Three vague replies from thirty emails is the normal outcome — and a design problem. Here is the version that gets usable answers.
You emailed thirty churned customers and three replied with something vague. That is the normal outcome, and it is a design problem rather than a sign that churned customers will not talk to you.
Expect 15–30% on a one-question personal email, against 5–10% on a survey link.
The reason this matters: churn diagnosis needs to know which type of churn you have, and five types need opposite fixes. Without responses you are guessing — the churn framework covers the five.
Four reasons, all fixable.
It arrives too late. Six weeks after cancelling, people have genuinely forgotten. The answer you get is a plausible reconstruction, which is why it reads as vague — because it is.
It asks why they left. That question invites a socially easy answer. "Too expensive" and "not the right fit" are the polite exits, and both are usually wrong. Price is blamed for problems that are really about value never being reached.
It looks like marketing. A branded template with a survey link reads as automated, and people do not owe automation a reply.
It asks too much. Every question after the first cuts your response rate. A five-question survey to someone who already left is an unreasonable ask.
Plain text, from your own address, no logo, no template. Subject line should look like a message from a person: their first name, or four lowercase words.
The body is three sentences. Acknowledge they cancelled without apologising. Ask one specific question about behaviour. Say explicitly that you are not trying to win them back.
That last line does more work than anything else. The main reason people ignore these emails is the assumption that replying triggers a retention sequence. Removing that fear is most of the battle.
| Instead of | Ask |
|---|---|
| Why did you cancel? | What were you doing the last week you used it? |
| What could we have done better? | What are you using instead now? |
| How would you rate your experience? | When did you last get something useful out of it? |
| Would you consider coming back? | What happened the week you stopped logging in? |
The right-hand column asks about events. The left asks for opinions and self-diagnosis, which people are bad at and generous with — the same principle as the Mom Test, applied at the other end of the lifecycle.
Send within 72 hours of cancellation, while the specifics are still available to them. Set it as a manual trigger rather than an automation, because the value is in it being visibly personal.
For customers who already churned months ago, expect lower quality answers and treat the exercise as directional. Prioritise the most recent twenty over a larger list of older ones.
Ten useful responses is enough to categorise. You are looking for a pattern in when people stopped, not a representative sample — churn benchmarks at solo scale covers why rates matter less than shape at small numbers.
Offer nothing in exchange. No discount, no gift card, no reactivation offer. Incentives attract people who want the incentive and change what they tell you, and they reintroduce exactly the suspicion that stops people replying.
You are sorting responses into buckets, not averaging them. Two data points do most of the work: when they stopped using it relative to signup, and whether they were active before cancelling.
| What they say | Actual churn type |
|---|---|
| "I never really got it set up" | Onboarding — fix activation |
| "We finished the project it was for" | Value — the job ended, structural |
| "It was fine, just cutting costs" | Price, but check they were active |
| "I did not realise it was still charging" | Never activated — onboarding |
| "I switched to [competitor]" | Positioning — you were substitutable |
Note the third row. "Cutting costs" is the most over-reported reason and the least reliable. If they were not actively using it, the cost was simply the visible part of a value problem.
Count the buckets and act on the largest one only. Splitting effort across three churn types is how founders spend a quarter and move nothing.
If onboarding dominates, which it usually does early, the work is upstream of churn entirely — the onboarding framework covers closing the gap to first value. For the fixes for each type once identified, the churn reduction guide covers them in order.
And if the answers reveal that your churned customers were a different type of person from your retained ones, the problem is upstream of both — positioning or product covers that diagnosis.
Usually four reasons: it arrived weeks after they left, it asked why they cancelled rather than what they were doing, it looked like automated marketing, and it asked more than one question. A one-question personal email typically gets 15–30% against 5–10% for a survey link.
Ask about behaviour rather than reasons: what were you doing the last week you used it, what are you using instead now, when did you last get something useful out of it. People reconstruct their reasons unreliably but remember what they actually did.
Within 72 hours of cancellation, while the specifics are still available to them. For people who churned months ago, expect lower quality answers and prioritise the most recent twenty over a longer list of older ones.
No. Incentives attract people who want the incentive, change what they tell you, and reintroduce the suspicion that replying will trigger a retention sequence. Saying explicitly that you are not trying to win them back does more for response rates.
Ten useful responses is enough to categorise. You are sorting answers into buckets rather than seeking a representative sample, and the pattern in when people stopped matters more than the raw churn rate at small scale.
Often not. Price is the most over-reported and least reliable answer because it is socially easy. If the customer was not actively using the product before cancelling, cost was the visible part of a value problem rather than the cause.
Bring what your churned customers told you. Marcus sorts it into the five types and names the one worth fixing.
Try GhostCoach free →14-day free trial · cancel anytime · 30-day money-back on Lifetime