What Cancellation Surveys Can Tell You About Churn

A customer who cancels and says nothing leaves you with a transaction record and a question. In its September 20 article, Thematic recommends collecting feedback from churned customers through short cancellation surveys, exit interviews and a review of support tickets, reviews and survey responses from before they left. Its argument is that customers who stayed can dominate feedback analysis, leaving the experiences of those who departed underexamined. Looking back through their earlier feedback expands what you can learn beyond the cancellation screen.

I agree with that expansion, but I would stop short of treating the resulting feedback as an explanation of churn across the business. A cancellation survey tells you why respondents say they left, not why everyone left.

The September 22 article “When Your Best Customers Teach You the Wrong Lesson” identifies that remaining visibility problem. Reaching beyond retained customers helps, but the people who answer are still a subset of the people who cancelled. Neither article establishes how large that gap is for your business. You have to investigate it locally, and some of it may remain unknown.

What should you investigate about recent cancellations?

I'd start with that question rather than “Why do customers churn?” The second question asks a small study to explain a whole business. The first can help you decide what to look at next.

For a subscription brand, you might define the group as customers who cancelled a paid subscription during the last complete calendar month. Decide whether that includes failed payments, first-order cancellations or people whose subscriptions remain active until the next billing date. A cancellation request and a subscription ending are different events. Pick the event that fits your question and use it consistently.

Write the definition down before reading the answers. I know how tempting it is to start with the comments. They're readable, and a vivid complaint feels like progress. But if I start there, I can end up building the study around whoever happened to give me the best story.

Your cancellation list is the starting point. The feedback attaches to that list, rather than quietly replacing it.

Who answered, and who is missing?

Alongside any themes, report how many customers were in the cancellation group, how many were invited to give feedback and how many answered. If you have partial responses or customers you couldn't contact, make those visible too.

This needn't become a response-rate exercise. The purpose is to show whose experience the findings describe. “Among cancellation-survey respondents” is a useful qualifier to keep attached to a chart when it moves into a presentation.

Then compare respondents with nonrespondents using information you already hold and are permitted to use. For a DTC subscription, that could include tenure, subscription frequency, product purchased, discount use or whether someone contacted support before leaving. Choose fields that could plausibly help explain different cancellation experiences.

For example, if customers with a support history are more common among respondents, your feedback may give more visibility to problems that reached support. That would justify looking separately at customers who left without contacting you. It would not tell you what those quieter customers experienced.

Even a close match on the fields you can check leaves room for differences you cannot see. Willingness to answer, frustration with the brand or having already moved on might affect participation. Those are possibilities to examine, not facts to assign to silent customers.

I wouldn't apply a correction factor based on a guess about those differences. You don't know how much is missing, or in which direction it would change the findings.

What the evidence actually says

Thematic's recommendation to review earlier feedback is useful because a cancellation answer can be thin. But combining evidence takes some restraint. I would keep three categories separate in the working notes:

• Stated reasons: what someone says in the cancellation survey or interview.
• Earlier complaints: what they reported before leaving, with the timing preserved.
• Inferred causes: your explanation of how an experience may have contributed to cancellation.

Imagine a customer selects “too expensive” when cancelling and previously contacted support about a late delivery. That's a hypothetical example, but the ambiguity is familiar. The delivery problem might have reduced the value they felt they were getting. Or the two events might be unrelated.

The records alone don't settle it. Record the price answer as a stated reason and the late-delivery ticket as an earlier complaint. Treat the connection as a hypothesis until you have more evidence, and perhaps even then.

I'd also resist calling the most common answer the “main cause of churn.” At most, it is the most common stated reason among the people who answered that question, under the choices and wording you offered them. If multiple answers were allowed, say so.

That sounds fussy until a team starts changing prices on the strength of the label.

Follow-up that tests your explanation

Use follow-up to explore a gap or an uncertain explanation. Interviewing only the people who wrote the longest comments gives you more detail from people who were already visible.

You could invite some nonrespondents from an underrepresented part of the cancellation group. You could also follow up with respondents whose stated reason and earlier feedback leave an unclear sequence. An invitation is only an attempt to reach them. The people who accept remain a selected group, and interviews don't remove that limit.

Keep the questions close to what happened:

• “Can you walk me through what was happening when you decided to cancel?”
• “What, if anything, had changed since your previous order?”
• “You mentioned price in the survey. What did that mean for you at the time?”

Let the person describe the decision before introducing an old support ticket. Otherwise, you risk supplying the connection you hoped to test. Afterward, you can ask whether that earlier experience played any part, leaving room for “no” or “I don't remember.” Our guide to writing survey questions that don't lead people covers that wording problem in more detail.

A survey may be enough to identify explanations worth investigating. An interview earns its place when you need to understand a sequence or clarify an ambiguous answer. More conversation doesn't automatically give you broader coverage.

A finding you can act on

I want a small churn study to end with a bounded decision. That might mean investigating delivery reliability among customers whose feedback points there, or checking whether subscription frequency fits how quickly customers use the product. Those are possible next steps, not conclusions you can draw without the corresponding evidence.

The report should let someone see the cancellation period, who answered and what remains uncertain without digging through an appendix. Keep reported reasons separate from your interpretation. Name the missing group you would most like to hear from next.

Before sending another cancellation survey, open the list of recent cancellations and mark which customers you already have feedback from. Choose the next follow-up from the unmarked rows, with a specific question you still need answered.