Subscription winback campaigns: the 30/60/90 playbook

What a winback campaign actually is (and isn't)

A winback campaign is a sequenced set of emails (and sometimes SMS) sent to subscribers who voluntarily cancelled, designed to convert them back to active subscribers — usually with an escalating offer and a hard expiration. It is not a one-off email. It is not the same as dunning (which targets failed-payment recovery on active subscribers). And it is not your generic re-engagement flow for inactive customers, because cancelled subscribers carry a specific signal that needs specific handling.

The distinction matters because the optimal cadence, copy, and offer differ. Dunning runs over 7-21 days and tries to fix a payment problem the subscriber didn't choose. Winback runs over 30-90 days and tries to reverse a decision the subscriber made on purpose. Treating them as the same flow is the single most common mistake — winback copy that reads like dunning ("your card was declined, please update") confuses the recipient, and dunning copy that reads like winback ("we miss you, come back") loses the urgency that recovers failed cards.

The 30/60/90 cadence (and why it works)

The default winback cadence that performs well across consumer-goods subscriptions is a three-touch sequence at day 30, day 60, and day 90 after cancellation. The reasoning isn't arbitrary. Day 30 is when the subscriber has run out of their last shipment and is feeling the absence. Day 60 is when they've either repurchased elsewhere or noticed they haven't bothered. Day 90 is your last reasonable touch — past that point, you're in cold-list territory.

Send earlier than day 14 and you look desperate, like the cancellation didn't register. Send at day 7 and the subscriber feels stalked. Wait until day 120+ and the relationship has gone cold; the open rate drops below your generic newsletter and conversion approaches zero.

1. Day 30 — the soft return. No discount yet. "Your last shipment of X is probably running low. Click here to resume — we'll start your next delivery whenever you want." Conversion: ~1.5-2.5%. The point is to remove the friction of resuming, not to bribe.
2. Day 60 — the offer. Modest discount (10-15%) on the next 1-2 shipments only. "Come back for the next two boxes at 15% off — same product, no commitment past those two deliveries." Conversion: ~2-4%. This is where the bulk of recovered revenue lives.
3. Day 90 — the last call. Stronger offer with a hard deadline (often 7 days). "This is the last winback we'll send. 20% off the next box if you resume by [date]." Conversion: ~1-2%. Below 90 days you're not winning back; you're discounting for free.

Total recovery across the three touches for a healthy program typically lands at 3-7% of cancelled subscribers — modest in percentage terms, meaningful in dollars because the cohort is large and the LTV is fully incremental.

Offer escalation: what to put in each email

The instinct is to lead with the deepest discount — "30% off, come back now!" — and call it a day. This is exactly what trains your existing subscribers to cancel and wait for the winback email. Offer escalation is the discipline of starting with the smallest possible incentive and stepping up only if the subscriber doesn't bite.

The escalation also signals scarcity and finality. By the third email, the recipient knows this is the best you'll do and there won't be a fourth. That's what creates the conversion bump on the day-90 message — not the size of the discount, but the framing of it as terminal.

Non-discount offers often outperform discounts when the cancel reason wasn't price. Free shipping for the next two deliveries, a bonus gift in the comeback box, an upgrade to a larger size at the smaller size's price, a free skip credit — all of these address the "I want to come back but I want a reason" signal without conditioning the cohort on discount-only return behaviour.

Segmentation: who gets the winback (and who doesn't)

Sending the same winback to every cancelled subscriber is the equivalent of running a single ad campaign to your entire customer base. Some segments will respond well, some won't respond at all, and some will mark you as spam. The minimum useful segmentation has three dimensions: cancel reason, tenure, and order count.

• By cancel reason. "Too expensive" → discount path. "Too much product" → cadence-extension path ("come back at every 6 weeks instead of every 4"). "Quality issue" → personal note from the founder, no auto-flow. "Moving / circumstance change" → pause-extension offer instead of winback ("resume in 3 months, no charge until then").
• By tenure. Subscribers who cancelled after 1-2 orders get a different message than long-tenure subscribers who cancelled after 12. Long-tenure cancellers have higher response rates and deserve a more personal tone; short-tenure cancellers may not have found the product fit at all and might respond to a sample of a different variant.
• By order count. Subscribers who never received a single order (signed up, cancelled before billing) should not be in the same flow as subscribers who churned at order 8. The former is a signup-flow issue; the latter is a product or fatigue issue.

Deliverability: keep your sender reputation alive

Winback emails go to a list that, by definition, has lower engagement than your active subscribers. Mailbox providers (Gmail, Outlook, Apple Mail) watch open rates, click rates, and complaint rates per sender domain. A bad winback campaign — wrong segment, too aggressive, low engagement — can tank your sender reputation and reduce deliverability on the rest of your transactional and marketing email. The cost of a poorly-run winback isn't measured in unrecovered revenue; it's measured in how many of your active subscribers' renewal-confirmation emails start landing in spam afterwards.

• Send winback from a sub-domain or sub-stream of your sending infrastructure (e.g. winback@ instead of your primary sending domain). If reputation tanks, your transactional renewal/dunning email is protected.
• Suppress hard bounces and spam complaints immediately. Don't include unverified email addresses in the sequence — they were valid when the subscriber signed up, but a year-old address has a meaningful bounce rate.
• Cap volume per send. If your cancel list is 50,000 people, don't drop the day-30 email on 50,000 at once. Stagger across 3-5 days so the per-day complaint rate stays under the mailbox-provider threshold.
• Make unsubscribe one-click and honor it instantly. The CAN-SPAM and GDPR rules are clear, but mailbox providers also weight unsubscribe behaviour into reputation — if your one-click unsubscribe doesn't actually stop the flow, Gmail will notice.
• Verify SPF, DKIM, and DMARC are set up on your sending domain. Without these, winback email to Gmail and Yahoo recipients (post-2024 policy changes) frequently goes to spam by default.

Channels: email, SMS, retargeting, none of the above

Email is the default channel for winback because it's cheap, async, and respectful — the subscriber opens it when they want. SMS is faster and has higher open rates, but it's a more aggressive channel and the unsubscribe cost is permanent (TCPA requires explicit re-consent). Retargeting ads can supplement both, especially for the day-60 and day-90 touches.

• Email — always. The primary channel. Costs nothing per send, doesn't burn relationship capital, easy to A/B test subject lines and offers.
• SMS — sparingly. Only for subscribers who explicitly opted into SMS during signup, and only for the day-60 or day-90 touch (not day 30). One SMS per cycle, not the full sequence.
• Retargeting ads — supplemental. Upload the cancelled-subscriber list to Meta/Google Ads as a custom audience. Run a low-budget retargeting campaign on the day-60 to day-90 window. The cost per recovered subscriber via ads is usually higher than email, but the combined open rate is meaningfully higher than email alone.
• Direct mail — for high-LTV only. A postcard to subscribers with $200+ historical spend, around day 45, converts at 2-3x the email rate. Per-touch cost is much higher, so reserve it for the top decile.
• Phone call — for the top 1%. If a subscriber spent $1,000+ over their tenure and cancelled, a five-minute personal call from the founder routinely converts at 30-50%. Doesn't scale, doesn't need to.

When winback stops being worth it

Every winback program has a point of diminishing returns. The question isn't whether to keep running it forever — it's at what tenure-since-cancel the math stops working, and at what point you should be reallocating budget to retention or acquisition instead.

The clearest signal is the conversion rate on each touch in your sequence. If the day-30 email converts at 1.5%, day-60 at 2.5%, day-90 at 1%, and you ran an experimental day-120 touch at 0.3% — you've found the floor. Anything below 0.5% conversion at a typical winback offer is below the cost of the send (email infra + complaint risk + the subscribers it costs at the next mailbox-provider engagement audit) and should be cut.

The other signal is cohort behaviour. If winback recovers subscribers who then churn again within 2 orders, the unit economics flip negative — you've paid for a discount and recovered nothing. Track post-winback churn separately; if it's more than 2x your baseline first-order churn, the program is recovering the wrong people and the offer needs to change.

What to measure (and what to ignore)

Winback metrics are noisy because the population is small and the conversion is delayed. The temptation is to measure open rate and call it a day — but open rate doesn't pay rent. The real KPI is recovered MRR per dollar spent, tracked at the cohort level over 90 days post-recovery.

• Recovered subscribers per cohort — count, not rate. A 3% conversion on 1,000 cancels is 30 recovered subs; 3% on 50 is 1.5. Tiny cohorts produce noisy rates; aggregate by month at minimum.
• Post-recovery retention — of the subscribers you won back, what percentage are still active 3 orders later? If it's below 40% you're paying to discount a one-time order, not building MRR.
• Revenue per winback email sent — total recovered LTV divided by total sends. This is the number that decides whether the program is worth running at all. Industry-typical range for consumer goods: $0.20-0.80 per email.
• Complaint rate per send — keep this below 0.1% (mailbox-provider threshold) or you'll lose deliverability across your entire sending domain. If your winback complaint rate is creeping toward 0.3%, your segmentation is wrong.
• Cancel-to-winback latency — the time between cancellation and first winback touch. If it's drifting past 35 days, something is broken in your trigger logic.

Common winback mistakes

Most winback programs underperform not because the concept is wrong but because the execution is rushed. The recurring failure patterns:

1. Leading with the deepest discount. Trains the entire subscriber base to wait for cancel-and-rejoin discount cycles.
2. One generic message regardless of cancel reason. Ignores the highest-signal data you already collected at cancel time.
3. Sending too soon. Day 7 winback reads as desperation and pre-empts the natural "I'm running out" moment that day 30 captures.
4. No expiration on the offer. Without a deadline there's no urgency; the email gets archived for "later" and never acted on.
5. Including quality-complaint cancellers. Spam complaints from this segment can tank deliverability on the rest of the program.
6. Running the campaign from your primary transactional domain. A winback reputation hit becomes a renewal-email reputation hit, and now active subscribers don't see their billing confirmations.
7. Not measuring post-recovery churn. Counting recovered subscribers without checking whether they stayed is the most common way to overstate the program's ROI.
8. Never ending the sequence. Subscribers in your winback list for 2+ years drag down your engagement metrics and never convert. Move them out.