There is no universal answer, and anyone who gives you one is describing a single agency. Some SBIR programs are built to fund concepts that have barely left the whiteboard; others sit in research cultures where a proposal without supporting data is not competitive. The first thing to establish is which kind you are writing for, and the honest way to establish it is to read the agency's solicitation and ask a program officer.
Why expectations differ so much
The agencies fund different things. Health-related programs review alongside academic research and reviewers are accustomed to seeing figures, so preliminary results carry real weight there. Defense components often solicit against a stated operational need, and a credible engineering path plus relevant prior work can matter more than bench data. Science-agency programs frequently exist precisely to fund early, unproven ideas. None of that is a rule you can rely on across a whole agency, because individual solicitations vary within one agency, but it explains why advice from a founder who won at one agency can be actively misleading at another. Start with choosing the right agency and read the specific topic you are answering.
What counts as preliminary data
More things count than founders assume. Bench measurements from a crude rig count. Simulation results count if the model is described and its assumptions are stated. A working breadboard prototype that demonstrates one mechanism counts, even if it looks nothing like the product. Published results from your own prior work count. So does a documented negative result that narrowed the design space, which many founders leave out because it feels like an admission.
What does not count is an assertion. "Initial testing has shown promising results" with no method, no conditions and no numbers is worse than saying nothing, because it invites the reviewer to assume the results were unimpressive. Replace it with the smallest honest specific: "On a bench rig, three samples held the seal past 500 cycles at room temperature; behavior at elevated temperature is unmeasured and is Objective 1."
Handling the result you do not have
Say so, and turn the gap into the reason for the project. A missing result stated plainly and assigned to a named task reads as competence. A missing result papered over with adjectives reads as either ignorance or evasion, and reviewers who work in your field will spot which key measurement is conspicuously absent.
If the gap is large enough that no agency will fund you yet, the answer may be a smaller source first — a state program, an internal build, or the routes in proof-of-concept funding. Applying too early and being scored on an empty evidence base burns a cycle and can prejudice a resubmission.
Medical-device applicants face the steepest expectations of all, which is worth reading about in NIH grants for device startups. Whatever your field, the cheapest way to calibrate is a short conversation before you write, and talking to a program manager covers how to have it.
Projects House builds the early rigs and prototypes that produce data worth citing. Tell us what you need to prove through the contact form.