Winback offer economics: how deep should the offer go
Set winback depth from the return curve, not from a CAC multiple. Antenna's panel shows 23% of 2024 cancels return within 3 months, 32% within 6 and 42% within 12, so more than half of all eventual returns happen before most sequences have spent a cent. Recurly reports 1 in 4 new signups are returning subscribers. The offer's job is to move the tail, not to pay people who were coming back anyway.
The 30-day winback sequence post covers timing and templates: what to send at day 7, 21 and 60, and what the subject lines should say. This post is the other half. What goes inside the day-60 email, how much it should cost, and how to tell whether it earned anything.
The cost multiple you should stop repeating
The standard opener for a winback article is that reacquiring a lapsed customer costs a fraction of acquiring a new one. Five times cheaper, seven times, twenty-five times, depending on the article.
We could not find a primary study behind any version of it. The multiples trace back through vendor blogs to other vendor blogs. Churnkey's own reactivation guide carries a "3x to 25x" figure and attributes it to a marketing blog rather than a study. Its retention KPI page claims merchants can "win back up to 34% of former customers" with no source at all. Churnkey publishes genuinely good primary data elsewhere in this post, which is exactly why the unsourced claims are worth flagging.
There is real academic work on winback, and it does not measure cost per reacquisition. Thomas, Blattberg and Fox (Journal of Marketing Research, 2004) modelled reacquisition price. Kumar, Bhagwat and Zhang (Journal of Marketing, 2015) modelled reacquisition likelihood and second-lifetime value. Neither answers "what does a winback cost versus a new customer", because that comparison depends entirely on a merchant's own channel mix.
So: if you are pitching a winback program internally, pitch it on the return curve. That number exists.
Returning subscribers are already a large share of signups
The most robust finding in this space is that reacquisition is not a marginal channel. It is a large, mostly unmanaged part of normal acquisition.
- Recurly's 2026 State of Subscriptions, built on 76 million subscribers across roughly 2,200 merchants, states that 1 in 4 new signups are now returning subscribers.
- Antenna measured 30% of 2023 Premium SVOD gross additions as resubscribers, defined as users who had previously subscribed to and since cancelled the same service within the prior twelve months. Apple TV+ ran highest at 37.2%; Netflix fell to 26% in 2023 from about 35% the year before.
- The gap this creates is visible in the churn rate itself. Antenna put Premium SVOD gross churn at 5.3% in September 2024 against net churn of 3.1%. Resubscribers were quietly cancelling out more than two points of monthly churn.
Read that last line carefully, because it changes the framing. Most merchants already run a winback program. They just run it passively, and they book the results as new customer acquisition. The question is not whether to start winning customers back. It is whether the offer you attach adds anything to what happens without you.
The return curve is the offer schedule
Antenna's win-back curve on 2024 cancels is the closest thing to a planning instrument the public data offers: 23% return within 3 months, 32% within 6, and 42% within 12. Netflix reached 42% within 6 months and 50% within 12; HBO Max and Hulu each recovered 44% within 12.
The shape matters more than the endpoint. Of the 42 points of eventual twelve-month return, 23 land in the first quarter. Returns decelerate hard after month 3 and again after month 6.
| Window after cancel | Cumulative return | New returns in window | What the offer is buying |
|---|---|---|---|
| Months 0-3 | 23% | 23 pts | Mostly nothing. They were coming back. |
| Months 4-6 | 32% | 9 pts | Some incremental lift, shallow offer |
| Months 7-12 | 42% | 10 pts | Genuine incrementality, deeper offer defensible |
| Beyond 12 | not published | - | Treat as cold acquisition |
This is streaming data, where re-entry is one tap and content release schedules pull people back. A WooCommerce store with a physical product and a re-onboarding cost will sit below these numbers. But the shape is the transferable part, and the shape says the same thing every time: a deep discount sent in week two is spent almost entirely on people who had already decided to return.
That is the single strongest argument for the day-60 placement in the sequence post. Not that day 60 converts better, but that day 60 is the first point where the concession is plausibly buying something.
Offer depth by recency
Nobody has published an experiment varying winback discount depth by time-since-cancel. We looked. If you see a table proposing "10% off at 30 days, 40% off at 120 days", it is somebody's judgement, not a finding, and it should be labelled as such.
What can be derived from published numbers is a break-even hurdle, and that is more useful anyway.
Churnkey's voluntary churn benchmarks put monthly churn at 12.0% for subscribers under three months. A reacquired subscriber restarts at the top of that hazard curve, not in the flat 3.2% band where twelve-month veterans sit. At 12% monthly churn, expected second-lifetime duration is about 8.3 months. On a $49/mo subscription that is roughly $407 of gross second-lifetime revenue.
Now the hurdle. You pay the concession to everyone who redeems, including everyone who was returning regardless. So the offer pays for itself only when the incremental share of redeemers clears the concession as a fraction of contribution.
| Winback offer | Concession on $49/mo | Digital margin (~100%) | Physical margin (40%) |
|---|---|---|---|
| 20% off, 3 cycles | $29.40 | 7% of redeemers must be incremental | 18% |
| 30% off, 3 cycles | $44.10 | 11% | 27% |
| 40% off, 3 cycles | $58.80 | 14% | 36% |
| 50% off, 6 cycles | $147.00 | 36% | 90% |
The last row is the one worth staring at. On a physical subscription with a 40% gross margin, a 50%-off-six-cycles winback needs nine out of ten redeemers to be people who would otherwise never have returned. Against a curve where 23% return in the first quarter unprompted, that is not achievable. The offer is a transfer, not an investment.
The same arithmetic explains why deep winback offers feel like they work. Redemption is high, the dashboard shows recovered subscribers, and the counterfactual is invisible. The winback revenue calculator lets you run your own price and margin through this; the honest version requires a holdout group, which almost nobody runs.
What the academic work says about offer type
Kumar, Bhagwat and Zhang analysed eight years of a US telecommunications firm's data, 2006 to 2014, modelling reacquisition likelihood, second-lifetime duration and second-lifetime profitability per month against the reason for defection and the type of winback offer. The specific coefficients sit behind a paywall, but the rank ordering was reported publicly and it is counterintuitive enough to be worth acting on.
- A bundled offer combining a discount and a service upgrade was the most effective at reacquiring customers. Those customers then left again the soonest and spent the least.
- A service upgrade alone was the least effective at winning customers back. Those customers stayed the longest and spent the most.
- Customers who left over price were more likely to return and stayed longer once back. Customers who left over service spent more per month once back. Customers who left over both switched to a competitor fastest and spent the least.
The lesson is that maximizing reacquisition rate and maximizing second-lifetime value are different objectives that pull in opposite directions. If your winback dashboard reports reactivations and nothing else, you will optimize your way into the bundled offer and the worst cohort.
For a WooCommerce merchant the "service upgrade" analogue is concrete: free expedited shipping, a tier bump held for the second lifetime, early access, a concierge onboarding call for a membership. These convert worse than a discount and are worth more per convert.
The second-lifetime price problem
Thomas, Blattberg and Fox modelled reacquisition price and second-tenure duration together and concluded that the optimal approach is a lower price to win the customer back, followed by higher prices once the relationship is re-established.
That is a straightforward description of what a time-boxed winback discount does, and it is also a warning. The returning subscriber is price-anchored on the reacquisition price. If your winback offer is 40% off and never ends, you have permanently repriced that customer at 60% of list and they will treat the next increase as a betrayal rather than a return to normal. Cap the cycles. Say the cap in the email.
Does a winback discount train churn-and-return?
This is the most repeated warning in retention writing, and we could not verify it. No source we could reach measures whether customers who receive a reactivation discount cancel again in order to earn another one.
The nearest real evidence points the other way. Antenna's promotional cohort analysis found 2024 Black Friday signups, who took the deepest discounts in the market, retained at 36% at twelve months against 33% for other promotional signups. Discounted cohorts were not worse cohorts. Churnkey's in-flow data agrees in a different direction: customers who accepted a cancel-flow discount stayed 5.1 months longer on average, and 11% were still subscribers over a year later after the discount had expired.
None of that proves winback discounts are safe. It does mean the "you will train them to churn" claim is currently an argument rather than a finding, and it should not be the reason you refuse to test an offer. The reason to keep the offer shallow is the break-even table above, which does not require any behavioural theory at all.
The decision rule
- Do not attach an offer before day 30. The first-quarter returns are largely free. Spending on them is the most common way a winback program produces a negative result while reporting a positive one.
- Size the offer from your margin, not from a benchmark. Run the concession against second-lifetime contribution. If the required incremental share is above about 25%, the offer is too deep.
- Cap the cycles and say so. Two to four cycles. The reacquisition price is not the new price.
- Test a non-discount offer against the discount. Shipping, a tier bump, a service concession. Expect lower redemption and higher second-lifetime value, and measure both.
- Hold back 10% of the churned file. Without a holdout you cannot distinguish an offer that works from a return curve that was always going to run. This is the single measurement that separates a real winback program from an expensive one. ChurnStop reports reactivations by cohort but does not randomize the holdout for you; that split has to be made deliberately when the sequence is configured.
- Report second-lifetime revenue, not reactivation count. The academic work is unambiguous that the offer maximizing one degrades the other.
Winback is worth running. The 42% twelve-month return curve makes that clear enough. What is not clear, and what almost no merchant measures, is whether their offer moved it.
What's next
- The 30-day winback sequence for the timing, the three emails and the stop-rules.
- How deep should a save discount go for the same margin math applied inside the cancel flow, where the counterfactual is different.
- Reading cohort retention curves for why a reacquired subscriber re-enters at the cliff rather than the plateau.
Sources: Antenna, Resubscription is on the Rise and State of Subscriptions: Premium SVOD 2025 Year in Review, accessed August 2026 · Recurly, 2026 State of Subscriptions · Churnkey, voluntary churn benchmarks (November 2025) and discounting at cancellation (August 2025) · Kumar, Bhagwat and Zhang, "Regaining 'Lost' Customers", Journal of Marketing 79(4), 2015 · Thomas, Blattberg and Fox, "Recapturing Lost Customers", Journal of Marketing Research 41(1), 2004.