Cancel Flows That Convert

A cancel flow is what a subscriber sees after they click cancel and before the subscription actually ends. Its job is to find out why they're leaving and, where there is something you can reasonably do about it, to offer that thing before they go.

The reason is what makes the rest of it work. Someone whose project just wrapped and someone who thinks your price climbed too far are both standing on the same page, and almost nothing you'd show one of them is right for the other. Build the flow backwards from the reasons your own subscribers give: ask first, then answer what they told you.

Recurly asked consumers who had canceled something in the previous twelve months why they did it, and allowed more than one answer, so the ordering matters more here than any individual share.

51% said they were not using the subscription enough, against 45% who said it was too expensive (Recurly State of Subscriptions, 2026). Disuse edges out price, and the two call for opposite responses. A discount lowers the price of something the subscriber has already stopped opening. A pause stops the charge and leaves the account where it was.

Further down, the reasons stop being about value and become about the product itself: expectations that were not met, support that did not help, bugs, billing nobody could follow. Those aren't offer problems, and it usually pays to separate them out early instead of pointing one discount at all of them.

If the most common reason for canceling is that someone stopped using the subscription, pause is usually the response that fits. It's the one offer built for a need that has paused instead of ending: a season closing, a project wrapping, three months abroad.

Consumers already behave this way where they are given the option. 58% say they have paused a subscription instead of canceling it, and 79% say an easy pause is something they need before they'll subscribe at all (Chargebee Global Consumer Insights, 2026). The merchant side agrees. Where a pause option exists, 25% of subscribers take it instead of canceling (Recurly State of Subscriptions, 2025), and 38% of consumers say they would prefer pausing to canceling outright (Recurly State of Subscriptions, 2026).

Read the resumption figure carefully

Recurly reports an unpause rate of 61.9%, and that is a peak month rather than a steady-state rate (Recurly State of Subscriptions, 2026). It is good evidence that a paused subscriber is a plausible returning subscriber, and it isn't a promise about how many of yours will come back.

The decision you actually have to make is duration. Offer pause lengths that match how your own customers go quiet, a few weeks for a meal kit and a quarter for a B2B tool, and set the resume date automatically so nobody has to remember to come back.

45% of cancellations cite price (Recurly State of Subscriptions, 2026), and the reflex is to reach straight for a discount. There are two cheaper rungs to try first.

The first is a smaller plan. A subscriber who says you're too expensive is often telling you they are on the wrong tier rather than finished with the product, and moving them down keeps a paying customer instead of losing one. The usual worry is that a cheap tier fills up with worse subscribers. Streaming is the clearest place to check that, and it does not really bear the worry out: twelve-month survival between ad-supported and ad-free tiers is close either way, and which tier leads depends on the cut. Netflix's ad tier is slightly ahead at 54% against 52%, while the category as a whole runs the other way at 39% ad-free against 37% ad-supported (Antenna State of Subscriptions: Adds and Ads, 2026). The margins are small and they do not agree on a winner, so the tier barely moves survival in either direction.

The second rung is the billing cycle. An annual renewal arrives as one large charge, and a subscriber who blanches at it may be perfectly happy paying monthly for the same thing. Switching the cycle lowers the bill in front of them without lowering your rate.

What a discount actually costs

Only then the discount, and it helps to know what you're giving up. Customers who churn and later return mostly do not return cheaper: 42% come back on the same plan and 33% on a higher one, with a median revenue change of nothing at all (ChartMogul SaaS Winbacks Report, 2026). A discount granted at the cancel screen is a concession you keep paying every cycle, while a clean exit followed by a win-back usually isn't. Keep the discount for the subscriber the first two rungs didn't reach.

A good deal of the reason list has no offer behind it. Someone who switched to a competitor, whose expectations were not met, who had a bad support experience or hit a bug they could not get past isn't going to be talked round by money. Taken together those account for a substantial part of the list (Recurly State of Subscriptions, 2026), and a flow that answers all of them with the same discount is both ineffective and slightly insulting.

What those subscribers give you instead is worth more than the save you were never going to make. While a departing customer will rarely fill in a survey you send them afterwards, most will pick a reason on the way out, and the free-text box under something like bad experience or switched to a competitor is usually the highest-quality churn data a subscription business collects. A pattern in that text is a roadmap item with the demand already attached to it.

So route those reasons to a feedback step instead of an offer, and read what comes back. If unmet expectations climbs three months running, you have found something no amount of save-offer tuning would have surfaced.

Three things make a flow work, and none of them is the copy on the screen.

The first is a set of reasons that are genuinely distinct. If two options would lead you to the same offer, they are one option. Five or so tends to be about right: enough to route on, and few enough that people read the list instead of picking the first plausible line.

The second is an ordered response per reason. Not one offer but an order, with the strongest match first, a fallback behind it, and a feedback step at the end so the subscribers you can't save still leave you something. The table below is the default SubJolt ships, and it is meant to be edited, not accepted as it stands.

Keep the exit reachable

The third is that canceling still has to work, which is where flows most often go wrong. Compliance is the usual reason given for getting this right, and it is not the only one: 82% of consumers say they need easy cancellation options before they will subscribe in the first place (Chargebee Global Consumer Insights, 2026). Making the exit hard costs you subscribers you never acquired, in exchange for a few you kept against their will. A flow can sit in front of the cancellation mechanism, but it can't become a condition of reaching it.

Built this way it works. SubJolt's cancel flows can save 20-40% of would-be cancellations. The ones it doesn't save leave cleanly, having told you why, which is worth more than a reluctant extra month.

One number isn't enough. A blended save rate tells you the flow is working or it is not, and nothing at all about what to change.

Measure save rate by reason and by offer, as a grid rather than a total. That is the view that shows pause carrying the disuse bucket while the discount goes to people who were never going to take it, and it is what makes the flow tunable. Watch the reason mix over time too, because it moves, and when it moves the cause is usually in the product rather than in the flow.

The win-back window is short

Keep track of what happens to the ones you did not save. Returns are front-loaded: 45% of customers who come back do it within 30 days of churning and 66% within 90, with a median gap of 38 days (ChartMogul SaaS Winbacks Report, 2026). A reactivation offer aimed at a six-month-old cancellation is mostly aimed at people who have already made other arrangements.

For the first month, get the reason data flowing and leave the offers alone. You'll design a better flow from four weeks of your own reasons than from any benchmark on this page.