An NSF I-Corps instructor named John Main wrote something this week that deserves to be printed above every founder's desk: ten encouraging interviews can still give you the confidence to build something nobody wants to buy. The problem may be who you interviewed.
Most validation advice obsesses over the number. Do twenty interviews, do fifty, do a hundred. Almost nobody audits the list of names. And the list of names is where the confidence gets manufactured.
Why do encouraging interviews mislead founders?
Because people interview whoever is easiest to reach, then treat those conversations as evidence about a market. The easiest people to reach are, in order: friends, people who already like you, and people with time to talk. None of those traits correlate with buying.
A friend who knows your idea wants to encourage you. An industry expert may understand the market and still have no firsthand experience of the task you are fixing. A user can describe the pain accurately and know nothing about how the purchase gets approved. Each gives you real sentences. Stack enough real sentences and you get a spreadsheet that looks like proof.
Who should you actually be interviewing?
Three filters do most of the work. First, firsthand experience: the person has personally done the task or suffered the problem, recently, not managed someone who did it years ago. Second, no relationship to you: a stranger has no social reason to be kind, which is exactly what makes their reaction usable. Third, proximity to the money: they pay for solutions in this space, or they can walk you through who does and how.
A solo founder who went through three pivots and six launches before YC described the same shift this week. He stopped asking people whether his idea sounded useful and started asking for a recent example of the problem. What triggered it, how they handled it, what that cost them. If someone has never tried to solve the problem, that is not a smaller signal. It is a different answer, and usually the honest one.
Is a bigger sample the fix?
No. Ten more interviews with the wrong people is more wrong confidence, delivered with better statistics. A VP of product wrote this week about two founders she met at meetups. One had built a health records app from his own real pain and still could not say who pays, patients, hospitals or insurers. The other was piloting his tool for free and could not say whether anyone would pay or whether the need existed outside two businesses. Both had done conversations. Neither had talked to the person who signs.
The screening also has to happen without leading. If your recruiting message describes the solution, you have pre-selected people willing to nod at it. Recruit on the problem, never on the pitch.
What does a trustworthy interview set look like?
Small is fine. Five strangers with fresh scar tissue beat fifty acquaintances with opinions. The set should include the different roles in the decision, the person who suffers, the person who pays, the person who can veto. And it should include someone whose current approach works well, because the people with no workaround are not always the people with a budget.
The uncomfortable summary: interview evidence is only as good as the sampling behind it. Founders check their code, their funnel math, their burn. Almost nobody checks their sample.
Key takeaways
- Encouraging interviews from the wrong people produce confidence, not evidence. The sample matters more than the count.
- Filter for firsthand recent experience, no relationship to you, and proximity to the money.
- Ask for recent examples of the problem, not reactions to your idea.
- Recruit on the problem, never on the pitch, or you pre-select the polite.
Count your interviews if you like. The market will only count the ones with the right people in them.