Ask ten founders how many customer interviews you need and you will get ten confident numbers. Five, because that is what fits into a week of building. A hundred, because someone impressive said so on a podcast. This week one founder story did the rounds about a team that ran twenty customer calls a week for over a year, more than two thousand conversations, on their way to seven figures in revenue. The same week, a first-time founder shared that her very first discovery interview became her first paying customer, sixteen weeks into the program that finally made her do one. Both stories are real. Neither one hands you your number.
Why is there no magic number?
An interview count is not a quota to hit, it is a measurement instrument, and instruments only work when they are pointed at the right thing. How many conversations you need depends on three variables: how sharp the question is that you are trying to answer, how well you chose the people, and how honest the setting allows them to be. Twenty interviews with people who live inside the problem will teach you more than two hundred with people who merely find your idea interesting.
This is why the famous numbers vary so wildly. The team doing twenty calls a week was not validating one assumption, it was running a permanent research operation across discovery, usability and retention. The founder who converted her first interview had spent weeks narrowing exactly who to talk to before she ever booked it. Same activity, completely different jobs.
Who actually counts as a customer interview?
A conversation counts when the other person has the problem, has already tried to do something about it, and spends the time describing what they actually did rather than what they might do someday. That last part is the whole game. What people say they would do is a compliment. What they did last Tuesday is data.
By that standard, a lot of things founders count are not interviews. An industry association reviewing your concept is not an interview. A friend telling you the idea is clever is not an interview. A survey is not an interview, because a form cannot ask the follow-up question that separates a habit from a hypothetical. If you finished a conversation without learning what the person did the last time the problem showed up, and what that cost them, it was a chat.
What does enough actually look like?
Researchers call it saturation, and you will recognize it without the vocabulary. In your first conversations everything is surprising. Somewhere between fifteen and thirty conversations with well-chosen people, the surprises thin out. You start hearing the same three complaints, the same duct-tape workaround, the same person who controls the budget. When two conversations in a row teach you nothing you did not already know, the next interview is worth less than the next experiment.
So track the only metric that matters here: surprises per conversation. While it is high, keep going. When it hits zero and stays there, stop scheduling calls and go test whether the pattern you found survives contact with real behavior, real strangers and real money.
What if everyone keeps saying nice things?
Then you have measured kindness, not demand, and more interviews of the same kind will only measure it more precisely. Saying I would definitely use this costs nothing, which is exactly why people say it so freely. The fix is not volume, it is temperature. Talk to strangers instead of your network. Ask about the past instead of the future. And watch for the endings that cost something: an unprompted introduction, a request to be told when it ships, a follow-up message you never asked for. Interviews that all end warmly and never end in any of those have answered your question, just not the way you hoped.
Key takeaways
- There is no universal number. Saturation, not volume, is the stop signal.
- Fifteen to thirty conversations with the right people is where patterns usually stabilize.
- Count surprises per conversation. When they stop, stop.
- Only people who have actually tried to solve the problem count as interviews.
- Warm answers from your own network are the most expensive kind of noise.
FAQ
Is 5 customer interviews enough to validate an idea?
Rarely. Five well-chosen conversations can kill an idea fast, for example when nobody you talk to has actually tried to solve the problem. But five is almost never enough to confirm one, because early patterns are noisy and politeness inflates everything.
Are surveys a substitute for customer interviews?
No. Surveys collect stated opinions at scale, and stated opinions are the least reliable signal in validation. Interviews earn their value from follow-up questions about what the person actually did, which a form cannot ask.
Should I interview people I already know?
Only as practice. People who know you are paying you kindness, not attention. The interviews that count come from strangers who have the problem and owe you nothing.
How long should a customer interview be?
Twenty to thirty minutes is usually enough. Past that point most people drift from what happened into what they imagine, and imagination is exactly the input you are trying to avoid.
Do AI-simulated interviews count toward the number?
They can sharpen your questions before you spend real conversations, which is useful. They cannot replace them, because the signal you are after is what a real stranger does with their limited time, attention and money.