The cancellation survey is one of those things that sounds like it should be the best feedback you ever get. Someone is leaving, they're emotionally invested in the decision, they've got a real reason in their head, and you have one last chance to ask them what it is before they're gone. On paper, asking makes perfect sense. The data should be gold.
In practice, it rarely is. I've sat with founders going through six months of cancellation reasons and watched the same scene play out. "Too expensive" leads the list, then "I don't need it anymore," then "I'm switching to something else." A blank box where the user can elaborate, mostly empty. Most of the responses fit into one of a few buckets and most of those buckets are useless on their own.
The problem isn't that people are lying when they pick "too expensive." It's that "too expensive" is doing a lot of work in a single phrase. It can mean "the value isn't there for me at any price." It can mean "the value would be there if I were a different kind of user, and I'm not." It can mean "I joined for one specific reason that turned out not to apply, and the rest of what you offer doesn't matter to me." It can mean "I just got laid off and I'm cutting everything."
The cancellation form doesn't distinguish between those. So you end up with a leaderboard of words that look like reasons but are really categories of reason, and a roadmap based on that leaderboard will lead you somewhere that doesn't help you.
There's a second problem that's harder to see. The people answering the cancellation survey have already decided. They aren't writing to help you improve the product. They're writing because the form is in their way and they want to click through it as fast as possible. Whatever sits at the top of the dropdown will get more clicks than what's lower down. Whatever needs typing will be skipped. Whatever takes thought will get answered with the first plausible thing that comes to mind.
This is true of any survey, but it's especially true here, because the user is in a hurry to leave. You're surveying people whose attention you've already lost.
A lot of cancellation flows make the second problem worse by offering a discount or a "pause" option somewhere in the middle. The intention is fine: give people who can be saved a path back. The effect on the data is awful. Now you don't know if someone said "too expensive" because it's true or because they spotted the discount up the road and were curating their answer to qualify. You don't know if the person who clicked "I'll come back later" actually intends to. The flow has become a negotiation, and surveys-as-negotiation produce answers shaped by what the user thinks will get them what they want.
The cancellation survey isn't the place to learn why people are leaving in a way that helps you change things. It's the place to confirm patterns you've already picked up elsewhere, and to flag categories worth investigating somewhere else.
The somewhere else is what most teams skip.
What you actually want to know is the moment the user stopped using the product. That moment usually happens weeks before the cancellation. By the time someone clicks "cancel subscription," they've spent some period not logging in, or logging in and not finding what they came for, or logging in and finishing what they came for and realising they don't need to come back. The cancellation is the receipt for a decision that happened earlier.
That earlier decision is where the useful feedback lives. Catching it means asking active users about specific friction points while they're still active, not asking departing users to reconstruct a story they've already half-forgotten. Short, contextual questions inside the product, at the moments where dropoff usually happens, will tell you more in two weeks than a cancellation survey will tell you in a year. Why timing matters as much as the question covers this point in more depth.
A related thing worth doing: actively re-engage cancelled users a month or two after they've left. By that point they're not in a hurry. They've either signed up for something else and have an opinion about how it compares, or they haven't and they have a clearer view of what they actually missed. A short, non-recovery email asking what they're using now and what's working better, with no offer attached, will get more honest answers than the survey did at the moment of departure.
Cancellation surveys still have a place. The data is cheap to collect, and the categorical patterns will tell you something at the rough level. Watch the categories for sudden shifts. A spike in one bucket from one week to the next is worth investigating even when the bucket itself is vague.
Qria routes cancellation responses into the same themes as your active-user feedback, which matters more than it sounds like it should. A "too expensive" coming in at the exit is read against the same theme as "too expensive" coming in from someone still using the product. Words that read one way at the exit mean something else when they come from a user who's still on the platform, and seeing both in one view tends to surface the cases where you've miscategorised a real product issue as a pricing one.
The people who answer the cancellation survey aren't really your customers anymore. The people whose answers matter most are the ones still around, who haven't cancelled yet, and who you've probably never asked the right question to.
