The Churn Survey
Ask why at the exit, briefly, without using the question to block the door.
The problem it solves
Cancelling users are the only people who can tell you precisely what failed. They are also the least willing to spend time on you, so the survey has to be small and the cancellation has to complete either way.
A churn: The rate at which customers stop paying or stop using the product. survey is one screen in the cancellation flow. A short list of reasons, one optional free text field, and a cancel that goes through whether or not anything is answered. That last constraint is the whole ethics of the pattern.
The multiple choice list produces the counting. The free text produces the understanding. Most teams read the chart and skip the text, which is backwards: the chart tells you that 22 percent said too expensive, and the text tells you they meant it stopped being worth what it costs, which is a different problem with a different fix.
Writing the options
- Six options at most, plus other, in a randomised order below the first two.
- Include an option that blames the product directly, phrased plainly. If your list has no honest way to say it was bad, your data is decoration.
- Separate no longer need it from it did not work, because they lead to opposite decisions.
- Keep price and value distinct.
- Make every field optional.
Then route the answers. A cancellation citing a bug should reach the team that owns it with the account attached. A survey whose output is a quarterly slide is a survey that changes nothing, and the users who filled it in were right to suspect that.
Resist the intercept. Anything that uses the survey answer to launch a retention offer converts the question into a sales step, and people who have worked that out stop answering honestly or pick whichever option looks likely to end the conversation fastest.
Worked example
One question that was worth it
Imagine a tool where the free text on cancellations keeps mentioning an export that silently truncates. Nobody filed a bug, because people who are leaving do not file bugs. The multiple choice would have logged those as missing features. The text field is the only place that problem was ever going to appear.
When it fits
- Cancellation is self-serve, so there is a flow to place the question in.
- You have enough monthly cancellations for patterns to be readable.
- Someone owns the responses and can route individual ones to the right team.
- The survey can be skipped entirely without affecting the cancellation.
When it backfires
- The survey is required to cancel. Forcing someone to answer questions to leave is a dark pattern: An interface built to get an outcome the user would reject if it were stated plainly., it is regulated in several places, and it poisons every answer you collect.
- Answers trigger an offer or a retention pitch. Once the question is a sales trigger, the responses stop being research.
- The option list has no honest way to say the product was bad, so the data reassures you and teaches you nothing.
- Nobody reads the free text, which means the only channel capable of surprising you is the one being thrown away.
Products using it
Netflix
Places a short optional reason list in the cancellation flow and completes the cancellation in the same session regardless of the answer.
Basecamp
Keeps account cancellation self-serve and immediate, with an optional feedback field rather than a retention gate.
Mailchimp
Asks a short reason question during downgrade and cancellation without requiring a support conversation to proceed.
The psychology under it
The takeaway
Keep the exit survey optional and one screen, read the free text rather than the chart, and never wire an answer to a retention offer.
Finished the teardown? Bank it and the day counts toward your run.