The win-back flow that recovers 12% of churned subscribers

What a win-back flow is actually for

Most lists carry a quiet problem. Somewhere between 40% and 60% of subscribers stop opening within six months of their last purchase, and they sit on the list bleeding deliverability reputation while contributing nothing. A win-back flow exists to do two things at once. Recover the buyers who can still be recovered, and cleanly retire the ones who cannot.

We run this as a Klaviyo win-back flow triggered on a lapse window, not a calendar broadcast. Across the accounts we manage, a well-built version brings back about 12% of the people who enter it. The other 88% give us a clean signal to suppress, which is almost as valuable because it pulls dead weight off the sending list.

The trigger and the lapse window

The entry trigger is the last purchase date crossing a threshold. We do not use last email open, because opens are noisy after Apple Mail Privacy Protection inflates them. We anchor on purchase behavior.

The threshold is category-dependent. For consumables with a 30 to 45 day repurchase cycle, we set the lapse window at twice the median repurchase interval. So a coffee brand with a 32-day median enters subscribers at day 64 with no second order. For considered purchases like furniture or electronics, the window stretches to 150 or 180 days. We pull the median interval from the customer's own order history in Klaviyo rather than guessing, because a generic 90-day window misfires badly on both ends of that range.

We also gate entry on a minimum lifetime value. Someone who bought once on a 70% discount and never returned is not worth a recovery offer. We require at least one full-price order or two orders total before the flow will accept a profile.

One thing we are strict about is the entry frequency. A profile can only enter the win-back flow once per long cooldown, usually a year. Without that cap, a customer who reactivates and lapses again can cycle through the flow repeatedly, getting trained that discounts arrive on a schedule. The cooldown keeps the offer feeling like a real reach-out instead of a recurring coupon drip.

Sequence timing

The flow is four emails over 18 days. The spacing matters more than the copy.

Email 1 goes out at entry. No discount. It is a re-engagement check that leads with what they bought and a reason to come back that is product-led. We see most of the 12% recovery concentrate here, before any money leaves the table.

Email 2 sends 4 days later if no order and no site session. This is the first soft incentive, usually free shipping or a low-value bundle, not a percentage off.

Email 3 sends 7 days after email 2. This carries the real offer, typically 15%, with a hard expiry of 5 days. We put the expiry in the subject line because urgency that is true performs and urgency that is fake trains people to ignore us.

Email 4 sends 4 days later as a last call, restating the same offer with the clock running out. We do not escalate the discount past email 3. Escalating teaches buyers to wait for the bigger number, and it poisons full-price behavior across the whole list.

The reason the spacing matters more than the copy is that a lapsed buyer's inbox is crowded and their relationship with the brand has cooled. Send too fast and you read as desperate, which raises complaints. Send too slow and the intent that the lapse trigger captured has gone cold by the time the offer lands. Eighteen days with widening gaps is the window we have settled on across dozens of accounts. It is long enough to feel considered and short enough to keep momentum.

Offer logic

The rule we hold to is that the discount climbs with proven value and stops early. A first-time buyer who lapsed gets a smaller offer than a three-time buyer, because the three-time buyer has shown they pay and they are worth more margin to reactivate.

We set offer tiers off lifetime value bands. Under one threshold, free shipping only. In the middle band, 15%. In the top band, 20% plus a personal-sounding note from a founder address. Every code is unique and single-use, generated through the ecommerce platform, never a shared public code that leaks to coupon sites and cannibalizes margin on active buyers.

Suppression rules that protect the domain

This is the part most teams skip, and it is the reason their sending reputation slides. A win-back flow that never suppresses anyone is just more mail to people who already told you no.

The non-negotiable exit: if a profile reaches the end of the flow with zero opens, zero clicks, and zero sessions across all four emails, we move it to a suppressed segment and stop sending campaigns to it entirely. Continuing to mail a profile that ignored four targeted attempts is how you land in spam folders for everyone else.

We also build conditional splits inside the flow. Any open or click routes the profile out of the discount track and into a lighter re-engagement path, because they are warming up and do not need a coupon. Any purchase exits immediately and adds them back to the active buyer segment.

One more guard. We exclude anyone with a recent customer-service ticket or a recent refund from the flow, so we never send a cheerful come-back-and-buy email to someone who is mid-complaint.

We also exclude anyone who is already active in another flow that is mid-conversation, so the win-back does not collide with, say, a post-purchase or replenishment sequence and double-mail the same person on the same day. In Klaviyo this is a flow filter checking recent message receipt, and it is the kind of thing that looks minor until you see a customer get three emails from the same brand in 24 hours and unsubscribe from all of it.

A concrete example

One skincare brand we work with had 22,000 lapsed buyers and a list that was slowly losing inbox placement. We built the flow with a 90-day lapse window pulled from their 44-day median repurchase cycle, four emails over 18 days, and a value-banded offer.

Over the first 60 days, 11.7% of entrants placed an order. Average order value on recovered buyers came in 8% below their historical average, which we expected given the discount, and acceptable because these were customers we had written off. More important for the long game, we suppressed 6,400 profiles who ignored the full sequence. Within three weeks, the brand's overall campaign open rate rose by 4 points and their spam-complaint rate dropped below 0.1%, because we stopped mailing people who were dragging the sender reputation down.

What we measure to know it is working

We judge a win-back flow on three numbers, not one. The recovery rate, the percentage of entrants who place an order, is the headline, and our bar is around 10% to 12%. But on its own it can be gamed by handing out a deep discount, so we pair it with two others. Recovered-order margin tells us whether we are buying back customers at a loss, and we hold the average recovered order to within roughly 10% of the brand's normal margin. Suppression rate, the percentage we move off the list at the end, tells us the hygiene half is doing its job, and on a stale list that is often 25% to 30% of entrants in the first pass.

We also watch the downstream effect on full-list metrics for the month after the flow goes live. If campaign open rates rise and complaint rates fall, the suppression is working as intended. If they do not move, the flow is recovering buyers but the list cleaning is too timid, and we tighten the exit rule. The flow is not a fire-and-forget asset. We review these numbers quarterly and adjust the lapse window, the offer bands, and the suppression threshold as the brand's repurchase behavior shifts.

How we think about it

A win-back flow is a revenue recovery tool and a list-hygiene tool wearing the same skin. The 12% you bring back pays for the work. The 88% you suppress is what keeps the rest of your program landing in the inbox. We build both halves on purpose, because a recovery flow that ignores the second half quietly costs more than it earns.

If you want the full sequence map and the suppression segment logic, we wrote it up at https://www.arthea.ai/article/winback-flow-sequence.