ValidationJuly 9, 2026By Maarten Follaets

How many customer interviews to validate an idea?

A row of interview markers where new insight keeps rising until it flattens out around the twelfth interview

Quick answer: For most early ideas, around ten to twelve good customer interviews is enough to validate the problem, as long as you talk to people who actually have it. The research agrees: an analysis of Guest, Bunce, and Johnson's landmark saturation study found that most recurring themes had appeared within the first twelve interviews, and the basic ones within six. The real rule is not a number. You stop when new interviews stop teaching you anything new.

Every founder asks this eventually. I have done five interviews and everyone agrees. Is that enough, or do I need fifty?

The honest answer is that the number matters less than who you talk to and what you listen for. But people want a target, so let me give you one, and then the rule that actually decides it.

The number, and where it comes from

Aim for about ten to twelve solid conversations with people who genuinely have the problem.

That is not a made up figure. In one of the most cited studies in qualitative research, Guest, Bunce and Johnson interviewed sixty people and tracked when new themes stopped appearing. Most of the recurring themes had surfaced within the first twelve interviews, and the basic ones were clear within six. After that, extra interviews mostly confirmed what was already there.

Your validation interviews are not an academic study, but the shape holds. The big, repeated patterns show up early. The tail is diminishing returns.

Saturation is the real finish line

The number is a guide. The actual signal to stop is called saturation.

Saturation is the moment new interviews stop surprising you. The same problem, described in the same way. The same clumsy workaround. The same reason they have not fixed it yet. When you can predict what the next person will say before they say it, you have heard enough on that question.

Sometimes that happens at eight interviews. Sometimes you are still learning at fifteen because the problem is messier than you thought. Let the pattern tell you, not the counter. Counting to a magic number and stopping is how founders convince themselves they are done when they are not.

Who you talk to beats how many

Here is the trap in chasing a number: you can hit saturation on the wrong answer.

Ten interviews with people who actually live the problem are worth more than fifty with whoever answered. If your sample is your friends, your Twitter followers, or anyone being polite, you will get a clean, consistent, and completely misleading picture. Everyone agrees, and you are validated into a wall.

So before you count a single interview, define who has the problem. Then only count conversations with people who fit. If they do not have the pain, their opinion is not data, it is noise that happens to sound like agreement.

The other half is asking the right way. It is easy to lead people to the answer you want without noticing. I wrote a full set of customer discovery interview questions that keep you focused on what people do instead of what they would say about your idea.

Reading ten interviews for signal

Once you have your ten or twelve, the work is not tallying yes votes. It is separating real signal from politeness.

Sort what you heard into three buckets. Validated, where the evidence clearly supports the assumption. Invalidated, where it clearly does not. And open, where you genuinely cannot tell yet. Most founders skip the invalidated and open buckets and only collect the quotes that flatter the idea. Do not do that. The invalidated notes are the most valuable thing you will get.

If pulling signal out of a pile of conversations feels like guesswork, that is normal, and it is where I lean on tooling. I broke down the method in how to analyze customer interviews.

Interviews are the first gate, not the whole thing

One caution. Interviews validate the problem, not the business.

Ten good conversations can tell you the pain is real and widely felt. They cannot tell you people will pay. Words are cheap in an interview, and everyone is generous when it costs them nothing. So treat a strong interview round as permission to run the next test, not as proof the idea is validated. Pair it with a real ask, a pre order or a paid pilot, before you bet on it.

Where Ventropolis fits

I built Ventropolis because doing this well, on your own, takes a kind of objectivity that is hard to hold when it is your idea.

Foxy, our AI co-founder, helps you decide who to talk to, plan questions that do not lead the witness, and read the transcripts for signal instead of for reassurance. It will tell you when you have hit saturation, and it will point out the answers you would rather not have heard.

You still make the call. Foxy just makes sure you are counting the right interviews for the right reasons.

If you want help running an interview round that actually settles the question, see how Ventropolis works for founders.

So the next time you wonder whether five interviews is enough, ask yourself a better question: has anything the last three people told you actually surprised you?

Frequently asked questions

How many customer interviews do I need to validate an idea?

Aim for around ten to twelve good conversations with people who genuinely have the problem. Qualitative research backs this up: in a well known study, most recurring themes had surfaced within the first twelve interviews, and the basic ones within six. You keep going only until new interviews stop teaching you anything new. If interview eleven and twelve tell you nothing you have not already heard, you are done.

What does data saturation mean?

Saturation is the point where new interviews stop producing new insight. The same problems, workarounds, and objections keep coming up, and each additional conversation just confirms the pattern. It is the honest signal that you have heard enough on this question. It does not mean you know everything, only that more of the same interview will not change your answer.

Do the interviews have to be with the right people?

Completely. Ten conversations with people who actually live the problem are worth more than fifty with a random crowd. If you interview the wrong segment, you can hit saturation on the wrong answer and feel validated while being wrong. Define who has the problem before you start, and only count interviews with people who fit.

What if all my interviews say different things?

That usually means one of two things. Either you are talking to too broad a group and should narrow to one segment, or the problem is not as sharp or as widely felt as you hoped. Scattered answers are a finding, not a failure. It is often a sign to tighten who you are building for before you count any interview as validation.

Is ten interviews really enough to bet on?

Interviews validate the problem, not the whole business. Ten to twelve good ones tell you whether the pain is real and shared. They do not tell you people will pay. Pair the interviews with a willingness to pay test, a pre order or a paid pilot, before you treat the idea as validated. Interviews are the first gate, not the finish line.