Win-Back Email Flows: Time Them Off Churn Risk, Not Calendar Dates

Most win-back guides will tell you to agonize over the subject line of your day-90 email. They're optimizing the wrong variable.

A win back email flow earns its revenue from two things: triggering off a real churn signal (not a calendar date), and aggressively sunsetting the recipients who don't respond. The copy matters third. Get the first two wrong and the best "we miss you" line in the world lands in spam, because you spent six months mailing people who stopped opening in March.

This piece lays out the operator version: when to fire the flow, how to structure three messages without breaching complaint thresholds, when to suppress, and what to actually measure. It draws on Klaviyo's churn-risk model, Google's 2024 sender guidelines, and ZeroBounce's list-decay research — all linked below — plus the patterns we see running email programs for clients.

TL;DR

• Trigger your win-back off churn-risk probability or expected-next-order date, not an arbitrary 90-day mark.

• Klaviyo's churn-risk score activates after 500+ repeat customers and predicts who's actually leaving.

• Keep the flow to 3 messages over roughly 14-21 days to stay under Google's 0.3% complaint ceiling.

• Sunset non-openers after the final message; ZeroBounce data shows lists decay 23%+ annually anyway.

• Measure recovered revenue, but weight deliverability lift equally — the cleaner list pays out across every flow.

1. Why "we miss you" at day 90 is the wrong trigger

The day-90 convention is a holdover from when ESPs had no other signal to fire on. It assumes every customer has the same purchase rhythm. A coffee subscriber at day 90 is deeply churned. A mattress buyer at day 90 is in the honeymoon phase. Same email, opposite meaning.

The second problem is harder: a fixed-window trigger mails customers who are still perfectly engaged with your brand, just not buying yet. You're spending sender reputation on people who didn't need to be persuaded back. That cost is invisible until you breach a complaint threshold and your whole program tanks.

The trigger should be a probability, not a calendar. If your platform can predict who's leaving, fire off that. If it can't, fire off a personalized expected-next-order date built from the customer's own history.

2. Timing off the data: churn risk and expected next order date

Klaviyo's churn-risk prediction scores the probability a customer won't purchase again. It activates once a store has 500+ customers who've placed orders, which is the volume the model needs to learn purchase rhythm. Once it's live, you have a real signal: fire the win-back when churn probability crosses a threshold you set, typically somewhere in the 60-80% band depending on margin.

For stores below that volume, or non-Klaviyo stacks, use expected-next-order date. The math is straightforward — take each customer's average days between orders, add a buffer (we usually start at 1.5x), and trigger the flow when that date passes without an order. A customer who buys every 30 days gets the flow at day 45. A customer who buys every 180 days gets it at day 270. Same logic, calibrated to the individual.

This is also where most teams discover their data plumbing isn't ready. If your order history lives in one system and your email tool can't see it cleanly, the trigger is theoretical. Fix that first; the ecommerce email flows guide covers the wiring.

The win-back flow as one mechanism: recovery on the left, list hygiene on the right.

3. The 3-message win-back structure that respects complaint thresholds

Google's bulk sender guidelines, in effect since February 2024, hold senders to a 0.3% spam complaint ceiling with a 0.1% target. Mailing customers who've stopped opening is the fastest way to breach that ceiling — and once you breach it, every flow you run suffers, not just the win-back.

That constraint shapes the structure. Three messages, no more, spaced to give signal time to surface:

Message 1 — Day 0 (trigger fires). Soft re-engagement. Lead with what's new since they last engaged. No discount yet. The goal is to separate "I'm just busy" from "I'm gone." Roughly 40-50% of recoveries happen here in the programs we run.

Message 2 — Day 7-10. Acknowledge the silence directly and put one clear value proposition in front of them. This is where a modest incentive can enter if your margins allow. Keep it specific to their last purchase category when you can.

Message 3 — Day 14-21. The honest version: "We'll stop emailing if we don't hear from you." This message does two jobs. It recovers the last sliver of engaged-but-distracted customers, and it gives the rest a clean exit before you suppress them. Counterintuitively, this message often has the highest click rate of the three.

Nothing else after that. A 5-message win-back doesn't recover more revenue; it just generates complaints from people who already decided.

4. The sunset branch: suppressing non-responders before they hurt you

This is the half of the re-engagement email sequence nobody writes about, and it's the half that determines whether your program is healthy 12 months from now.

After the third message, anyone who hasn't opened, clicked, or purchased moves to a suppression segment. They stop receiving campaigns. They can still receive transactional mail. They're not unsubscribed — they're sunset.

Why this matters: ZeroBounce, analyzing 11 billion+ addresses, found at least 23% of email lists decay annually through job changes, abandoned addresses, and disengagement. That decay happens whether you act on it or not. The difference is whether the decay shows up as quiet suppression on your side or as spam complaints and spam-trap hits on the mailbox provider's side.

A sunset flow email marketing policy is maintenance, not optional cleanup. Run it monthly. Move the unengaged out. Your inbox placement improves on the next send, every active subscriber sees more reliable delivery, and your complaint rate falls back under Google's ceiling. If you're already seeing inbox issues, our piece on marketing emails going to spam walks through the deeper diagnostic.

The operator framing: a suppressed subscriber costs you nothing. An engaged subscriber who never opens costs you every other subscriber.

5. Win-back for non-ecommerce: lapsed clients and dormant leads

The same architecture works for B2B and services, with different inputs.

For a lapsed customer email campaign in a services business, the trigger is engagement decay, not purchase rhythm. Define what "active" means for your relationship — meeting attended, ticket opened, portal login, email opened — and fire the flow when those signals go quiet for 1.5x the customer's normal interval.

For dormant leads, the trigger is content engagement decay. A lead who downloaded three pieces in Q1 and zero in Q2 is a churn-risk signal in B2B terms. Three messages, same cadence, different value props: one with new research, one with a specific outcome from a similar client, one honest "should we close the loop?"

The sunset logic is identical. CRM hygiene is just list hygiene wearing a different hat.

6. What to offer (and when a discount is genuinely the right call)

Default to no discount in message 1. You don't yet know if the customer needed one. About half of recoveries happen on attention alone, and every one of those is a customer you didn't have to discount to keep.

Discount in message 2 when three conditions are true: your margin can absorb it without making the recovered customer unprofitable, the product category has low natural repurchase pull (commodities, replenishables), and your competitors discount routinely so the customer expects the lever. If any of those is false, lead with value instead — a how-to, a new collection, a use-case they haven't seen.

For services and B2B, the equivalent of a discount is a no-commitment offer: an audit, a working session, a short diagnostic. Same logic — it should appear in message 2, not message 1, and only when the deal economics support it.

Message 3 carries no offer. It's the honest exit. Adding a discount there teaches customers to ignore you until you panic.

7. Measuring win-back: recovered revenue and deliverability lift

Most teams measure win-back on recovered revenue alone. That's half the picture and the smaller half.

Track four numbers monthly:

Recovered revenue. Orders from customers who engaged with the flow within 30 days of the final message, attributed to the flow. This is what most dashboards show by default.

Suppression rate. Percent of triggered customers who exit through the sunset branch. A healthy flow suppresses 50-70% of the people who enter it. If you're suppressing under 30%, your trigger is firing too early — you're mailing people who weren't actually at risk.

Complaint rate on the flow itself. Should stay well under Google's 0.1% target. If it climbs toward 0.3%, your trigger or your messages need work before the rest of your program pays the price.

Deliverability lift on the rest of your program. Compare inbox placement and open rates on broadcast campaigns in the 30 days before and after a sunset cycle. A working sunset typically lifts open rates 8-15% on the remaining list in the programs we've measured. That lift compounds across every campaign and every flow — which is why the sunset half of the win-back often produces more revenue than the recovery half, just not in a column anyone's tracking.

That second-order revenue is the case for treating win-back as a list-hygiene mechanism first and a recovery tactic second.

FAQ

How often should a win-back email flow run?
The flow itself is evergreen — it fires on a per-customer trigger, not a campaign schedule. Each customer enters once when their churn-risk score or expected-next-order date crosses your threshold. Audit the trigger threshold and message performance quarterly. Run your sunset suppression monthly so list decay doesn't compound between cycles.

What's the difference between a win-back flow and a sunset flow?
A win-back flow tries to recover lapsed customers with messages and offers. A sunset flow removes non-responders from your active sending list to protect deliverability. Most mature programs combine them: the win-back's final message is the entry point to the sunset branch, so recovery and hygiene run as one mechanism.

When should I add a discount to my re-engagement email sequence?
Not in message 1, and not in message 3. Message 1 tests whether attention alone recovers the customer; about half do. Message 3 is the honest exit and shouldn't carry an offer or it teaches customers to wait you out. Message 2 is the right place — but only if your margin and category support it.

Does sunsetting subscribers really improve deliverability?
Yes, and the effect is measurable within 30 days. Mailbox providers weight engagement rates heavily when deciding inbox placement. Removing the unengaged raises your aggregate open and click rates, which signals quality to providers like Gmail and Yahoo. Google's bulk sender rules make this less optional than it used to be — the 0.3% complaint ceiling is enforced.

What if my store doesn't have enough data for churn-risk prediction?
Klaviyo's churn-risk model activates at 500+ customers who've placed orders. Below that, use expected-next-order date instead: calculate each customer's average days between purchases, multiply by roughly 1.5, and trigger the flow when that date passes with no order. It's less precise than a probability model but materially better than a fixed 90-day window.

Ship this week

Pull a list of customers whose expected-next-order date passed in the last 30 days. Send them message 1 of the structure above — no discount, just attention. Watch what comes back. That's your baseline for everything else this article describes.

If you'd rather have the flow wired, measured, and sunsetting on its own, start a conversation.