Quick answer
Win-back performance is set by cadence and suppression discipline, not by the size of the offer. Segment churned accounts by why they left, because each reason has a different useful window. Send only when something has genuinely shipped, cap attempts, and treat the suppression list as the compliance artefact. Check your consent basis first: a former customer is a former customer, and regulators do not treat that relationship as permanent.
Segment by churn reason first
“Churned” is not a segment. It is five problems wearing one label, and each has a different useful window.
- Involuntary (card failure). The account never decided to leave. The window is days, and the fix is payment recovery, not marketing.
- Unmet need. They wanted a capability you lacked. The window opens the day you ship it, and stays shut until then.
- Budget. Nothing about the product was wrong. The window follows their fiscal calendar, not yours.
- Champion left. Two windows exist: the account, and the person at their new company.
- Competitor switch. The longest window — it usually opens at their renewal date with the competitor, which you can often infer from their contract term with you.
A win-back is not a save. A save happens while the customer is still yours and still deciding, a detection problem we covered in our piece on churn-save agents and cancellation intent. A win-back happens after the relationship has ended — which is exactly what makes it legally harder, not easier.
Timing: get the window from your own data
Two shapes are defensible and most teams need both. The short-window touch runs inside the first fortnight; it is only honest for involuntary churn and for cancellations whose stated reason you can fix immediately. The “something changed” touch runs months later, triggered by an event on your side — a shipped feature, a supported integration, a removed limitation — matched to the reason a cohort gave when they left.
Plot twelve months of reactivations against days-since-cancellation, split by churn reason. You will usually find two clusters and a long dead zone between them, and that dead zone is where most programmes waste their attempts. The discipline behind follow-up pacing on live pipeline applies here with a stricter cap.
The only durable message is a shipped change
A discount tells a churned buyer they were overpaying the whole time, and it selects for the accounts that will churn again at the next renewal. What converts is narrow and verifiable: you left because X; X now works; here is the changelog entry. If you are not capturing a structured churn reason at cancellation you cannot build this message later, and you are reduced to sending everyone the same discount. The cancellation flow is where the campaign is actually built.
None of this is exempt for being aimed at former customers. ACMA’s guidance on common telemarketing and e-marketing mistakes lists customer retention messages as a complaint category: “We have had complaints about alleged e-marketing messages sent by businesses attempting to regain customers who have cancelled subscription services (retention messages). These messages are commercial and must comply with the Spam Act.”
Retry logic and suppression
Involuntary churn has retry rules you do not set
For failed-payment recovery the retry ceiling is not purely a product decision. Visa’s rules bulletin on declined transaction resubmission, dated 3 September 2020 and effective 17 April 2021, states that when “a Category 1 (Issuer will never approve) decline code is used, the merchant is not permitted to reattempt the transaction”, and moved several codes into Category 2 (Issuer cannot approve at this time) specifically “to allow merchants to reattempt up to 15 times in 30 days”. It adds that “Merchants must not reattempt any transaction using the same account number following a decline for Response Code 14”. That bulletin is addressed to Canada, Europe, LAC and the U.S., so confirm what applies in your own region — but read the Visa bulletin before letting a dunning tool’s defaults stand. Network rules and a processor’s retry defaults are not the same thing: branch on the response code category, not a fixed calendar.
Suppression is the artefact that matters
Four controls do most of the work: max attempts per window, counted per person and per channel with a programme-level ceiling on top; cool-down between attempts and a longer one after an unanswered sequence; permanent suppression on request, which must survive CRM migrations, exports and re-imports; and cross-channel honouring, so an opt-out given on a call suppresses the email and the SMS too.
Under the Spam Act every commercial message must carry an unsubscribe facility that, in the ACMA’s words, “honours a request to unsubscribe within 5 working days”, “does not require the payment of a fee”, and “is functional for at least 30 days after you sent the message”. In the United States, the FCC’s news release on its February 2024 Report and Order (FCC 24-24) says the new rules require “that robocallers and robotexters honor do-not-call and consent revocation requests within a reasonable time, not to exceed 10 business days from receipt”. That US clock runs from receipt, so a suppression list reconciled monthly is already too slow.
The consent question is the hard part
Australia
ACMA says inferred consent is “usually when a person has a provable, ongoing relationship with your business, and the marketing is directly related to that relationship”, and that it “does not cover sending messages after someone has just bought something from your business”. A customer who cancelled is hard to fit inside an ongoing relationship on any ordinary reading. That is the whole problem.
ACMA’s Statement of Expectations on the use of consent in telemarketing and e-marketing [DOCX] puts an age on it. The document states up front that it “is not legal advice nor is it a definitive compliance guide to the Rules” — it is guidance, not law. On inferred consent it tells businesses to “use it only where there is a clear, current or ongoing relationship with the individual and the goods or services being marketed are directly related to that relationship”. Among the practices it lists as consumer-unfriendly: “Do not use consent that is old, where a consumer would not expect it to still apply (for example, consent to receive telemarketing that is more than 3 months old becomes stale unless a consumer has agreed to a longer period under terms and conditions)”. It also says: “Do not re-contact consumers who have unsubscribed to encourage them to resubscribe, unless they have subsequently given their consent to do so.”
So if your terms never said how long consent would be used for, ACMA’s stated expectation is that telemarketing consent goes stale at three months — shorter than almost every win-back window in circulation. Calls are treated harder than messages, too: the same document says “Commercial calls made using an auto-dialler (robocalls) or that use pre-recorded or AI voices are all telemarketing calls.” There is no AI carve-out; we covered the enforcement side in our post on the Do Not Call Register and AI voice agents.
United States
The FTC’s Telemarketing Sales Rule has an established business relationship exemption, and it is narrower than most people assume. Sellers “may place live telemarketing calls from a sales agent (but not automated calls or robocalls) to a consumer with whom a seller can demonstrate it has an established business relationship, provided the consumer has not asked to be on the seller’s entity-specific Do Not Call list”. The transaction-based version applies “within 18 months preceding a telemarketing call”, and “The 18-month period runs from the date of the last payment, transaction, or shipment between the consumer and the seller.” An 18-month clock, live agents only, and an entity-specific request overrides it — so the consent basis and its date belong on the record as a field, not as a judgement call at send time.
Channel comparison
| Channel | Reach after churn | Consent posture | Evidence quality | Suppression risk | Where it fails |
|---|---|---|---|---|---|
| Win-back email | Highest, but decays on a dormant list | Spam Act consent still required; ACMA names retention messages as commercial | Weak — opens are unreliable; only clicks and replies count | Moderate — unsubscribe must work and be honoured within 5 working days | When they left over the product, not attention |
| SMS | High and immediate, but intrusive | Same consent test as email | Good — a reply is unambiguous signal | High — ACMA flags unsubscribe instructions asking people to reply to an alphanumeric Sender ID, which generally cannot receive replies | No room for a “here is what changed” story |
| AI voice agent | Moderate — bounded by calling hours and register washing | Hardest. ACMA: “Commercial calls made using an auto-dialler (robocalls) or that use pre-recorded or AI voices are all telemarketing calls.” The FTC exemption covers live agents “but not automated calls or robocalls” | Best — a recording captures the reason in the person’s own words | Highest — opt-out must be actioned on the call, not batched | Anywhere you lack express consent, or the number is registered and no exemption applies |
| Human AE outreach | Lowest — headcount-bound; only the top of the list | Best in the US, where the exemption is written for live agent calls; still register-bound in Australia | High in the room, inconsistently written down afterwards | Underrated — suppression depends on a rep remembering to log it | Cost per attempt, and the long tail it can never reach |
Measurement: reactivation by reason, plus a holdout
Report reactivation rate split by churn reason. A blended number hides the only finding that matters: your involuntary cohort reactivates at a completely different rate to your competitor-switch cohort, and averaging them tells you to do the wrong thing to both. Then hold out a randomised slice of each cohort and contact no one in it. Some churned customers return on their own; without a holdout you will bill that to the sequence and scale a programme that is mostly measuring natural return.
One figure to avoid: the claim that it costs five times more to acquire a customer than to retain one is repeated everywhere without a traceable origin. Ipsos Loyalty devoted a chapter of its book Loyalty Myths to it; the published excerpt (© 2005) notes that “it is difficult to determine the exact origins of this platitude” and that the earliest sources they could find attribute it to late-1980s work by the Technical Assistance Research Project, before concluding that “supporting any retention strategy based in whole upon this myth is a recipe for financial disappointment” (Loyalty Myth #8). Saying plainly that it has no traceable source is stronger than repeating it.
FAQ
How long after cancellation should the first win-back touch go out?
There is no defensible universal number. Derive it by plotting reactivations against days-since-cancellation, split by churn reason. Timing is also constrained by consent. ACMA’s Statement of Expectations — guidance rather than law; it states it “is not legal advice nor is it a definitive compliance guide to the Rules” — lists as consumer-unfriendly the use of “consent to receive telemarketing that is more than 3 months old”, which it says “becomes stale unless a consumer has agreed to a longer period under terms and conditions”.
Can we call a churned customer whose number is on the Do Not Call Register?
The ACMA states that after a number has been on the register for 30 days, telemarketers can only call it if you have consent or if they are exempt, and that “Even if you give a company consent to call you, you can tell them at any time if you want to withdraw that consent.” Past custom is not consent, and the register does not block calls — it lets businesses wash their lists before calling.
Does an established business relationship let us send an automated win-back call in the US?
No. The FTC’s guide to Complying with the Telemarketing Sales Rule is specific that the exemption covers “live telemarketing calls from a sales agent (but not automated calls or robocalls)”, and applies only where the consumer “has not asked to be on the seller’s entity-specific Do Not Call list”.
How quickly do we have to action an opt-out?
In Australia a Spam Act unsubscribe request must be honoured within 5 working days, and the facility must stay functional for at least 30 days after the message was sent. In the United States, the FCC’s news release on its February 2024 Report and Order (FCC 24-24) says robocallers and robotexters must “honor do-not-call and consent revocation requests within a reasonable time, not to exceed 10 business days from receipt”. That US clock runs from receipt, so batching suppression into a monthly reconciliation is already too slow.
Where to go from here
Zian AI builds autonomous sales agents running live phone, SMS, email and WhatsApp outreach in 30+ languages, with CRM integrations. SmartReach AI™ orchestrates message, channel and timing by country, industry and profile — the part that matters for win-back, because the cadence, cool-down and suppression state are the product, not the copy. If you are rebuilding a win-back programme and want consent, cadence and suppression logic designed in from the start rather than bolted on, Apply For Partnership.