Testing it: experiments that are allowed to say no

The short version. The point of testing is not to gather encouragement — it is to make a decision. That requires three things most idea-stage "validation" quietly lacks: a specific hypothesis, a threshold chosen before you see the results, and a genuine willingness to act on a no.

1. Write the bet down before you run it

"We believe [these people] will [do this costly thing]. We'll test it by [this] for [this long]. If fewer than [number] do it, the idea as stated is wrong." The number is the part people skip, because it is the part with teeth. Chosen afterwards, any result can be argued into a pass; chosen before, the test can actually teach you something.

2. Test with costs, not compliments

People are kind. They will say your idea is great and never think of it again — kind words are the fool's gold of idea-testing. Evidence is what people give up for the thing, and it comes in grades: paid money is the strongest; a deposit or pre-order next; time invested (a booked call kept, a form completed, a trial actually used) next; an email address is weak; a compliment is nothing. Design every test to ask for the highest-grade evidence you plausibly can. A "fake door" — a page selling the product as if it exists, with an honest "not ready yet, join the list" after the click — is often the cheapest way to ask strangers for a real signal.

3. Interview about their life, never about your idea

The moment you pitch, the data spoils — now they're being nice. Ask instead about what they already do: "When did this problem last come up? What did you do about it? What did that cost you? Have you tried to fix it — what happened?" Past behaviour is evidence; opinions about a hypothetical future are weather. If they have never tried to solve the problem, that is a finding too — it usually means the problem isn't painful enough to pay for.

4. Small numbers are fine; vague questions are not

You don't need statistics at this stage — five honest conversations with the right people will demolish or transform an idea faster than five hundred survey responses. What you need is sharpness: the right people (the ones from your one-sentence proposition, not your friends), questions about behaviour, and notes written down verbatim before memory rounds them up.

5. Let it fail properly

When a test fails, resist the two reflexes: softening the threshold ("nearly passed"), and re-running it until it passes. A failed test is a success of the method — it just saved you the months the build would have taken. The useful move is diagnostic: did the wrong people see it, was the promise unclear, or do the right people simply not care? The first two are fixable; the third is your answer.

6. Keep a log

One page per test: the bet, the threshold, what happened, what you decided. Memory is an unreliable narrator with a motive; three months in, the log is the only honest account of what you actually learned — and it is exactly what a future investor, partner or report will ask you for.

The whole thing, on one line

Write the bet and the pass mark first → ask for costs, not compliments → interview about their life, not your idea → let a failed test actually fail → and log it, because memory takes your side.