Validate Demand With a Behavioral Commitment

Validation starts after you can name a specific customer and painful moment. It ends only when a qualified person does something that costs them a little to keep moving with you.

This guide is for a domain professional with a customer/problem hypothesis but no behavioral evidence. Conversations matter, but compliments do not validate demand. Traffic, likes, survey enthusiasm, waitlist signups without follow-through, and AI-generated market analysis also do not count as commitment.

Your output is an evidence ledger that separates what people said from what they did.

Decision

Decide which uncertainty could stop the offer even if everything else works. Usually it is one of these:

• the problem is not frequent or costly enough;
• the person you can reach does not own the decision;
• the current workaround is acceptable;
• the promised change is not valuable enough to justify switching; or
• the buyer will not commit time, access, a referral, a pilot, or payment.

Pick one uncertainty. A validation test that tries to prove the entire business at once produces ambiguous evidence.

For example, suppose your synthetic hypothesis is that independent property managers need a cleaner way to collect maintenance details. The riskiest uncertainty may not be whether maintenance is frustrating. It may be whether a manager will grant access to a real, sanitized request so you can test a manual intake process. That access is a behavior. “Sounds useful” is not.

Action

Run two connected tests: last-instance interviews and a commitment ask.

1. Conduct last-instance interviews

Speak with people who match the hypothesis. Ask them to reconstruct the last time the problem happened:

1. What triggered the situation?
2. What did you do, in order?
3. Who else became involved?
4. What information or approval was missing?
5. What did the workaround cost or delay?
6. What have you already tried to improve it?
7. What happens if nothing changes?

Stay with the event. If the person offers a general opinion, ask for the most recent example. Do not pitch while you are still learning the sequence.

Record only information you are allowed to retain. Remove names and identifying details before using AI to organize notes.

2. Ask for one concrete commitment

Match the ask to the uncertainty and the maturity of the relationship. A valid commitment can include:

• scheduling a working session and showing up;
• granting bounded access to a sanitized example;
• introducing you to the actual decision-maker;
• referring another qualified person;
• signing a narrowly scoped pilot;
• reserving a delivery date; or
• paying a deposit or the full price.

Do not jump straight to payment when access or a working session would test the current uncertainty more directly. But do not remain forever in free conversations to avoid hearing “no.”

Use a direct sentence:

The next useful test is [specific step]. It will require [time, access, referral, pilot, or payment] by [date]. Would you like to do that?

Then stop talking. Record the response and what happens by the date.

Build the evidence ledger

Use one row per person with these columns:

• qualification: why this person matches the hypothesis;
• last instance: when the problem most recently occurred;
• consequence: what changed because of it;
• current workaround;
• previous attempts;
• commitment requested;
• commitment accepted or declined;
• behavior observed by the deadline;
• objection or condition;
• next decision: continue, revise, or stop.

Keep “accepted” separate from “completed.” A person who agrees to a meeting but does not schedule or attend has not supplied the same evidence as someone who follows through.

Interpret the result

Look for behavior patterns, not a winning quote. Continue when at least one qualified person completes a commitment tied to the risky assumption and the interviews show a coherent painful moment. Revise when people experience the problem but resist the specific commitment for a repeated, testable reason. Stop when the consequence is weak, access is unavailable, or the workaround is consistently good enough.

One commitment is not proof of a scalable market. It is the minimum behavioral signal to justify the next bounded investment.

What AI can help with

AI can draft neutral questions, role-play difficult follow-ups for practice, convert sanitized notes into ledger rows, compare objections, and highlight where your records mix statements with actions. It can help you write competing interpretations so you do not automatically favor the optimistic one.

Synthetic role-play is useful before an interview. It is never evidence in the ledger.

What AI must not decide

AI must not count its own generated responses, web summaries, or persona simulations as demand. It must not label a polite comment as commitment. It also must not decide that a person is qualified when you do not understand their role, authority, or context.

The human operator owns the ask, the relationship, the privacy boundary, and the continue/revise/stop decision.

Completion signal

Validate is complete when your ledger contains:

• last-instance evidence from qualified people;
• an explicit commitment request;
• the observed behavior after that request; and
• at least one completed commitment involving time, access, a referral, a pilot, or payment.

If no qualified person commits, do not disguise that result with more content, traffic, or AI research. Revise the person, problem, consequence, or ask and test again. If someone does commit, carry that evidence into Build and define the smallest useful offer you can actually deliver.