How to find out why your users churn
A practical method to understand churn: start from the data, talk to users who already left, find the moment the decision forms, then size each cause with a survey.
- Do not start with a survey. Start with the data: when do users leave, and what changes before they do?
- Interview users who left recently and rebuild the timeline of their decision, not their opinion of the product.
- Look for the moment the decision forms. It is often weeks before the cancellation, when you can still act.
- Size every cause with a survey and keep only those confirmed by two sources.
1. Start from the data, not from a questionnaire
Before talking to anyone, spend a few days on cohorts, usage curves and support tickets. You want to know who leaves, when, and what their behaviour looks like in the weeks before.
This turns interviews into tests of hypotheses instead of open-ended chats. At Betclic, one week on cohorts and support tickets showed where churn concentrated before the first interview.
2. Talk to users who already left
Active users explain what they like. Users who left explain why they left. Recruit people who churned recently, ideally within the last 90 days, while the decision is still fresh.
Ask them to tell the story: the last time they used the product, what changed, what they tried instead. You are rebuilding a timeline, not collecting a satisfaction score.
3. Find the moment the decision forms
Most teams react at cancellation. By then the decision is already made. The useful question is when it started.
In the Betclic churn study, play frequency dropped by 23% three weeks before cancellation, 50% at two weeks and 77% in the last week. The decision formed about three weeks earlier, which is the window to act.
4. Size each cause with a survey
Interviews tell you which causes exist. A survey tells you how much each one weighs. Send a short questionnaire to a larger group of users whose activity is dropping, built from the causes the interviews surfaced.
Then cross the two: a cause that appears in interviews and is confirmed by data or survey is solid. A cause seen in only one source stays a hypothesis.
5. Turn it into a decision
End with three things: the causes ranked by weight, the moment to intervene, and what to fix first. A long report nobody reads changes nothing; a short priority list changes the roadmap.
Questions teams ask.
- How many churned users should I interview?
- In my experience, 8 to 12 interviews per user segment are usually enough to see the same causes come back. Add a survey when you need to know how much each cause weighs.
- Is an exit survey enough to understand churn?
- Rarely. Exit surveys capture the last reason people give, not the story that led there. Use them to size causes, after interviews have told you which causes to ask about.
- How long does a churn study take?
- Usually 2 to 6 weeks: about a week of data, two weeks of interviews, a week of survey and a few days to turn findings into priorities.