The 6 Questions to Ask a Customer Who Just Churned
Not ten questions. Six, in this order, each with the follow-up that unlocks it — because the order is doing more work than the questions.
The single most common mistake on a churn call is opening with “so, why did you cancel?”
It feels like the obvious question. It is the wrong one, and asking it first ruins the rest of the call. You get the rehearsed answer — the tidy one they already gave their boss, their finance team, and themselves — and everything after that is them defending a position rather than remembering what happened.
Six questions. This order. Each one earns the right to ask the next.
The whole design principle: ask about events, not reasons. Reasons are stories people construct afterwards. Events are things that happened, and things that happened can be fixed.
TL;DR
- Never open with "why did you cancel?" — it triggers the rehearsed answer and the rest of the call defends it.
- Q1 establishes the timeline: when did you first think about cancelling? The gap to the actual cancel date is your intervention window.
- Q2 gets the concrete last straw. If you hear a category ("too expensive"), you have not finished this question.
- Q5 is the roadmap question, phrased as a counterfactual: what would have had to be true for you to still be a customer?
- Q6 is where the best sentence of the call usually lands. Never skip it for time — skip Q4 instead.
Before You Dial: Four Things, Ten Minutes
Pull the account
Signup date, plan, MRR, last login, last three support tickets. You are not going to recite these — you are going to notice when their story and the data disagree, which is one of the most useful moments on the call.
Write down what you already suspect
One sentence. You are testing this hypothesis, not confirming it. Writing it down first is what stops you unconsciously steering toward it.
Check the date
Call within 7 days of the cancellation. After about 14 days, what you get back is a summary rather than a story — the specific frustration has been replaced by a tidy narrative, and the narrative is not what happened.
Pick the wrong person to call
Never let whoever owned the relationship run the call. People soften bad news for someone they like. A colleague they have never met gets a straighter answer.
The Opening Line That Gets the Honest Version
“Hi [name], this is [you] from [product]. I'm not calling to win you back and there's no offer at the end of this — I promise. You cancelled last week and I'd rather learn something from it than guess. Ten minutes, and I'll take whatever you give me. Is now bad?”
Two things are doing the work. Naming the absence of a save attempt removes the defensive posture they arrived with — most people expect a retention pitch and are braced for it. “Is now bad?” gives them a costless exit, which paradoxically raises the number who stay on the line.
If they say now is bad, book a time. Do not push through. A call someone is enduring produces short answers.
Q1 — “Walk me back. When did you first think about cancelling?”
Why this first: churn is a decision with a date, and that date is almost never the cancellation date. In our calls the gap runs from a few weeks to several months. Whatever it is for you, that gap is your intervention window — the period where a different email, a different in-app nudge, or a call from a human would still have mattered.
Follow-up: “What was happening that week?”
This second question is the one that produces the story. People remember context better than they remember decisions, so anchoring to a week retrieves detail that anchoring to a reason never does.
Asking “why did you cancel” here. You get the rehearsed answer and spend the next nine minutes inside it.
Q2 — “What was the last thing that happened before you decided?”
Why: reasons are abstractions; events are specific. Compare:
| What you get from a reason | What you get from an event | |
|---|---|---|
| Speed | "It was too slow" | "I waited 40 seconds for the export while a client was on the call" |
| Price | "Too expensive" | "Renewal landed the same month we lost the retainer" |
| Features | "Missing features" | "I needed to send it to our accountant and there was no export" |
| Complexity | "Too complicated" | "I set it up twice and both times the import dropped our custom fields" |
Every left-hand answer is unactionable. Every right-hand answer is a ticket.
Follow-up: “Was that the first time, or had it been building?” — this separates a single incident from a pattern, and patterns are worth more.
Accepting the abstraction. If you hear a category, you have not finished this question. Stay on it: “What did that look like, specifically? What were you doing when you noticed?”
Q3 — “What are you using instead?”
Why: the answer is rarely a competitor, and each possible answer means something completely different.
| Answer | What it actually means | Follow-up |
|---|---|---|
| A competitor | Feature or positioning gap | "What does it do that we didn’t?" |
| A spreadsheet | Your product cost more effort than it saved | "What made that good enough?" |
| An intern / manual process | Value was real but not worth the price | "How long does that take each week?" |
| Nothing | The need went away — often not your fault at all | "What changed so you didn’t need this?" |
“Nothing” is the most under-appreciated answer here. If a meaningful share of your churn is people who simply stopped needing the job done, you do not have a product problem — you have an acquisition targeting problem, and the fix is upstream.
Getting defensive about the competitor. You are collecting, not selling. The moment you argue, the call becomes a negotiation and the honest answers stop.
Q4 — “When you signed up, what did you expect this to do for you?”
Why: most churn is the distance between the promise they heard and the thing they got. This question measures that distance directly, and the answers usually indict marketing or onboarding rather than the product.
Follow-up: “Did that ever happen? When?”
If the answer is “no, never”, you have an activation problem. If it is “yes, at first”, you have a sustained-value problem. Those two get fixed by entirely different teams.
Correcting them. If they expected something you never claimed, that is still your problem — it is a positioning bug, and it is producing customers who were always going to leave.
Rob Fitzpatrick's The Mom Test is the best short book on this and its core rule is the one underneath every question above: talk about their life, not your idea. The failure it names — asking questions so leading that even your mother's answer would be useless — is exactly what “would you have stayed if we'd built X?” does to a churn call.
Q5 — “What would have had to be true for you to still be a customer today?”
Why: this is the roadmap question, and the phrasing is load-bearing. “What should we build?” asks a customer to do product management with no knowledge of your codebase or strategy. A counterfactual asks for a condition, which is exactly what you can prioritise against.
Follow-up: “And if we'd shipped that six months ago, would you have stayed — or was something else already in motion?”
That follow-up is a lie detector, and it is kind rather than confrontational. It gives them permission to say “honestly, probably not”, which is enormously valuable information. Some churn is unwinnable, and knowing which is which stops you building for customers you were always going to lose.
Treat the answer as a condition, not a commitment. They are not promising to come back. They are telling you what mattered — which is what you asked for.
Q6 — “Anything else you'd want the person who builds this to hear?”
Why: the single most useful sentence of the call lands here more often than anywhere else. The framing matters — “the person who builds this” makes it concrete and slightly personal, and gives them permission to say the thing that didn't fit the previous five questions.
Follow-up: silence. Count to five. Do not fill it.
This is the hardest instruction in the whole script and the highest-return one. The second half of an answer is where the truth usually is, and most interviewers talk over it because four seconds of silence feels like forty.
Running out of time and skipping it. If you are at ten minutes and have two questions left, drop Q4. Never drop Q6.
The close
“That's genuinely useful, thank you. If we fix [the specific thing they said], can I email you and tell you? No pitch — just so you know it wasn't wasted.”
Roughly one in five say yes. Those become your win-back list, and they convert considerably better than a discount campaign, because you return with evidence rather than desperation.
What To Do In the Ten Minutes After
Log it the same day. Memory of a call decays as fast as memory of a frustration, and what replaces it is your own interpretation.
6
Questions, in this order
12 min
Typical call length
8–15
Calls before answers stop surprising you
1
Verbatim you must record exactly
The verbatim column is the one people skip and the one that matters most. Aggregated themes persuade nobody in a roadmap meeting. One exact sentence from a real, named customer ends the argument — which is why the coding scheme in the feedback-to-decision framework keeps verbatims attached to themes permanently.
Stop when new calls stop surprising you. That is usually somewhere between eight and fifteen for a single segment — consistent with the diminishing-returns curve Nielsen Norman Group documented for qualitative research. You do not need thirty.
The 6-Question Churn Call Script
The full script: pre-call checklist, the opening line verbatim, all six questions with follow-ups and failure modes, the close, and the logging sheet.
- Pre-call checklist and the "call someone they don’t know" rule
- The opening line, word for word
- All six questions with their follow-up and their failure mode
- The close that builds your win-back list
- A nine-field logging sheet, including the verbatim column
“Talk about their life instead of your idea. Ask about specifics in the past instead of generics or opinions about the future.”
Frequently asked questions
- How many churned customers should we interview?
Eight to fifteen per segment, then stop. Nielsen Norman Group's research on qualitative studies found five participants surface the large majority of issues, with sharp diminishing returns after that; churn interviews behave similarly. If call twelve is still surprising you, your segment is too broad — split it and run each half separately.
- Should we offer an incentive for a churn interview?
A small thank-you after the fact is fine. An incentive offered up front changes who agrees — you recruit people motivated by the incentive rather than people motivated to tell you something — and it makes the “no pitch” promise less credible. If you must, mention it at the end.
- What if the customer is angry?
Let them be. An angry churned customer is giving you the highest-signal call you will get that month, and the only wrong move is defending. Say “that sounds genuinely frustrating, tell me what happened” and take notes. Do not explain, do not contextualise, do not offer a credit.
- Can we send these six questions as a survey instead?
You can, and they will work better than a dropdown, but you lose the follow-ups — and the follow-ups are where most of the value is. Q2 in particular almost always needs one more push to get past the abstraction, and a form cannot push. If a survey is your only option, ask Q2, Q3 and Q5 only, and leave every box open.
Sources & further reading
- 1The Mom Test — Rob FitzpatrickThe reference text on asking customers questions that survive their politeness.
- 2Why 5 Participants Are Okay in a Qualitative Study, but Not in a Quantitative One — Nielsen Norman GroupWhy a small number of conversations is sufficient for discovery, and insufficient for sizing.
- 3Making Exit Interviews Count — Harvard Business ReviewOn interview structure and why organisations fail to act on what they hear.
- 4How To Talk To Users — Startup School — Y Combinator
Keep reading
These six questions work. Someone still has to make the calls.
saasfeedback.ai runs this exact script with your churned users, trial drop-offs and downgrades — recordings, transcripts and MRR-weighted themes back within 48 hours.
Book a demo