Twenty people said yes. A month later a different twenty said yes. Nothing has contradicted the idea, and the absence of contradiction has started to feel like evidence. This is usually the point where a founder stops asking and starts building.
It is also the point where the process is least trustworthy, and the reason is not obvious from the inside.
Why does everyone agree with my startup idea?
Because agreement is the cheapest thing a person can hand you, and because the people you can reach easily tend to resemble each other.
Early feedback comes from a founder's own surface area. Friends, colleagues, followers, the people who reply to a post, the forum where new ideas get a warm reception by default. None of them are lying. They are all standing on the same tilt.
Ask forty people from that group and you have not collected forty data points. You have collected one behaviour, sampled forty times.
What is the difference between reproducible and correct?
A data team recently described running the same churn analysis forty times over six weeks. Four different people triggered it. Every run matched to the decimal. All forty were wrong, because of a bad join upstream.
Reproducibility did exactly what it promised. It carried the same error faithfully every single time, and the consistency is what stopped anyone from looking.
Idea validation has the same failure mode without the audit trail. A process that returns the same answer is consistent. Consistent is not the same word as correct.
Does a bigger sample fix biased feedback?
No, and this is the most expensive misunderstanding in early validation.
If the bias lives in who you can reach, adding more of the same people multiplies it instead of diluting it. Fifty conversations with people who want you to succeed is fifty times the same conversation, and the larger number makes the result feel more solid rather than less.
Narrowing the group helps, but only in one direction. A tight, well defined group is useful because disagreement inside it means something. It is not useful as a way of reaching yes faster.
What does a real demand signal look like?
The signals worth trusting have one thing in common. They cost the other person something before you showed up.
A workaround somebody already built. A spreadsheet maintained by hand every Friday. A budget line already going to a worse solution. A deposit. A subscription cancelled somewhere else.
Stated interest costs nothing. Signing something costs almost nothing either, which is why a signed agreement can sit for four months and produce a single order. Said is not paid, and that gap does not close as a company gets bigger.
How do I know whether my validation process is broken?
Look back over the last three months of what your research told you and ask one question. Was any of it inconvenient?
A working process regularly produces answers the founder did not want. It changes a plan somebody liked. If nothing has ever come back uncomfortable, that is not confirmation that the idea is good. It is an instrument stuck at its default reading.
The fix is not more interviews. It is changing who is in the room, and changing what it costs them to be there.
Key takeaways
- Consistent feedback usually reflects a consistent sample, not a validated idea.
- Reproducible and correct are different properties. A process can return the same wrong answer indefinitely.
- A bigger sample does not remove bias that lives in who you are able to reach.
- Real demand signals cost the other person something before you arrived.
- If your research has never told you something inconvenient, it is probably not measuring anything.
FAQ
How many people should I talk to before building?
The number matters far less than who they are and what the conversation costs them. Five conversations with people who already built a workaround will tell you more than fifty with people who find the idea interesting. Pick the number after you have decided how you will separate a polite yes from a real one.
Is unanimous positive feedback ever a good sign?
Yes, when the people giving it have no relationship with you and something at stake in the answer. Unanimity among strangers who owe you nothing is a genuine signal. Unanimity among people who want you to succeed tells you about the relationship, not the market.
My friends and my network are my target customers. Does their feedback count?
Sometimes, and it is very hard to separate the relationship from the judgement. If they are genuinely in your market, use them to understand the problem and go elsewhere for the decision about demand. Those two jobs need different people.
What is the fastest way to get an honest answer?
Ask for something small that costs them. A deposit, a pre-order, a calendar slot they have to defend to somebody else, an introduction they have to spend reputation on. Anything free comes back positive from almost anyone.
Should I stop running customer interviews?
No. Interviews are very good at explaining behaviour and very bad at predicting it. Use them to understand why something happens, and use behaviour to decide whether anyone will pay.