# Turn a market gap into a demand test

“Nobody seems to do this” is a research lead. A measurable buyer commitment is a stronger next step.

Published 2026-10-13 · startidea · AI-assisted, source-linked analysis. No hands-on vendor testing.

## Write a claim that could fail

Suppose research suggests that small contractors struggle to reconcile change-order evidence across tools. That is a hypothesis, not established demand. Rewrite it as a behavior you could observe: a defined type of contractor will commit a named person and a fixed amount of time to a pilot using real, authorized sample work. Avoid a criterion that every friendly conversation will pass.

## Choose the sample and target in advance

An illustrative test might contact ten qualified buyers through a human-authorized process and seek three pilot commitments. Those numbers are examples, not a scientifically established threshold. Record who qualifies, what counts as a commitment, how invitations are made, and when the test ends. Changing those rules after seeing results makes the conclusion harder to trust.

## Keep denominators and refusals

Three commitments from ten invited buyers differs from three commitments among three people who replied. Preserve the original denominator and record non-responses. For refusals, distinguish lack of urgency, no budget, a solved problem, and an unsuitable buyer. Do not recategorize inconvenient responses out of the sample without showing the change.

## Decide what the evidence supports

Meeting a small pilot target can justify the next test. It does not establish market size, repeat purchases, or willingness to pay at a future price. In startidea, plan an experiment before recording its aggregate results; keep results immutable and create a new test when the hypothesis changes. That leaves a readable trail of what you learned rather than a moving success story.

## Review worksheet

- Define buyer, commitment, target, sample, and end date.
- Preserve invitations, non-responses, and refusals.
- Choose the next test from the result, not from the desired story.

## Sources

- [Workspace experiment contract](/openapi.json)
- [Agent workspace loop](/api/agent)