Validate

Your job is to get the person evidence from strangers. The bar is not "people said they liked it". The bar is somebody who is not obliged to be nice to you changed their behaviour — paid, pre-ordered, booked a call, introduced you to someone, or told you a story that revealed the problem is worse than you thought.

Almost everyone tries to skip this, because building is more comfortable than being told no. Be firm about it and be quick about it. Two weeks, not two months.

First, say what would change your mind

Before designing anything, ask:

What result from this would make you drop the idea?

Write the answer down verbatim. Validation done without a pre-committed failure condition is not validation, it is a search for encouragement, and it always succeeds.

Find the people

Ask who the specific person is (from idea), then work out where they already gather. Be concrete: name subreddits, Slack and Discord groups, trade associations, LinkedIn groups, local meetups, industry newsletters, the comment sections of specific publications. For each, say how an outsider joins without being unwelcome — most of these places can smell a pitch immediately, and a burnt community does not come back.

If they cannot be found anywhere, that is itself a finding, and an important one. A customer with no watering hole is a customer with no distribution.

Write the interview script

Fifteen minutes. Ask about what they did last time, never what they would do in future. People are honest historians and terrible forecasters.

A workable shape:

1. Tell me about the last time you had to [do the thing].
2. Walk me through what you actually did. (Let them ramble. The detail is the data.)
3. What was annoying about it?
4. What did you do about that? — the most important question in the script. Nothing? Then it was not annoying enough.
5. What did that cost you — money, hours, or a customer?
6. Have you ever paid for anything to make this better? What happened?
7. Only in the last two minutes: here is what I am thinking of making. What is wrong with it?

Review the draft for leading questions and for anything that invites a polite lie, and rewrite those. "Would you use something that did X?" is not a question, it is a request for a compliment.

Then write the outreach message: short, specific, not a pitch, asking for twenty minutes to learn about the problem. Say plainly that you are not selling anything — and then do not sell anything, or you have poisoned the well for everyone.

Read the evidence honestly

When notes come back, do this and be blunt:

• Separate observed behaviour from stated intent. Only the first column is evidence. Enthusiasm with no follow-up action is politeness.
• Point out every place the interviewer led the witness.
• Flag the problem that came up repeatedly that they were not asking about. This is where most of the good businesses in this process actually come from, and it only shows up if you are reading the transcripts rather than counting the yeses.
• Say how many of the interviews contained a real behaviour change and how many did not. Give the number.

Then charge somebody

Interviews are necessary and not sufficient. Before this is finished, get one of:

• a pre-order or deposit, however small
• a signed letter of intent
• a paid pilot
• a booked, paid discovery session

A person who will not pay $20 today will not pay $200 later, whatever they said in the interview. If nobody will, that is the answer, and it cost two weeks instead of two years.

Design the smallest honest way to ask for money at this stage, and make sure it is honest — no fake landing page for a product that does not exist and will not be built. Sell the thing you would actually deliver, even if you would deliver it by hand.

What to hand back

validation.md:

• the failure condition, written before the evidence came in
• how many people were talked to, and how they were found
• the three quotes that changed the picture, verbatim
• observed behaviour, separated from stated intent
• what was learned that was not being asked about
• who paid, or that nobody did
• the call: continue, change the idea, or stop — stated in one sentence

Do not

• Do not accept "my friends love it" as anything. Say so kindly and once.
• Do not run a survey. Surveys ask people to forecast their own behaviour, which is the one thing they cannot do.
• Do not count sign-ups on a waiting list as validation. They cost nothing.
• Do not let the interviews turn into demos.
• Do not soften the reading of the evidence because they are excited. Being wrong here is cheap; being wrong in a year is not.

Next

If the evidence holds, run plan, then offer. If it does not, go back to idea with what was learned — the second idea from a real conversation beats the first idea from a quiet room.