Email and retention
Win-back flows: timing them to the product's real cycle
A win-back flow that fires at ninety days is guessing. Compute the reorder cycle from your own orders, define lapsed from it, and time the sequence to it.
By CartKernel · Published
A win-back flow fires when a customer is lapsed, and most stores define lapsed with a round number: sixty days, ninety, one hundred and twenty. The number should come from how the store’s own customers reorder. A coffee customer with no order in forty-five days is lapsed. A mattress customer at forty-five days is on schedule for the next decade. Here is how to compute the cycle from orders, set the trigger from it, and build a sequence that respects the difference.
Compute the cycle from your orders
Four steps, done in a spreadsheet or a notebook, repeated each quarter.
- Export a year or two of orders with customer identifier, order date and the category of the first item in each order.
- Keep customers with two or more orders and compute the days between each consecutive pair.
- Group by the category of the earlier order and take the median, the 75th percentile and the 90th percentile of the gaps.
- Treat the band between the 75th and 90th percentiles as the end of “usual”. A customer past it is lapsed for that category.
Say a store’s skincare buyers who do reorder come back at a median of fifty days with a 90th percentile of one hundred and ten. The first win-back touch belongs just before day one hundred and ten, not at ninety and not at a hundred and eighty. The figures are illustrative; the method is the point.
Two checks before trusting the result. If only a small share of first-time buyers ever place a second order, the store does not have a win-back problem, it has a second-purchase problem, and the post-purchase flow is the place to work. And if the gaps cluster around a season rather than a number of days, the trigger is a date, covered below.
Lapse is a property of the category, not the store
| Category | What sets the cycle | Where lapse starts |
|---|---|---|
| Consumables such as coffee, supplements, pet food | Usage rate and pack size | A little past the pack’s expected life |
| Skincare and haircare | Product size and routine | Past the 90th percentile of observed gaps |
| Apparel and footwear | Season and occasion | The next season with no order |
| Gifts and occasions | The calendar | The occasion’s anniversary with no order |
| Furniture, appliances, equipment | Years | No lapse in the usual sense; the flow is cross-category |
| Hobby and craft supplies | Project rhythm | Past the observed gap between project orders |
For seasonal and occasion categories the trigger is anchored to a date, such as the anniversary of the last order or the start of the buying season, rather than a count of days.
Who to leave out
Exclude active subscribers, customers whose last order ended in a return or refund, anyone who unsubscribed or marked a message as spam, anyone with an open support conversation, customers whose only purchase was a gift card, and buyers of a discontinued product, who need a different message about the replacement. A win-back sent to someone who returned their last order reads as if the store was not paying attention.
The sequence, keyed to the cycle
Touch one, at the lapse point. No discount. What has changed since their last order in the category they bought from: new products, an improvement to the one they had, a restock. A reminder of what they bought with a one-click path to reorder it. A useful piece of content if one exists.
Touch two, a fraction of the cycle later. The ask. A one-question survey: did not need it yet, found it elsewhere, price, a problem with the product, something else. Route each answer. A product problem goes to support that day. “Did not need it yet” extends the cycle for that customer and pauses the flow. Price moves to the offer branch. Found it elsewhere gets one honest message about what is different and then stops.
Touch three, only for the offer branch or for silence. Where store policy allows an incentive, tier it by the customer’s past value and give it an expiry. The tiering decision cuts both ways: a store can reward its best lapsed customers most, or reserve incentives for customers who have never bought at full price, and either is defensible if written down.
Touch four, then the sunset. A last message that says so. Customers who engage with nothing across the sequence move out of regular campaigns into a low-frequency segment, and after a further period of silence are suppressed. Continuing to mail people who never open is how spam rates cross the thresholds in Gmail’s sender guidelines, and Klaviyo emails going to spam describes what happens when they do.
SMS. One message at the offer stage where consent exists and the cycle is short enough for a text to make sense. Is SMS marketing worth it for ecommerce sets out when it is.
Hold some customers out
A share of lapsed customers come back on their own. Without a holdout, the flow claims them. Randomly withhold a portion of each month’s newly lapsed customers, send them nothing, and compare their reorder rate over the following sixty days with the recipients. The difference is what the flow added, and it is the only number that justifies the incentive spend. Incrementality explains the method, and how do you measure incrementality for ecommerce ads applies the same logic to paid media.
What to measure
- Reactivation rate by category segment, against the holdout.
- Revenue per recipient by touch, so the cost of the incentive can be compared with what touches one and two recovered without it.
- The share of reactivations that used a discount.
- Whether reactivated customers place a further order, which is the difference between winning a customer back and buying one order.
- Survey answer distribution, which is product and pricing feedback the rest of the business should see.
Recompute the cycle every quarter. Pack sizes change, subscriptions absorb the most regular buyers, and the median drifts.
When win-back is the wrong tool
Durable goods have no reorder cycle to time against. The right move there is a cross-category introduction timed to delivery and a referral ask, not a lapsed sequence. Stores whose customers mostly buy once have a product or post-purchase problem that a win-back cannot fix, and how do you increase repeat purchase rate is the better starting point. Everything else, from consumables to seasonal apparel, has a cycle worth measuring, and a flow timed to it is the difference between a message that arrives as a reminder and one that arrives as noise. The repeat purchase rate and customer lifetime value entries define what the flow is trying to move, and the email marketing service builds it from the order export up.