Cancellation Survey

Run a subscription product and the count of people who left accumulates by itself. How many teams cancelled this month, and whether that share rose or fell, comes straight out of the billing records. What those numbers can't tell you is what to fix. A move from three percent to five percent says something is wrong; it doesn't separate a pricing problem from a missing feature from people who gave up because they couldn't figure the product out.

So teams guessed. Put the churn rate on a screen once a month and everyone describes what they can see from where they sit. Sales says competitors are cheaper. Engineering says the product is slow. Design says the screens are confusing. All three are plausible and none of them come with evidence. The meeting ends with whoever argued hardest, a discount goes live next quarter, and churn doesn't move.

The person who actually knows the reason is the one leaving. And their last contact with the company is the screen where they press the cancel button. A cancellation survey grabs that moment to ask one thing. It isn't there to turn that person around — it's there to get one line out of them that becomes the basis for keeping the next person.

A survey is made of five parts: where you ask, the list of reasons, a free-text box, a follow-up question, and the account details attached to each answer.

1. Where you ask. Put it on the screen immediately before cancellation is confirmed. Email it afterward and response rates fall into the single digits, because someone who has already left has no reason to answer.
2. The list of reasons. Five to seven options works. Too expensive, missing a feature I need, couldn't figure out how to use it, switching to another product, the project ended and we don't need it — those are the usual starting set. What matters is that each option leads somewhere different. Rising "too expensive" answers send you to pricing; rising "couldn't figure it out" answers send you to what happens right after signup.
3. A free-text box under the options. Most answers leave it blank, and the handful that don't tell you what the option list failed to capture. When "switching to another product" arrives with the name of that product written next to it, the competitive picture starts to appear.
4. A follow-up question, asked based on the first answer. Someone switching gets asked which product; someone missing a feature gets asked which feature. Two questions per person beats five questions for everyone — it protects the response rate and the usefulness of the answers at the same time.
5. Account details, which you attach rather than ask for. Store the plan, how long the account existed, and how many seats it had alongside every response. Without them every answer carries the same weight. "Too expensive" from a free-trial account that left after two weeks calls for something completely different than the same words from a 30-person team that paid for two years.

Designing and running the survey usually sits with someone on growth. One person owns the wording of the options and the decision, made with the product team, about where in the flow it appears. The draft of the option list is best written with customer success or sales in the room, since the people who hear the same complaint every day know the words customers actually use.

Four groups read the results. The product team looks at which missing features drove people out and uses that when picking next quarter's work. Whoever sets pricing looks at which plan the price-related answers cluster in. Customer success and sales learn which kinds of customers leave for which reasons, and reach out early to accounts showing the same signs. And whoever designs the save offer mines this data for which offers work against which reasons.

A monthly cycle suits the summarizing. A week's worth is too small a sample, so the ranking flips every week; once a quarter is too slow to react. Circulate one page a month with the share of each reason and the change from last month, and churn becomes something the company discusses in sentences rather than in a single percentage.

"Missing features moved to number one in last month's cancellation survey."

The cause to chase is in the product itself, not in the price.

"Only 40 percent answer the cancellation survey. How much can we trust it?"

The people who answered may differ from the people who didn't, so be careful with conclusions.

"Let's drop that option. Even when people pick it, there's nothing we can do."

An option that doesn't lead to an action just splits the answers.

When Ravi first put a survey on Pagedeck's cancellation screen, 180 responses came in that month. Nearly half said the product was too expensive, and the team started discussing a price cut immediately. Then they looked at the account details attached to the responses: most of those answers came right after a free trial ended, and among teams who had actually been paying, only ten said price.

Ravi recounted with the two groups separated. Among paying teams the top reason was a missing feature, and the free-text box had the same feature name written in it over and over. Mia moved that feature up for the next quarter, and the free-trial side got redefined — not a pricing problem, but a failure to show the product's value inside the trial window.

The most common misreading is treating the survey as a device for preventing cancellations. It sits on the cancellation screen, right beside the save offer in the cancellation flow, so it looks like one. But this survey doesn't try to keep the person in front of it. Keeping them is the save offer's job; the survey leaves a record for the next person. Blend the two and start slipping discount offers between the questions, and the respondent can't tell whether this is a survey or a negotiation, so they click any option to get out. Now you've failed to keep them and lost the data too.

The other mistake is reading the collected answers as causes exactly as written. People pick whichever explanation is easiest to pick. "Too expensive" comes first in nearly every product, and a large share of those answers don't mean the price is high — they mean the value didn't feel like it matched. Take that at face value, cut prices 20 percent, and churn stays where it was while revenue drops. The survey points at where to dig; the actual cause takes one more layer of usage data or a direct conversation to confirm.