How do you run a customer interview about churn?

A churn interview is not a damage-control call—it's the highest-signal research a founder can do. Done right, it surfaces the gap between what you think your product does and what users actually experienced. The goal is to leave with a specific, falsifiable reason someone stopped paying, not a polite rationalization.

Why most churn interviews fail before they start

The typical mistake is treating a churn interview like a support ticket: the founder already has a hypothesis, asks leading questions to confirm it, and walks away feeling validated. This is what Garry Tan flags as the 'as expected' failure mode—when nothing in a user conversation surprises you, it means you filtered everything through assumptions you already held. Surprise is the signal. If you finish a churn call without hearing at least one thing that contradicts your mental model, you ran a bad interview.

The second pre-call failure is choosing the wrong medium. Surveys and email follow-ups systematically miss the truth. People click the polite answer when given a multiple-choice exit form. A synchronous conversation—phone or video, never async—creates enough social friction that users will eventually say the real thing if you stay quiet long enough. Voice hesitations, topic changes, and unprompted complaints are data that a survey checkbox can never capture.

Finally, many founders wait too long. The best window to reach a churned user is within 72 hours of cancellation, while the frustration or disappointment is still emotionally available. After two weeks, the user has mentally moved on and will give you a sanitized retrospective. Put an automated trigger on cancellation that queues a personal outreach from the founder within one business day.

Structuring the conversation: the arc of a churn interview

Open with context, not questions. Tell the user exactly why you're calling: 'I'm the founder. You recently canceled, and I'd rather understand what went wrong than guess. This call is just for me to learn—there's no sales pitch.' This framing disarms defensiveness and signals that honesty is safe.

Start with their timeline, not your product. Ask them to walk you back to the moment they first signed up: what problem were they trying to solve, what made them choose you, and what changed. You're reconstructing their job-to-be-done arc. The answer to 'what changed' is almost always more diagnostic than any direct question about your features. People churn because something in their life, workflow, or expectations shifted—and your product didn't shift with them, or never actually fit to begin with.

Once you have the timeline, go specific on the moment of decision. 'Can you remember the last time you logged in before you canceled? What were you trying to do?' This is the crux of the interview. Paul Graham's observation about direct user engagement—that watching someone use your product in their own environment is irreplaceable—applies here in reverse: you're reconstructing that scene from memory, and the details matter. Push past 'it just wasn't a fit' with follow-up questions: 'What were you hoping would happen when you did that?' and 'What actually happened instead?'

Close by asking where they went. What are they using now, or what are they doing instead? If they switched to a competitor, ask what that product does differently. If they're doing nothing, that tells you the problem felt lower priority than you assumed. Either answer is valuable and neither should be treated as an insult.

The questions that generate real signal

The most productive churn interview questions are backward-looking and concrete. Avoid 'what would have kept you?' because users are bad at predicting their own future behavior and tend to request features they wouldn't actually use. Instead, anchor everything in what actually happened.

High-signal questions include: 'What were you trying to accomplish the last few times you used the product?' 'Was there a specific moment when you decided it wasn't working?' 'Did you ever tell a colleague about us—what did you say?' That last question is especially useful: the language someone uses to describe your product to a peer is often more honest than anything they'll say directly to you.

Garry Tan's framework for spotting genuine insight applies here too—you're looking for the specific surprise, the moment where the user did something or felt something that you didn't design for or anticipate. If a churned user says 'I kept trying to use it for X but it didn't do that,' and X is something you never intended, that's not a one-off complaint. That's a signal about where your positioning is leaking, or potentially where a better product is hiding.

Avoid piling on multiple questions at once. Ask one, let there be silence, and resist filling it. The answer that comes after a three-second pause is usually more honest than the first one. Take notes but don't type audibly—it makes users feel like they're giving a deposition.

What to do with what you learn

After five to seven churn interviews, you'll start hearing patterns. Categorize the reasons not by feature request but by the stage in the user journey where trust broke down: acquisition (wrong user acquired, misaligned expectation), activation (user never got to a meaningful outcome), habit formation (product didn't create a reason to return), or value delivery (the core job-to-be-done wasn't reliably completed). The category tells you whether you have a marketing problem, an onboarding problem, a retention mechanic problem, or a product problem—and those require completely different responses.

Paul Graham's point about early users in 'Do Things That Don't Scale' is directly relevant here: the feedback you get from direct, engaged conversations with individual users is qualitatively different from anything you'll get at scale. Churned users are, paradoxically, some of the most honest people you can talk to—they have no reason to be nice to you anymore. Treat this as a gift rather than an embarrassment.

For each churn reason you identify, trace it back to a specific point in your onboarding or product flow and ask: what would have had to be true for this user to succeed? Sometimes the answer is that they were never the right user and you need to tighten acquisition targeting. Sometimes you'll find that one missing feature or broken flow accounts for 40% of churn, and fixing it is a clear priority. Either way, you're now operating on evidence rather than assumption—which is the actual job of a founder in the early stages.

“The feedback you get from engaging directly with your earliest users will be the best you ever get.”

— Paul Graham, source

The one thing to do

Book five churn interviews this week—call within 72 hours of cancellation, ask users to walk you through their last session, and don't stop questioning until you hear something that genuinely surprises you.

Frequently asked questions

How many churn interviews do you need before the data is useful?

Five to seven is usually enough to spot a dominant pattern. If you're hearing five different reasons from five different people, you either have a segmentation problem or you need to dig deeper—the surface reasons are rarely the real ones.

What if churned users won't respond to interview requests?

Response rates improve dramatically when the outreach comes from the founder personally, not a support alias, and when the message is short and explicitly says there's no sales pitch. Offering a $25 gift card as a thank-you also works and is money well spent compared to the cost of ongoing churn.

Should you try to win back the churned user during the call?

Not during the interview itself—it contaminates the data and makes users less honest. If at the end of the call you've genuinely learned something that you're already fixing, you can mention it briefly, but make the primary goal learning, not recovery.

How is a churn interview different from a standard user research interview?

A churn interview is forensic rather than generative—you're reconstructing a failure rather than exploring possibilities. This means the questions are more specific, the timeline matters more, and the emotional stakes for the user are higher. Treat it accordingly.

Sources

More playbook answers · Growth Prophet home