Founders ask us how many customer conversations are enough. It is a reasonable question and the honest answer is unsatisfying: somewhere around ten, and after that you are usually buying comfort rather than evidence.
The number matters less than what comes back. Ten people describing the same problem in their own unprompted words is a signal you can act on this week. Forty people politely agreeing that your idea sounds useful is not a stronger version of that signal — it is a different thing entirely, and it is worth nothing.
What we look for instead
Three things, in order. Each one is harder to fake than the last.
- The same words. Not the same agreement — the same specific phrase, used by people who have never met.
- Something already paid for. A spreadsheet, an intern, a workaround. Money already leaving for this problem beats any stated intention.
- An action after the call. A forwarded email, a colleague pulled in, a date offered. Enthusiasm that survives the call ending.
The stall has a recognisable shape. Conversation twenty-two is scheduled, the deck gets another pass, and the week produces no new evidence. It feels like diligence from the inside. From the outside it looks like a founder who has decided not to find out.
Ten is a sample. Forty is a stall.
So we cap it. Ten conversations, then a written answer, then a decision about what to build. If the answer is thin after ten, that is itself the finding — and it is a cheaper finding at ten than at forty.1
1 The ten-conversation cap is not a research method, it is a decision deadline. We have broken it twice, both times in regulated markets where the buyer and the user were different people. ↩
Chapters
- 00:00Why we stopped the week
- 06:40The assumption we were testing
- 15:12What the week actually produced
- 24:35Telling the founder
- 33:08What we would do differently
Transcript
Lightly edited for readability. Timestamps jump the player.