The 'What Made You Cancel?' Rooter-Cause Diagnosis
Ask what made them cancel (not why), then dig past the surface reason to the real causes.
- Difficulty
- Easy
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 93%
A method for extracting honest, actionable churn reasons from customers who are already mentally out the door. It combines a precise question wording, open-ended (not multiple-choice) responses, and a discipline of digging past the surface answer ('too expensive', 'project ended') to the deeper causes. Use it whenever you're trying to learn why customers leave.
Origin
Cohen learned the wording insight from a Groove case study and the digging discipline from a healthcare 'cause of death' analogy. He also discovered at Smart Bear that fixed-order cancellation dropdowns produce fake data — the first option was over-selected until they randomized order, after which all reasons were picked equally (pure noise).
Core principles
- 01Churning customers have stopped investing in you, so they won't spend effort diagnosing — make it easy and open-ended
- 02Fixed multiple-choice reasons produce noise (people pick the first option)
- 03Wording matters: 'what made you cancel' outperforms 'why did you cancel'
- 04The surface reason is the 'proximate cause'; keep asking to reach the 'rooter causes'
- 05Reject the idea of a single root cause — complex systems have many interlocking causes
How to run it
- 1
Use open-ended questions, not a dropdown
Ask a free-form question so you capture what the customer generates, rather than forcing them into pre-set buckets that produce noise.
Watch out Randomize any list you do show — a fixed order makes the first item over-selected and the data meaningless.
- 2
Ask 'what made you cancel?' not 'why did you cancel?'
The 'why' phrasing invites a simple abstract answer like 'budget'; 'what made you cancel' asks what about the product or situation caused it, yielding richer answers.
Pro tip Groove doubled usable responses (10% to 20%) on the same email just by switching 'why did you cancel' to 'what made you cancel'.
- 3
Dig past the proximate reason to the rooter causes
Treat the first answer as a proximate cause and keep asking. 'Too expensive' is almost never the real reason — they already saw your pricing page and chose to buy, so something else broke the promise.
Pro tip 'Project ended' may be real, but ask whether more successful software would have kept the project alive, or whether you targeted a fragile segment.
Watch out Don't collapse this into a naive 'five whys' that assumes one root cause at the bottom — look for an array of contributing causes.
- 4
Catch at-risk customers before they cancel
Talk to customers showing trouble signals (never uploaded data, too many or too few support calls, no logins) while they're still reachable — you can both save them and learn more.
Pro tip With enough data, correlate behaviors with cancellation, but note what churned customers have in common that retained ones don't.
- 5
Default to fixing onboarding
If you don't know what else to do, focus on onboarding: most cancellation happens in the first 30-90 days, and small onboarding shifts produce large downstream retention gains.
Pro tip Like a YouTube retention curve, shifting the early drop-off by 10% can lift end-retention by 20-30%.
Watch out Early churn is especially unprofitable — you paid to acquire them and they left before paying it back.
In the wild
Cohen noticed the first dropdown reason was picked most. After randomizing the order so each user saw a different order, all reasons were selected equally — revealing the original data was complete noise.
→ Proved fixed-order cancellation surveys generate fake signal; open-ended questions are safer.
A death's proximate cause is 'stopped breathing', but asking further reveals a car crash, caused by passing out at the wheel, caused by undiagnosed diabetes — plus systemic non-preventative healthcare.
→ Illustrates that 'undiagnosed diabetes' is a far more actionable cause than 'stopped breathing' — the same logic applies to 'too expensive'.
Common mistakes
Believing 'too expensive'
The customer already read your homepage and pricing page and chose to buy, so price was acceptable; something else broke the expected promise (e.g. a missing integration).
Abdicating responsibility on 'project ended'
You can't stop a project ending, but you can ask whether a more useful product would have kept it going or whether you targeted a fragile segment — closing that door ignores fixable issues.
Over-trusting AI to synthesize responses
LLMs are averaging machines — good at themes, bad at the surprising, actionable details; the details are the triggers for action, so read the responses yourself.
Is it for you?
Best for
SaaS operators diagnosing churn who want honest, specific reasons instead of misleading survey buckets
Not ideal for
Massive-volume consumer apps where individual outreach is impractical and only aggregate patterns are feasible
From the transcript
“What you want to do is say what made you cancel?”
“they started by asking why did you cancel? They got 10% usable responses. They changed it same email to why what made you cancel and…”
“too expensive is often the number one or at least like top three reason in one form or another. And that is never ever ever…”
From the episode
5 questions to ask when your product stops growing
Jason Cohen (2x unicorn founder)