Customer discovery interviews: capture what was said

Customer discovery has a failure mode nobody names: you hear confirmation. Signals that fit the hypothesis are remembered and repeated; signals that challenge it are rationalised or simply not retained. This is ordinary cognition rather than dishonesty, and it is strongest in exactly the people whose conviction is funding the company. You cannot fix it by trying to be objective. You fix it with a verbatim record you read later, when the social energy of the conversation is gone and you can see what you actually asked.

How to keep the research honest

01 / Capture verbatim, because the words are the asset

"I always get stuck where I have to connect my calendar" is a ticket. "The setup is a bit confusing" is noise. Those can be the same interview, ninety seconds apart, and only one of them is usable. A summary keeps the second.

02 / Read the transcript for your own questions

This is the uncomfortable part and the most valuable. Leading questions, subject changes when an answer got awkward, moving on before someone finished — all obvious in transcript and invisible in memory. It makes you a better interviewer faster than anything else.

03 / Count unprompted mentions

The number that matters is how many people raised a thing without being asked. That count is only available if you captured all of them the same way; from memory you will count the ones that stuck, which is the ones that agreed with you.

04 / Keep the language for the positioning

The phrase a customer uses about their problem is what should appear on the landing page. It cannot be manufactured and it cannot be recalled accurately — "something about being overwhelmed by email" is not a headline, and the sentence they actually said probably is.

05 / Make it available to people who were not there

The person writing the copy, building the feature or taking the sales call was usually not in the interview. A searchable archive is how customer understanding reaches them as evidence rather than as your summary of evidence.

Questions

How do you capture customer insights from interviews?

Record the interview with consent, then work from the verbatim transcript rather than notes. Extract the exact phrasing customers use, and count how many raised each theme unprompted.

How do startups avoid misreading customer research?

By separating the conversation from the analysis in time. Read the record days later, count unprompted mentions rather than trusting recall, and read your own questions for leading language.

What is the best way to find product-market fit?

Talk to customers constantly, capture what they say accurately, look for the same language recurring across different people, and update the hypothesis on what was said rather than on your summary of it. The capture step is the one most teams are weakest at.

Should I record customer interviews?

With consent, yes — and consent is straightforward here, because participants generally understand that accuracy is the reason. Tell them what you are using, what happens to the recording and how long you keep it.

How many customer interviews do I need?

Fewer than most people think if they are captured properly, and more than most people do if they are not. Ten well-recorded interviews you can actually search beat forty you half remember.

Scriben is recording people know about. You say what the pen is and what it does, they agree, and then it stays out of the way for the rest of the conversation — that second half is the product, and it only works after the first. Recording law varies by jurisdiction and by profession: see recording people lawfully before you start.