A customer stopped using the product in March. The survey knows this, because it asked, and it knows they chose 'no longer needed' from a list of six reasons. The interview found out that the weekly summary email stopped arriving, and that without it they forgot the product existed, and that they would come back if it came back.
Nothing in the survey's six reasons was wrong. None of them was the reason.
Why surveys stop short
A survey has to decide its questions before it hears any answers. That is fine when you already know the shape of the problem and want to count it. It fails when the interesting answer is one you did not think to offer, which is most of the time with churn, with new markets, and with anything a customer finds hard to say.
Why interviews used to be rare
Interviews were expensive because a person had to run each one, and slow because that person had a calendar. So teams ran eight, called it qualitative, and went back to the survey for anything they needed a number from.
An interviewer that asks why, waits, and asks again, costs nothing to run at three in the morning in Portuguese. So the trade-off changes. You can run the interview with a hundred people, get the number from the same study, and still have every conversation to read.
What to do with a hundred interviews
- Read ten. Not the summary, the transcripts. The pattern you are looking for is usually visible by the sixth.
- Count the rest. The interviewer coded every answer against what it heard, not against a list you wrote in advance.
- Ask the respondents. Everything those hundred people said is now something your digital respondents know. Ask them the follow-up you wish you had asked.