How to Run a Customer Interview That Actually Tells You Something

Most customer interviews confirm what you already believe. Here is how to run ones that change your mind — including a script you can use today.

Customer interviews are one of the most powerful tools available to any team building a product or service — and one of the most consistently misused. Teams conduct them, write up notes, share a document, and then continue building exactly what they had planned to build before the conversation. The interviews felt productive. Nothing changed.

This happens because most interviews are structured to confirm rather than challenge. The founder or product manager already has a theory — the customer has this problem, our product solves it — and the questions, sometimes without the interviewer realising it, are designed to find evidence for that theory rather than test it. Customers, who are generally polite and want to be helpful, oblige.

Real customer discovery requires a different approach: a different frame of mind, a different set of questions, and a different relationship to what you hear. This article gives you the practical tools for each of those.

Why most interviews fail before they start

The most common mistake is confusing a customer interview with a usability test or a sales call. They are different activities with different goals. A usability test asks: can people use what we have built? A sales call asks: can we get this person to buy? A customer interview asks: what is actually true about this person's life, and where does a problem worth solving live? The goal is understanding, not validation, not conversion.

When founders pitch their product during a customer interview — even subtly, even by explaining what they are building before asking questions — they have already lost the most valuable signal available: the customer's unprimed, honest description of their world. Once you have described your solution, the customer begins trying to help you with it, rather than describing their experience as it actually is.

Who to talk to — and who to avoid

The sample you choose shapes everything you learn. Talking to the wrong people produces confident conclusions that are confidently wrong. Talk to people who have the problem you are investigating, not people who might have it someday. If you are building tools for marketing teams, talk to people currently working in marketing — not people who have vaguely considered marketing, or who have a friend who does marketing.

Avoid talking only to people who are already interested in your product. Early adopters and people who signed up for your waitlist are not representative of the broader market. They have already demonstrated they find the idea compelling, which means they will validate your assumptions even when those assumptions are wrong. Talk to them too — but talk to people who have never heard of you in roughly equal numbers.

Also avoid the warm network trap. Friends, family, and former colleagues will almost always tell you what you want to hear. Even when they try not to, the social dynamics of existing relationships make honesty difficult. Aim to get at least half of your interviews from people you do not know personally.

A structure that works

The script you can use today

What follows is a tested interview structure. Do not read it verbatim — adapt it to your context and let the conversation breathe. The questions in italics are the core ones; the notes are for you.

How to listen: what you should be hearing beneath the words

Follow the emotion, not the logic. When a customer says something with energy — frustration, resignation, unexpected delight — pull on that thread. "You mentioned that part is annoying — can you say more about what that's like?" Energy is a signal about what actually matters. When someone says something flatly, with no affect, they probably don't care about it as much as their words suggest.

Ask for specifics constantly. "Sometimes" is a word that hides almost everything. "Sometimes I forget to follow up." When? How often? What happens when you forget? What did you do last time that happened? General statements are safe for the customer to make; specific stories are where the truth lives.

Sit with silence. When a customer finishes an answer, do not rush to ask the next question. Let three seconds pass. In almost every case, they will add something — and the addition is usually more honest and specific than the initial answer. People fill silence, and what they fill it with tells you something about what is actually on their mind.

Every question in the "closes down" column is about your product or your hypothesis. The customer's job is not to evaluate your idea — it is to describe their world. Questions that invite them to evaluate produce politeness, not insight.

After the interview: from notes to insight

A customer interview that does not get processed within 24 hours is mostly wasted. Memory degrades fast, and the things you remember without notes are the things that confirmed your existing beliefs — the surprises and contradictions, which are the most valuable signals, tend to fade first.

Within an hour of each interview, write down three things: the most surprising thing the customer said, the most important thing they said, and any direct quotes that stuck with you. Do this before you talk to anyone about the interview, because discussion shapes and flattens the memory.

After several interviews, look for patterns across your notes. What problems come up repeatedly? What language do multiple customers use to describe the same thing? What workarounds have multiple people independently invented? Convergent patterns across interviews are the raw material of genuine insight. Single-interview observations are just anecdotes until they appear again.

How many interviews do you need?

Less than you think, done well, beats more than you can process, done quickly. For most product decisions, five to eight interviews with people who closely match your target customer will surface the most important patterns. The first few interviews reveal the most significant themes; each subsequent one adds marginal additional signal while repeating what you have already heard.

The exception is when you are segmenting: if you have multiple distinct customer types, you need five to eight per segment, not in total. A pattern that holds across small business operators but not enterprise buyers is a different pattern, and mixing the two groups will obscure that distinction.

The right question is not "have I done enough interviews" but "have I stopped hearing new things?" When three or four consecutive interviews produce nothing you have not already heard, you have probably reached saturation for that round. Do another round when your questions change — after you have built something, shipped something, or significantly adjusted your hypothesis.