Churn interview questions: learn why customers really leave

Stop reading the cancellation form as the answer. "Too expensive" is the easy box to tick when nobody has asked for the ten-minute story. It is a label. It tells you where to start asking, not what to fix.

This guide gives you a complete exit-interview script and a worksheet for weighing rival explanations. It works for subscriptions, memberships, service contracts and retainers, and for users who stop logging in.

The short version:

• Talk to three groups: recent leavers, near-leavers, and a few people who stayed.
• Ask for the story of the decision, event by event, not for opinions about your business.
• Write down two or three competing explanations before you start, and test each one against what you hear.
• Choose the one cause you can change, test cheaply, and check against records you already hold.

Treat the cancellation reason as a starting point

Do not treat "why did you cancel?" as a question with a straight answer. Nielsen Norman Group's guidance on interviewing users notes that human memory is fallible and that people¹. A one-line reason is that kind of story. The cancellation form invites it: one tick, no room for the Tuesday it all went wrong.

So use the form's reason to decide what to explore, then ask about events. "Too expensive" becomes a question about the day the price first bothered them and what they did next.

Who to call, and when

Recent leavers are your core group. The switch-interview method, designed for talking to people who changed what they do, recommends talking to people who switched². Book the call inside that window: call within a few weeks of the cancellation, well before the 90-day mark. Ask once by email or text, say plainly that nobody will try to win them back, and offer a small thank-you for 30 minutes of their time.

Near-leavers are the easiest group to forget, and among the most useful to add. These are people who started the cancellation flow and stopped, who paused, who downgraded, or who told you they were thinking about it. The same method says to recruit².

A few who stayed give you contrast. Choose people who look like your leavers on paper: same plan, similar tenure, similar size of business. If you only talk to leavers, you cannot tell a cause of leaving from a habit everyone shares. Purposeful-sampling guidance describes deliberately seeking examples that run counter to a trend, and calls³. Treat your stayers as that comparison group. Once a pattern shows, add the stayers who look most like your leavers and test the pattern on them. Add two or three so you do not tell a neat story.

Plan a first round of about ten to twelve conversations: five or six leavers, three near-leavers, two or three stayers. Include at least one person who left quietly. Stop when new calls stop changing your list of explanations.

Interview as a researcher, not a retention agent

Do not try to save the account on the call. It turns the interview into a negotiation, and you stop learning anything. Keep these two jobs apart: retention is someone else's call, on a different day.

During the call:

• Stay quiet longer than feels natural. The GOV.UK guidance on in-depth interviews is blunt:⁴.
• Never feed them your words. If they say "it just wasn't working", do not offer "was it the price?" Ask "what was not working?" The switch-interview method puts it plainly:².
• Do not defend. If they criticise your onboarding, say "tell me more about that" and write it down.
• Ask someone who was not on the account to run the call, if you can. That way the interviewer has nothing to defend and no reason to argue.

Follow your organisation's consent and privacy rules for recording and note-taking.

The exit-interview script

Work through five phases in order. You are reconstructing a timeline: start at the cancellation, go back to the first doubt, walk forward to the decision, then ask what they did next. Phases one to four ask about specific events and moments, because the point is to⁴ rather than generalities.

Copy this, replace the words in square brackets, and use it as a guide rather than a form. Skip what they have already answered and follow the story.

Opening (2 minutes)

• "Thanks for making time. I'm [name]. I'm trying to understand what happened, in your words. I'm not going to try to change your mind."
• "It will take about 30 minutes. Is it all right if I take notes? [Explain recording and storage in line with your own rules.]"

Phase 1: The moment they decided (4 minutes)

• "Think back to the day you cancelled. Where were you and what were you doing?"
• "What happened in the hour before you did it?"
• "What did you do straight afterwards?"

Phase 2: The trigger (8 minutes)

• "When was the very first time you thought about stopping, even for a second? What was going on?"
• "Between that first thought and cancelling, what happened? Walk me through it."
• "Was there a day it went from 'maybe' to 'yes'? What was different that day?"
• "What had you already tried before cancelling? Who did you talk to?"

Phase 3: The underlying problem (7 minutes)

• "Tell me about the last time you used [the service] before you stopped. What were you trying to get done?"
• "What happened? What did you do next?"
• "Describe a time it worked really well. When did that stop being true?"
• "In the weeks before you cancelled, what were you doing instead?"

Phase 4: What they switched to or did instead (4 minutes)

• "What are you doing now about [the thing it used to cover]? How did you pick that?"
• "What else did you look at? What made you drop it?"
• "If you are doing nothing: what happens now when that need comes up?"

Phase 5: What would have had to be true (3 minutes)

• "Think about the week before you cancelled. What would have had to be different for you to still be with us today?"
• "Was that about us, about you, or about something outside both?"

Treat this phase as a pointer, not evidence.

Closing (2 minutes)

• "What should I have asked that I didn't?"
• "Who else who left would be worth hearing from?"

Stayers and near-leavers

Keep the Opening and the Closing. Replace Phases 1 to 3 with the questions below (the last one is Phase 3's "last time you used it" question, reworded for someone who stayed), and drop Phases 4 and 5, which only make sense for someone who left. Allow about 12 minutes for the first three questions and 12 for the last-time question, so the call still fits in about 30 minutes.

• "Think about the last time you nearly stopped. What was going on?"
• "What did you do next? What made you stay that time?"
• "What would have to happen for you to stop?"
• "Tell me about the last time you used [the service]. What were you trying to get done? What happened, and what did you do next?"

Probes that work anywhere

"What happened next?" "Can you give me an example?" "Who said that?" "What did you do then?" For more on follow-ups that get past the first answer, see our guide to interview questions beyond surface answers.

Keep rival explanations alive

Every leaving story has a trigger and, behind it, a cause. The trigger is the event that tipped the decision: a price rise email, a missed month, a bad delivery. The cause is the condition that made the trigger matter: the service had stopped fitting their week, or a competing habit had already taken its place. Fix the cause, not just the trigger.

Phase 2 finds the trigger. Phase 3 finds the cause. The danger comes afterwards, when every story starts to support the explanation you liked before the interviews. Analysts who weigh alternative explanations are advised to⁵ rather than test one favourite. The rule that gives the method its force is about evidence:⁵. A leaver saying "the product was fine" fits every explanation, so it separates none of them.

Before your first call, write two or three rival explanations. Make them different enough that one fact could favour one over another. Then fill in this worksheet after every interview.

How to use it:

1. Write facts, not impressions. A fact such as "booked three evenings in a row and found them full" goes in. "Seemed frustrated with availability" does not.
2. Mark every row against every explanation. Most rows will be Neutral. Then fill the last column: Yes if the fact favours one explanation over the others, Partly if it splits some explanations from others but leaves two or more tied, No if it treats them all alike. Cross out every row marked No, because it tells you nothing.
3. Look for rows that fight an explanation. One solid fact against an idea is worth more than five that agree with it.
4. Keep an explanation Alive until a row marks it Fights. Mark it Weakened once one does, and Dropped only when two or more rows mark it Fights and none mark it Fits. Write the result in the Running view row.
5. After four or five interviews, write in the last column what you still need to hear. Add a question to the next call to get it.

A worked example: the pottery studio

Here is an invented example to show the method. Picture a neighbourhood pottery studio that sells monthly memberships. The cancellation form's top reason is "too expensive". The owner wants to know what to fix first.

The owner interviews eleven people: six leavers, three near-leavers and two members who stayed. Before the first call she writes down three rivals:

• A, Price: the fee is more than the studio is worth to them.
• B, Access: they cannot get a bench when they want one.
• C, Habit: one missed stretch breaks the routine and nothing brings them back.

She does not call the stayers yet, because they only help once a pattern shows. Her first seven calls are four leavers and three near-leavers. After five calls she checks the worksheet and writes in the last column what she still needs: do leavers with no missed month also find full evenings? She adds that question to the next calls. After seven calls, the worksheet looks like this.

Price is weakened, not dropped. The good-value row is Neutral for all three, because good value on the days they went says nothing about the days they stayed away, so it is crossed out. Access and Habit both survive. They also pull in different directions: Access says the studio should change its booking, Habit says it should change what happens after a missed month.

Her running view after seven: A weakened, B alive, C alive. She interviews two more leavers and asks the new question. Both say yes, and neither had missed a month.

Now she has a pattern to test, so she picks the two stayers: members on the same plan as the leavers, who also come in the evenings. Both book their evenings weeks ahead. She adds the row: Neutral for Price, Fits for Access, Neutral for Habit, and Yes in the last column.

That is enough to act on. She does not conclude that most members leave over bookings, because eleven conversations cannot tell her that. She checks the booking records she already keeps to see how often evenings sell out, and decides to test reserved recurring slots for the next group of new members.

Turn the stories into one decision

Reduce the pile of exit stories in four moves.

1. Sort by event, not by opinion. Write each interview as a short timeline of events, keeping their exact words where they matter: the first thought, the trigger, the decision, what happened next. The GOV.UK guidance on analysing research says to⁶. Then group those facts into themes.

2. Run the worksheet across all interviews. Count how many surviving explanations you have left. Two is a good answer. One is worth a second look: check you did not stop at the first explanation that fitted.

3. Choose the cause to act on with four questions.

4. Write the decision in one sentence, with the evidence beside it. For the pottery studio, that reads: "We will hold a recurring evening slot for new members for eight weeks, because four of six leavers hit full evenings and two stayers booked ahead; we will compare renewals against the previous group." The counts describe what she heard, not how common the problem is. For how to word the evidence behind a claim like that, see what makes a customer insight evidence-backed.

Then tell the people who gave their time what changed.

What this will not tell you

Interviews tell you which explanations survive close questioning. They do not tell you how many customers left for each reason. A round of interviews can surface causes you had not thought of; it cannot rank them by size. Use your records, a short survey built from the causes you found, or a test, to measure how common each one is.

Assume the leavers who agree to talk differ from those who do not, because you will never hear from the second group. Treat an explanation that survived as the best current account rather than proof, and expect to revise it after the next round.

If you want to see how this kind of study can be set up, you can see an example churn study setup.