In this article
Half the customers who answered your cancellation question selected “too expensive.” Should the next product meeting be about lowering the price?
Possibly. The answer could also describe a product someone barely used, a budget reduction at a happy customer, or a plan whose minimum commitment exceeds a small team’s needs. Each points toward different work. Treat the selected reason as the customer’s explanation at that moment, then establish what decision the available detail can support.
A useful cancellation review preserves the original response, adds relevant product context and gives unresolved questions a named next step. It should help the team choose between fixing a problem, investigating a pattern and accepting a departure that the product cannot reasonably prevent.
Keep the cancellation event separate from its explanation
Begin with a clear population. Are you reviewing completed voluntary cancellations, requests scheduled for the end of a billing period, or every subscription that ended? A failed renewal belongs in payment-recovery analysis unless the customer separately explains a decision to leave. Mixing these events changes what a reason distribution means.
Stripe’s subscription object has separate cancellation-detail fields for reason, feedback and comment. That separation is useful in your own records: the event’s system classification, the selected answer and any written explanation should remain distinguishable. Keep request and effective-end timestamps separately too. Stripe documents that its cancellation timestamp can reflect the request rather than the later end of service.
Retain the exact survey version with each answer. Save its wording, options and where it appeared in the cancellation journey. Pew Research Center’s questionnaire guidance explains that wording, the available response categories and their order can affect answers. A new “missing integration” option can change the apparent reason mix even if the product has not changed.
Read the denominator before the headline
Consider an illustrative monthly cohort of 20 completed voluntary cancellations. All 20 customers saw an optional question; 12 answered and eight skipped it. These are fictional records, not findings from a business.
| Selected reason | Answers | Share of the 12 respondents | Share of all 20 cancellations |
|---|---|---|---|
| Price | 6 | 50% | 30% |
| Missing capability | 3 | 25% | 15% |
| No longer needed | 2 | 16.7% | 10% |
| Other | 1 | 8.3% | 5% |
| No answer | 8 | Not included | 40% |
The defensible headline is that six customers selected price, representing half of respondents and 30% of the full cancellation cohort. You do not know what the eight nonrespondents would have selected. Assigning them to the observed categories would add information the customers never supplied.
Nor does a 60% answer rate, by itself, settle whether the responses are representative. Pew’s work on survey nonresponse distinguishes response rate from bias arising when respondents differ meaningfully from nonrespondents. Its telephone-survey results are not a benchmark for cancellation forms. The applicable question is whether the people who answered your form differ in ways relevant to leaving.
Compare available context such as plan, tenure and recent task completion between respondents and nonrespondents. A visible difference deserves investigation. Similarity on those fields still leaves unmeasured differences possible.
Code the explanation without rewriting the customer
Keep a raw-answer column and a separate analyst code. A short codebook should say what each code includes, what it excludes and when to leave a record unresolved. Permit more than one contributing factor in written feedback, but count a customer only once within each factor. Multi-code totals need not add to 100%.
These six selected records illustrate how to examine the fictional cohort. All comments and product observations below are invented for teaching.
| Record and stated reason | Relevant context | Working interpretation | Follow-up question |
|---|---|---|---|
| A: Price; “Our software budget was cut.” | Team completed its usual weekly work until departure | Budget constraint stated; value dissatisfaction unconfirmed | Which spending or scope changed? |
| B: Price; “I only need one report a month.” | One report completed in the last month | Low usage relative to commitment | Is the task occasional, or did something prevent use? |
| C: Price; “The bill was higher than expected.” | Invoice and plan terms have not been checked | Billing surprise, cause unresolved | What amount did they expect, and why? |
| D: Missing capability; “I couldn’t export.” | Their role lacked export permission | Access restriction may explain the failure | Did an admin approve the export or offer another route? |
| E: Missing capability; “It needs our warehouse connector.” | Required connector is absent | Specific product gap | Which recurring job depends on that connection? |
| F: No longer needed; “The project is finished.” | Final report completed before cancellation | Finite job completed | Is another project likely, or is this a permanent departure? |
The analyst should not silently change D’s answer to “user error.” The customer could not finish a task. The observed permission setting suggests a path to investigate, while the experience remains a real product problem. Likewise, A’s healthy activity does not disprove a pricing concern. Usage and willingness or ability to pay answer different questions.
Have another reader independently code a small shared set. Discuss disagreements, refine definitions and retain ambiguous cases. The point is a repeatable interpretation that someone else can inspect, with enough context to challenge it.
Route a pattern to the smallest useful decision
Price-related responses can lead to several investigations. A’s stated budget constraint may justify a conversation about narrower scope. B suggests examining whether packaging fits occasional work. C requires an invoice and expectation check before any pricing proposal. Sending all three a discount would test an offer while leaving their underlying situations unclear.
D belongs with the people responsible for permissions and task recovery. E deserves product discovery around the missing workflow, including customers who still use the product. F may call for a good exit and an easy return when the next project starts. Successful completion of a finite task is a different product situation from repeated failure at an ongoing one.
For each proposed change, write the observed pattern, affected segment, missing information, next investigation and decision owner. Ask what finding would make the team abandon the idea. That keeps a vivid comment from becoming a roadmap commitment without a clear reason.
Improve the question without turning departure into an interview
Stripe’s customer-portal cancellation page supports a selectable list of reasons and optional text for “Other reason.” It is a concrete collection mechanism, not a ready-made explanation of your product’s churn.
Keep your own question brief and voluntary. Preserve an honest route for an unlisted reason, and make follow-up contact optional. If “Other” contains a recurring, distinct problem, consider a new option; record the version change so the following month’s chart is not presented as a clean trend.
Separate two outputs at the review meeting. The first is the descriptive report: who canceled, who answered and what they said. The second is the decision log: what the team will investigate or change, and how it will judge the result. Use the discovery-to-opportunity process for that second step. A cancellation answer earns attention; the next piece of work determines what it should change.
Sources
Something we should clarify?
Suggest a correction


