The save-flow playbook.

What people actually say when they cancel.

Two things surprise most operators. First, how concentrated the reasons are: the top four categories account for about 77% of cancellations in Churnkey's exit-survey data (Churnkey, State of Retention 2025)and about 88% in RevenueCat's (RevenueCat, State of Subscription Apps 2025). Second, which reason comes first: it is not price.

1. Reason 1

   They stopped using it

   "Wasn't using it enough." "Got too busy this month." "On a break from the gym." This is the single largest category in both datasets — 30.6% of cancellations (Churnkey, 2025) and 37.0% (RevenueCat, 2025) — ahead of price. The save is usually a pause, not a discount: a customer who is not using it will not use it more because it got cheaper. Auto-resume in 30, 60, or 90 days.

2. Reason 2

   Price sensitivity (real or stated)

   "It's too expensive." About a third of stated reasons, and the most consistent finding across sources: 33.0% (Churnkey, 2025), 34.6% (RevenueCat, 2025), 31% (Chargebee, 2024). The save is a discount ladder — a modest offer first, a deeper one if they push back, then a downgrade tier at the bottom. We do not publish an acceptance rate by reason, because nobody measures one: see the sequencing section for the acceptance rates that are actually published.

3. Reason 3

   Product / fit mismatch

   "Not the right plan for me." "Don't need all these features." "Missing something I need." A smaller but real share of every published breakdown — the exact figure moves too much between providers to quote as one number. The save is a tier change or product swap, not a discount. Someone on the wrong plan does not benefit from paying less for the wrong plan.

4. Reason 4

   Genuinely done

   "Finished my project." "Moved cities." "Don't need it anymore." The save here is honest: don't try to save. Process the cancellation cleanly and capture the reason. It is not a wasted call — roughly one in four new subscription sign-ups is a returning subscriber (Recurly, 2026 State of Subscriptions), and people remember the teams that didn't fight them.

How to sequence the flow.

The single biggest mistake operators make is offering everyone the same thing. Sequence the conversation to classify the reason first, then match the offer. Four rules:

• Don't lead with price. It is not even the most common reason, and if you offer a discount to someone whose actual problem is fit, they will take the discount and still churn later — and you have trained them to expect a discount at renewal.
• Take the pause seriously. Recurly reports that three out of four subscribers who pause eventually return, and that pause usage rose 337% year over year across its 2,200-merchant, 76-million-subscriber base (Recurly, 2026 State of Subscriptions). When the reason is non-use, a pause addresses it and a discount doesn't.
• Calibrate your expectations on offer acceptance. Published acceptance rates for save offers are modest and clustered: about 17% for discounts (21% in B2C), 15% for pauses, and 14% for trial or term extensions (Chargebee, Q1 2024). A flow that assumes half of cancellers can be bought back is not a plan, it is a hope.
• Honor "genuinely done." Pushing back on a genuine cancellation creates the regulatory and reputational risk that ends save programs — and in at least one state it is now unlawful. Read the room, and read the next section.

How to measure it.

Three numbers. Anything else is a distraction at this stage:

• Save rateby cancellation reason. A 25% blended save rate looks fine until you discover it's 60% on price and 0% on product-fit — you have a fixable problem hiding inside an average.
• Post-save retention curve. What share of saved customers are still subscribed at 30, 60, 90 days? If your save rate is high but 90-day retention is bad, you're saving with offers that don't address the real reason.
• Offer mix. Which offer types are actually closing saves? You'll find one or two outliers — double down on those and prune the rest.

For context on the denominator: Recurly's benchmark data puts overall subscription churn at 3.60% per month, split 2.34% voluntary and 1.25% involuntary (Recurly Research, July 2026). A save flow only addresses the voluntary side; the involuntary side is a dunning and payment-recovery problem, and confusing the two is how retention programs get judged against a number they were never going to move.

Compliance, and why it decides your flow design.

This is the part most save-flow advice gets wrong or leaves stale, so here is the current position. None of this is legal advice — check it with your own counsel before you ship a flow.

• The FTC's "click-to-cancel" rule is not in force. The amended Negative Option Rule was vacated in full by the Eighth Circuit on July 8, 2025 in Custom Communications, Inc. v. FTC, on procedural grounds rather than on the merits, days before its compliance date. The FTC reopened the question with an advance notice of proposed rulemaking in March 2026 and has not proposed a replacement rule. Do not design on the assumption that it is gone: ROSCA and Section 5 of the FTC Act still apply and are still being enforced, and the FTC reported receiving nearly 70 consumer complaints a day about negative-option practices in 2024, up from 42 a day in 2021 (FTC, October 2024).
• California constrains the save offer itself. The Automatic Renewal Law (Cal. Bus. & Prof. Code §§ 17600–17606), as amended by AB 2863 (2024) and operative July 1, 2025, requires cancellation through the same medium the customer signed up in, and permits a retention offer only where a control that immediately effects the cancellation is displayed prominently alongside it. If cancellation is requested by telephone, the business must first tell the customer they can complete the cancellation at any point in the call.
• Minnesota bans the unsolicited save attempt. Minn. Stat. § 325G.58, effective January 1, 2025, prohibits presenting a retention offer without the consumer's permission — and permission only counts if it is obtained afterthe cancellation request. That is a constraint on the flow, not on the wording: in Minnesota, "before I do that, can I ask why?" has to come after the customer has agreed to hear it.
• Connecticut puts a clock on the callback. Conn. Gen. Stat. § 42-158ff, as amended by SB 3 (2025) and effective July 1, 2026, requires that a cancellation left by voicemail be processed, or the call returned, within one business day — and adds a private right of action. If your cancel line can go to voicemail, that is now a deadline.
• And the rest of the map. Colorado's one-step online cancellation requirement (Colo. Rev. Stat. § 6-1-732, as amended by SB 25-145) took effect August 6, 2025 and extends to business customers in February 2026; New York has its own regime at N.Y. Gen. Bus. Law § 527-a. More than 30 states now have at least one automatic renewal law. The short version everywhere: never make cancelling harder than signing up, honor a clear "I want to cancel" the first time you hear it, and log everything. A save flow is fine. A bad-faith save flow is what gets legislated.