The insight the whole flow rests on: your subscription platform already tells you why each customer cancelled. Recharge's subscription cancelled event carries the cancellation reason, and Skio, Stay.ai, and Smartrr emit equivalents. A price objection, a too-much-product objection, and a not-working objection are three different problems with three different fixes, and a cancel flow that sends everyone the same 20% off answers one of them while insulting the other two. Branch on the reason, and the stakes are bigger than the save itself: Recurly's data shows nearly one in four new subscriptions comes from a previously canceled customer, so how the cancellation ends shapes the next year of revenue either way.
Before the flow: the in-portal save comes first
Honesty about sequencing: the highest-converting save surface is inside the cancellation portal itself (the pause, skip, and swap offers your subscription platform shows before confirming). The email flow is the second line, catching people who completed the cancellation anyway. Build the portal offers first in the platform; this flow assumes they exist and failed.
The build, branch by branch
Trigger: the subscription cancelled event. First email within 1 to 4 hours, while the decision is warm. Then branch on the reason payload:
Branch 1: too much product / stockpiling
The most recoverable cancellation in consumables, because the customer likes the product; they just have too much of it. The save is never a discount: it is the pause and the frequency change, offered in one click. Ship every two months instead, skip the next one, downsize the bundle. The pause deserves top billing: pause-before-cancel adoption is up 337% in Recurly's data, and roughly three of four customers who pause return within months. This branch alone justifies the flow in most supplement and CPG accounts, and the fix costs zero margin.
Branch 2: price
The branch where a discount is actually the right tool, structured to protect the base: a one-time credit on the next renewal or a temporary rate rather than a permanent markdown, and framed as we want to keep you, not as confession that the list price was fake. Recurly's framing matches: a subscriber canceling over cost responds to a concrete offer, while one citing low usage needs a pause, and the reason they give should shape the response. Guard it mechanically: once per customer per year, and excluded for anyone whose cancellation pattern suggests offer-farming (cancel, save-offer, cancel again).
Branch 3: not working / results
The product-belief cancellation, and in supplements the compliance-sensitive one. The save is education plus honesty: results timelines restated, usage checked (the adherence problem from the post-purchase flow often surfaces here), and where genuinely fit, the swap to a different product for the same goal. What never goes in this branch: outcome promises to argue them out of their experience. If it did not work for them, the graceful exit below serves the brand better than a discount to keep taking something they believe failed.
Branch 4: no reason given / other
The generic branch gets the softest touch: one email acknowledging the cancellation cleanly, the pause option mentioned once, and the door left open. Resist the urge to interrogate.
The second email, all branches: the graceful exit
Two to three days later, whoever was not saved gets the exit done right: cancellation confirmed plainly, what they keep (account, history, any points), how restarting works in one click, and a genuine thank you. This email converts nobody today, and it is how you end up on the right side of the one-in-four statistic: the cancelled subscriber who returns in three months instead of buying from a competitor is the one whose last impression was competence, not clinging.
Routing and hygiene
- Cancelled subscribers exit all subscription-lifecycle flows immediately (pre-billing notices to a cancelled subscriber are the classic bug), and enter the one-time-buyer track: replenishment reminders on their product's cycle, per the setup guide, which quietly becomes the long-cycle winback.
- Suppress the save flow for involuntary churn: failed-payment cancellations are the dunning flow's job, a different problem entirely.
- Cool-down between save offers, tracked as a profile property, so serial cancellers cannot farm branch 2.
What to measure
- Save rate by branch, which is the whole point of branching: expect the pause branch to lead and the results branch to trail, and investigate any inversion
- 90-day survival of saved subscribers: a save that churns again in one cycle was a deferral, and if branch 2 shows this pattern, the discount is masking a value problem
- Return rate of gracefully-exited customers at 90 and 180 days, the number the exit email exists to move
- Portal saves versus email saves, so credit lands where it belongs and the email flow is judged on its real increment
Frequently asked questions
How fast should the first save email send?
Within 1 to 4 hours of the cancellation event. The decision is still warm, and the pause alternative is most persuasive before the customer has mentally moved on.
What is the highest-converting save offer?
For consumables, the pause and frequency change, because the most common real reason is too much product, and the fix costs no margin. Roughly three of four pausers return, which no discount matches. Discounts are the right tool only for the price branch.
What if our platform does not pass cancellation reasons?
Ask in the flow: first email poses the one-tap why question, and the click branches the sequence. One step slower, same architecture.
Should we try to save everyone?
No. The results-objection customer who is done deserves the graceful exit, and offer-farmers should be excluded mechanically. A save program's reputation with its own list is an asset worth protecting.