Budget engineering as separate lines tied to milestones, not as one contractor lump sum. In education-technology Phase II proposals, engineering usually takes 25–40% of the award. You land near the top of that range when the pilot needs a production platform, research-grade logging and district IT review, not a stretched Phase I prototype.
How to budget software engineering in an SBIR Phase II proposal
How much of the award should go to engineering
This is a general market range, not a rule from any agency. Expect 25–40% of a Phase II budget to go to software engineering when the product is software used by teachers, students or lab staff. The rest goes to the PI’s and research staff’s time, the efficacy study, participant incentives, evaluation, commercialization work and indirect costs.
Four things decide where you fall in that range:
| Factor | Lower end (about 25%) | Upper end (about 40%) |
|---|---|---|
| State of the Phase I prototype | Built on a maintainable stack, with some tests | Notebook or student-built code that has to be rebuilt |
| Who logs in during the pilot | A few classrooms, all supervised by the research team | Dozens of teachers across districts, using it on their own |
| Data requirements | Summary outcomes exported at the end | Event-level logging that has to hold up in an efficacy analysis |
| AI components | None, or one fixed model call | An orchestration layer comparing prompts or models across study conditions |
If two or more rows describe you at the upper end, budget near 40%. Teams that budget 15% for software usually end up taking the shortfall out of the study.
The budget lines to include
Reviewers and program officers trust a budget more when they can see what each dollar buys. Break engineering into these lines in the budget justification, each tied to a milestone in the work plan:
- Architecture and hardening (15–25% of engineering): decide what to keep from the prototype, then add authentication, roles, a real database and automated tests.
- Core features for the pilot (30–40%): the teacher and student workflows the study depends on, and nothing else.
- Research data logging and export (10–15%): event logging with stable identifiers, de-identified exports and a data dictionary your analyst can work from.
- Privacy, security and district documentation (8–12%): the technical disclosure package, the data flow diagram, answers to the district’s security questionnaire, and support for the data use agreement.
- Testing and deployment (8–12%): staging and production environments, a deployment pipeline, and load testing at pilot scale before teachers log in.
- Pilot support (10–15%): on-call fixes during study windows, plus content or configuration changes between cohorts.
Hosting, AI model usage and third-party services belong in a separate other direct costs line, not in engineering labor. Cloud hosting for a pilot of a few hundred students is usually modest. AI model costs depend on calls per session multiplied by sessions, so estimate them from the study design rather than guessing.
Worked example: a $1M Phase II with 30% for engineering
Take a hypothetical 24-month Phase II of $1,000,000 that puts $300,000 into engineering. At $85 an hour, that buys about 3,530 hours of engineering spread across five milestones:
| Milestone | Months | Deliverable | Budget | Approx. hours |
|---|---|---|---|---|
| 1. Hardening | 1–3 | Architecture decision, rebuilt core, auth, tests | $60,000 | 706 |
| 2. Pilot platform | 4–8 | Teacher and student workflows, logging, admin tools | $110,000 | 1,294 |
| 3. Pilot readiness | 9–10 | District disclosure package, load test, staging dry run | $35,000 | 412 |
| 4. Pilot support | 11–18 | Fixes during study windows, cohort changes | $55,000 | 647 |
| 5. Efficacy data and handoff | 19–24 | Final exports, data dictionary, documentation | $40,000 | 471 |
Each milestone lines up with an aim or a go/no-go point in the proposal. That lets you report progress in the program officer’s terms, and it means a fixed-price engagement can be invoiced against deliverables instead of timesheets.
Milestone 3 is the one teams leave out. District IT review can take a full semester if the documents arrive late. Budget the disclosure package before the recruitment window, not after it.
Staff, consultant or subcontract
Under SBA’s SBIR policy, the small business has to perform at least half of the Phase II work, so subcontracts and consultants together can make up no more than half. STTR has its own split between the small business and the research institution. Check your solicitation for how the agency measures this, then pick a structure that leaves room for the evaluation subcontract as well.
- A graduate student or postdoc keeps building when the pilot is small and supervised, the code is already maintainable, and someone senior reviews it. It stops working once a teacher is using the tool without the researcher in the room.
- An engineering firm on fixed-price milestones suits a defined path from prototype to pilot with known deliverables.
- A retainer or hourly arrangement suits the pilot period, when the work arrives as bugs and cohort changes rather than features.
On data rights: the small business should own the code, the data and the SBIR data rights. A vendor builds under a work-for-hire or assignment clause and should not keep a license that limits your Phase III. Your counsel or sponsored-programs office signs off on the actual contract language.
What a program officer reads in the engineering budget
A Phase II budget with one line reading "software development, $150,000" reads as a research project that hopes software will appear. A budget with milestones, a testing line, a logging line and pilot support reads as a company that can ship and sell. That difference matters in review because Phase II is meant to lead to a product.
Reliability is also research validity. If saves fail, sessions drop or logging misses events in some classrooms, those students drop out of your analysis and the efficacy study loses power. Clean, complete logging is what makes the outcome analysis possible, so put it in the budget as its own line.
We have done this before. We built the prototype that helped an EdTech team win $1M in Phase II funding, and then we built the production platform. On another project, we took a research prototype to a district-approved classroom pilot in under 30 days, with 20 teachers in two districts. The research findings were the teams’ own; our job was software that held up under the study.
Frequently asked questions
What percentage of an SBIR Phase II budget should go to software development?
For education-technology products, 25–40% of the award is a common range. You land near the top when the Phase I prototype has to be rebuilt, the pilot runs across several districts, or the study needs event-level logging or an AI orchestration layer. That range is general market experience, not an agency rule.
Can I hire an outside software firm with SBIR Phase II funds?
Yes. The firm is budgeted as a subcontract or consultant, and under SBA policy the small business has to perform at least half of the Phase II work. Check your agency’s solicitation for how that share is calculated, and leave room for evaluation subcontracts as well.
Should hosting and AI model costs be in the engineering budget?
No. Put them in other direct costs. Hosting for a pilot of a few hundred students is usually modest. Estimate AI model costs from the study design: calls per session multiplied by the expected number of sessions.
Who should own the code an outside developer writes for our SBIR project?
The small business should own the code, the data and the SBIR data rights, with the developer working under a work-for-hire or assignment clause. A vendor should not keep a license that limits your Phase III commercialization. Your counsel or sponsored-programs office should approve the contract language.
How long does district IT review take, and should I budget for it?
Plan for anything from a few weeks to a full semester, depending on the district and how complete your documentation is. Budget a separate engineering milestone for the technical disclosure package, data flow diagram and security questionnaire, and finish it before teacher recruitment starts.
Planning a Phase II budget or a pilot?
We offer a free 30-minute review of your proposal’s technical section or your pilot plan. Bring the prototype’s repository or a demo, your pilot timeline, and the funder’s solicitation requirements. Start a conversation.