Win-back flows: 12 patterns that earn their place

The most common way a win-back program goes wrong is the trigger itself — the rule that decides who lands in the sequence. Teams reach for hasn't opened an email in 90 days because it's easy to build. It's also a garbage signal. Apple Mail Privacy Protection — the iOS feature that pre-fetches images so every email looks "opened" whether it was or not — inflates opens for over half the market. And 90 days is a round number someone pulled out of a deck in 2014. Use a real behavioural signal or don't bother.

For a subscription product, lapse is no login in N days. For a marketplace, no purchase plus no session start. For a content product, no consumption event plus no app open. Pick what reflects actual product value. Then tune N to the natural cadence of the thing — 14 days for food delivery, 9 months for annual travel booking. A round number is not a cadence.

One sequence for everyone is how you end up sending a cheery please come back to someone who cancelled last week over a billing dispute. There are at least three populations inside the lapsed list, and tiering them — splitting the audience into bands of likely-to-return — is the whole game.

Recently lapsed, historically high engagement is the top tier. Most likely to return. Light touch, high-value reminder, short sequence of two to three messages across one to two weeks.
For the medium-term lapsed, moderate historical engagementband you need a real reason to reach out: a product update, a new feature, something that's changed. Slower sequence — three to five messages across four to six weeks.
The bottom tier — long-lapsed or cancelled — is usually the largest audience and the lowest reactivation rate. Most of these users should end up suppressed (quietly removed from marketing sends), and the win-back attempt is a final qualification before that, not a real conversion play.

Sequence length scales with tier. Anything more than five messages per tier is counterproductive — you're burning deliverability (your ability to land in the inbox rather than spam) on people who aren't coming back and calling it optimism.

The Orbit Win-Back Playbook skillgenerates the tier definitions, the sequence per tier, and the sunset policy — tuned to your lapse definition and your product's natural cadence.

Each one with the audience it suits, the product type it fits, and the failure mode it has when you push it past its lane. Use them as a menu, not a checklist — most programs only need three or four.

The "here's what you missed" digest.Three to five things: new features, popular content, community milestones. Works for content-heavy and community products. Skip it for transactional products — you'll sound out of touch with a shopping cart.

The "what changed since you left" update. Single biggest change since the user lapsed. Lands best when the change addresses a common complaint. Lead with the change. The ask to return is a footer, not a headline.

Personalised usage memory. Their last listen, their favourite item back in stock, how long since their last workout. Requires a real data model — the generic version of this reads as mildly surveillance-adjacent, not personal. More on where this tips over in the personalisation guide.

Low-friction re-entry.One-click resume, one-click restart, skip the password reset. The finding that surprises most teams: removing password friction alone can lift top-tier reactivation two to three times over. Users don't always leave because they lost interest. Sometimes they left because they couldn't be bothered to log in.

So far: content reminder, change-driven update, personalisation, friction removal. The next group is the incentive-and-research bracket — where most teams reach first, and where most of the failure modes live.

Direct discount.The default move for most teams. Works for the middle and bottom tiers, and the level matters more than the offer. 10% won't move anyone who stopped caring. 30% signals desperation and teaches your base to wait out campaigns. 15–25%, tied to a specific product or category, is the zone that works without training bad behaviour.

The "what went wrong" survey.A single open-ended question, nothing else. Reactivation conversion is low. The value is elsewhere — as product research, it's some of the most honest feedback you'll ever get. Slot it late in the bottom-tier sequence. The users who respond are doing you a favour.

Behaviour-based trigger. Not a campaign — an automation. If a lapsed user does anything at all (opens, clicks, visits a page), fire a targeted flow the same day. Intent-in-the-moment is a much stronger signal than the lapse itself, and this is the pattern most teams forget to build.

Peer-behaviour nudge. Users like you are now doing X. Social proof for reactivation. Needs a real user model behind it. Popular with other users as generic filler reads as weak and vaguely manipulative.

The final cluster is the relationship-preservation moves — the ones you reach for when you've accepted that not every user needs to come back as the user they were before.

The "we'll miss you" note. Direct from the team, signed by a real person, brief, no discount. Usually the last message before sunset. Works because it breaks the pattern of every other marketing email the user has received, and because the subtext — this might be the last one — is true.

Preference-change invitation. Too many emails? Tell us what you'd prefer. Preserves the subscriber relationship without forcing reactivation. A subscriber who opts down in frequency is worth considerably more to your sender reputation than one who marks you as spam because you kept sending anyway.

Seasonal tie-in. Align the moment with a natural re-evaluation point: new year, quarter end, seasonal product relevance, annual usage anniversary. Higher inherent relevance. Less forced than a cold where have you been.

The channel-switch.Move lapsed email subscribers to a lighter-weight channel — push, SMS, in-app only — instead of forcing more email they've stopped opening. Lower complaint rate (the share of recipients who hit "mark as spam"), preserved relationship, often re-engages users who were tired of the email cadence specifically. The constraint, always, is consent: you can't port an email subscriber to SMS without a separate SMS opt-in, and that's the end of the conversation.

A sunset policy is the rule that says: this user is done, stop sending them marketing. Without one, the lapsed list grows indefinitely and your sender reputation drags down with it, because mailbox providers weight engagement-per-send (opens, clicks, replies divided by emails sent) and your engagement-per-send is bleeding.

A policy that works. After the win-back sequence completes, if the user takes no engagement action — open, click, login, purchase, session start — remove them from marketing. Transactional mail (receipts, password resets, the stuff they actually need) still flows. A final goodbye email gives them one link back, and re-opt-in is always available via a product surface. That's it. Not complicated. Just written down.

The number to watch is list churn — the percentage of subscribers you sunset each month. Healthy programs sunset 1–3% of the list each month. Programs that sunset nothing end up with lists that are 40% dead weight inside 18 months. The deliverability guide covers how list hygiene compounds into sender reputation, and the Deliverability Management skill runs the full diagnostic when you want to check if yours is already bleeding.

The headline metric is reactivation rate: share of the sequence audience that took a defined reactivation action inside the measurement window. Define the action specifically. Completed a purchase is not the same as opened any email. Measure both if you want, but the purchase number is the one that decides whether the program earns its seat.

Watch three secondaries. Complaint rate — should stay under 0.1% even though this audience is ambivalent and will naturally push higher than a warm cohort. Unsubscribe rate — will run higher than your baseline, and that's healthy. It's the sunset working in the open. And downstream LTV (lifetime value — total revenue per user over their relationship) of reactivated users — do they stick for 30+ days, or do they churn again inside a fortnight? Reactivation that doesn't retain is a vanity metric with extra steps.

The trap worth naming: measuring the open or click rate of the sequence itself and declaring the program broken when the numbers look rough. Opens on this audience will sit at 30–40% of your normal baseline. They're supposed to. Use reactivation as the headline and judge the program on that, not on sequence engagement stripped of context.