What makes a good research-institution partner on a federal bid

The rule that creates the relationship

Every STTR proposal has a second organization inside it, because the statute says so. Under the eligibility rules at 13 CFR 121.702 and the SBA policy directive governing both programs, an STTR award is a cooperative research effort between a small business and a single research institution. The small business performs at least 40 percent of the work. The research institution performs at least 30 percent. The remaining 30 percent can go to either party or to a third party. Five agencies run STTR against a 0.45 percent set-aside of extramural research budgets: the Department of Defense, the Department of Energy, the National Institutes of Health, NASA, and the National Science Foundation. That one paragraph of law is why a firm like ours and a university lab end up on the same submission.

It is also why picking the wrong partner is expensive. A research institution on an STTR is not a vendor supplying a component. It is a co-author of the technical approach, a named entity in the personnel section, a signatory on an intellectual-property agreement, and a budget line a contracting officer will read closely. When the partnership is right, the lab supplies the piece of the method that would otherwise take two years to earn. When it is wrong, the proposal reads like two documents stapled together, and reviewers say so.

We run an STTR lane alongside our SBIR work, so we have a standing view of what separates the two outcomes. Five qualities predict most of it. None of them is scientific reputation.

Quality one: the sponsored-programs office answers

The SBA policy directive requires the small business and the research institution to execute a written allocation-of-rights agreement covering intellectual property and the right to carry on follow-on work. Components differ on timing. Some want the executed agreement before award; several want at least a signed model or a draft at submission. Either way, that document travels through the university's office of sponsored programs, usually with a stop at the technology transfer office, and both have queues.

Turnaround is the number that matters. A solicitation opens and closes in a window measured in weeks, and every day the paperwork sits is a day out of the technical writing. A lab where the principal investigator can name the research administrator, forward one email, and get a reply the same week is worth more to a bid than a lab with a longer publication list and a silent front office. That is arithmetic about a deadline, not a comment on scholarship.

What a responsive office looks like: a named administrator for the college, a stated turnaround on a subaward package, an allocation-of-rights template the institution has signed before, and a willingness to review terms during the proposal window rather than after selection. Some universities keep an administrator dedicated to SBIR and STTR. Those campuses are the easiest to bid with, and their PIs usually know it.

Quality two: a PI with time, and a calendar to prove it

STTR lets the principal investigator be employed primarily by either the small business or the research institution. SBIR does not; there the PI's primary employment has to sit with the firm. That flexibility is the single biggest reason faculty find STTR attractive, and it is also where the most common failure begins. A faculty member with four active awards, a full teaching load, and eight graduate students is genuinely enthusiastic in April and genuinely unreachable in October.

So we ask for calendar, not enthusiasm. Which weeks. How many hours per month. Which graduate student or postdoc is assigned, by name, and at what appointment percentage. Whether a sabbatical, a chair rotation, or a competing renewal lands inside the period of performance. A PI who answers those in one paragraph has already done the planning that makes the work happen. A PI who answers "we will make the time" has told us something too.

Phase I periods of performance in the DoD program commonly run six to twelve months at low six-figure funding, and 30 percent of that does not buy many hours. One measured result on real data, delivered on schedule by a student named in the proposal, beats an ambitious scope that quietly depends on the PI finding evenings.

Quality three: alignment that is real, not adjacent

Reviewers read the research institution's role directly. They are looking for whether the 30 percent sits on the critical path or decorates the margin. A lab that supplies the core method, the instrument, the reference dataset, or the validation that nobody else can produce reads as necessary. A lab that supplies "guidance and technical consultation" reads as a compliance move to satisfy the statute, and it costs points on both the technical and the personnel factors.

The failure mode is subtle because it usually looks like a match. The lab's expertise is one abstraction away from the requirement. A group with a strong record in graph learning over citation networks is not automatically the right group for entity resolution across freight records, even though both are graphs. The methods rhyme; the failure modes, data pathologies, and evaluation criteria do not.

Our test is cheap. Before anyone signs anything, we write one paragraph of the technical approach with the lab's method inside it, naming the input, the operation, and the measured output. If it comes out clean, alignment is real. If it needs three hedges to survive, alignment is adjacent, and we would rather say so in week one than find out in month four.

What we ask on the first call	A partnership that works	A partnership that stalls
Who signs the allocation-of-rights agreement, and by when?	A named administrator and a date inside the proposal window	"Legal will look at it once we hear something back"
How many PI hours fall inside the period of performance?	A number by month, plus a named student or postdoc	"Whatever the project needs"
What exactly does the lab deliver?	A dataset, a trained model, a measured result, a written section	Advice, review, and general oversight
May we name the PI and the institution in the volume?	Yes, with a short bio and a letter on letterhead	"Better to wait until after selection"
Has anything from this lab been licensed or fielded?	A license, a spinout, a prior Phase II, a tool other groups run	Citation counts and a journal list
What is the negotiated F&A rate on a subaward?	A rate, a base, and whether off-campus applies	Unknown until the budget office is asked

Quality four: a transition record beats a citation count

Bayh-Dole (35 U.S.C. 200 through 212, implemented at 37 CFR 401 and carried in the standard patent rights clause at FAR 52.227-11) lets universities elect title to inventions made with federal funding, on the condition that they work to bring those inventions into use. That is why every research university has a technology transfer office and a licensing history, and that history is proposal evidence.

What we look for is narrow and specific: a prior Phase II executed with a small business, a license to a company that still exists, a piece of software other groups actually run, a measurement technique a government laboratory adopted, an instrument design that left the building. Any of those facts belongs in the commercialization narrative, and any of them changes how a program manager reads the bid. We have written elsewhere about how a lab's transition record becomes proposal evidence, and the mechanics matter more than the volume.

Citations do not do this work. A reviewer scoring commercial potential is asking one question about the lab: does the method leave the campus? A lab that has answered it before will answer it again. A lab that has not is still workable, but the transition story then rests entirely with us, and we plan for that from the start.

Quality five: willing to be named

We do not put an unnamed person in a federal proposal. Not "a senior researcher at a leading university." Not "a graduate assistant to be identified." A named principal investigator with a short biography, a named student or postdoc where the institution permits it, and a letter of support on letterhead signed by someone with the authority to sign it.

That looks like an administrative preference. It is a scoring rule. Personnel is a scored element in most SBIR and STTR evaluations, and an evaluator cannot credit qualifications that belong to nobody. Every unnamed seat in a proposal is a seat the evaluator scores at zero, and every hedged commitment reads as a commitment that has not been made.

Willingness to be named is also a proxy. A PI who says yes on the spot has usually cleared the effort internally, checked the conflict-of-interest posture, and confirmed the department supports the subaward. A PI who wants the arrangement kept quiet until an award arrives is telling us the decision has not been made at the level where it has to be made.

The money conversation belongs in week one

Facilities and administrative costs decide how much science the 30 percent actually buys. Under 2 CFR 200.414, an institution charges its federally negotiated indirect rate, and on-campus organized research rates at major universities commonly land between 50 and 60 percent of modified total direct costs. Run the arithmetic on a Phase I. If the award is $150,000 and the institution takes 30 percent, that is $45,000 gross. At a 55 percent rate applied to modified total direct costs, roughly $29,000 remains for direct effort. That is a few months of a graduate student, part of the tuition remission, and a slice of faculty summer salary.

None of that is a problem. It becomes a problem when it surfaces four days before close. Some institutions will apply an off-campus or reduced rate to STTR subawards; some will not; some cannot without a waiver from their cognizant agency. All three answers are workable if we have them early, because the scope can be sized to the money instead of the money being asked to stretch to the scope.

Two costs get forgotten. Tuition remission is often excluded from the indirect base but is still real money against a small budget. Core-facility and instrument time is charged at published rates and needs its own line. Ask for both numbers the same day the rate is requested.

What makes these partnerships fail

Publication rights against controlled work. Universities protect the right to publish, and the fundamental research exclusion described in NSDD-189 and reflected in the export rules at 15 CFR 734.8 is what keeps most campus research outside export licensing. When a solicitation touches export-controlled data or a controlled deliverable, that exclusion stops applying, and the institution's standard publication clause collides with the sponsor's restrictions. The fix is structural, not legal: scope the institution's share onto the unrestricted portion of the work and keep the controlled portion inside the firm.

A subaward treated like a grant. A federal subcontract has deliverables, dates, invoicing terms, and reporting obligations that flow down. Labs accustomed to grant rhythm sometimes read a milestone schedule as aspirational. One planning call that walks the schedule line by line, with the administrator in the room, prevents most of this.

The silent middle. Engagement during proposal writing and radio silence in month three is the most damaging pattern, because it surfaces after the deliverable clock has started. A standing 30-minute call on the calendar from day one, with the student attending, costs almost nothing and catches drift while it is still cheap.

Intellectual property negotiated after selection. Background intellectual property, field-of-use, license options, and follow-on rights are far easier to settle when nobody has won anything yet. Once an award exists, positions harden. Settle it in the allocation-of-rights agreement during the proposal window.

The lab quietly bidding with two firms on the same solicitation. Nothing prohibits it, and it puts both proposals at risk. Raise it in the first call, and if the answer is complicated, our note on exclusivity in teaming agreements covers the ways to structure it without anyone feeling boxed in.

Data rights confusion. SBIR and STTR data rights under DFARS 252.227-7018 protect qualifying technical data and software for 20 years from the date of award. Institutional background intellectual property, open-source obligations, and pre-existing licensed code sit outside that protection. Write down which category each asset falls into before the first line of code is shared.

What a lab can send us in one email

• The PI's name, and two lines describing the method
• The agency and solicitation number under consideration, or "none yet"
• Whether the lab has run an STTR or a small-business subaward before
• The sponsored-programs contact and the usual subaward turnaround
• The federally negotiated F&A rate, and whether an off-campus rate can apply
• One prior result the method produced, with the number attached
• Whether the PI and the institution can be named in a submitted volume

Common questions from labs

Bottom line

The best research partner on a federal bid answers quickly, commits real hours to a named person, owns a piece of the method the requirement actually needs, has moved something off campus before, and will put its name on the page. Scientific standing matters, and it is not the deciding variable. All of it is knowable from one call and one email exchange, which is why we run the screen before a single hour of writing goes in.

Frequently asked questions