
The whole point of publishing these SBIR budget percentages is that you can screen your own budget in 20 minutes: compute the ratios from direct labor through indirect, compare against the table, fix or justify what falls outside.
If you want a second set of eyes, we do a free budget sanity check. Send your category percentages and target program, and we will tell you which ratios will draw scrutiny and whether a justification or a restructure is the right fix. Straight answer, no pitch, no obligation.
For the narrative side of the budget, the companion piece covers what makes a budget justification competitive, not just compliant.
Sources
- NIH SEED / seed.nih.gov -- Phase I budget guideline ($314,363), the 40% MTDC indirect provision for applicants without a negotiated rate, and the 7% fee allowance
- SBA SBIR/STTR Policy Directive via sbir.gov -- the two-thirds self-performance requirement and 33% subcontract limit for Phase I, and the STTR 40%/30% split
- NSF America's Seed Fund -- the $305,000 all-in cap, Line J/Line K fee mechanics, and outsourcing limits (solicitation 24-579, Section V.C)
- 2 CFR 200 -- the government-wide 15% de minimis indirect rate on modified total direct costs
- AFWERX and ARPA-H -- award structures and proposal formats; per-cycle ceilings verified against the live solicitation
- The ratio bands, red-flag thresholds, and AFWERX/ARPA-H review norms are Cada's own practitioner benchmarks from pre-submission budget reviews (hundreds of proposals across 30+ agencies), not published agency policy
Award ceilings, indirect-rate provisions, and fee caps reflect agency guidance as of August 2026. Solicitation terms change; verify against the live solicitation before budgeting. All company examples are fictional and used for illustration only.