Win-Back Emails: Define Lapsed From Your Own Data, Then Send at the Right Time

You set up the win-back flow six months ago. The trigger is "no order in 90 days". It sends a "we miss you" email, maybe a discount, maybe two follow-ups. Clicks are thin, second orders barely move, and the automation runs on while you forget about it. The copy is rarely the problem. The timing is. A customer who buys coffee every 28 days is gone by day 60. A customer who buys a serum every 75 days is not lapsed at 90, they are on schedule. One flat rule sends the right message to the wrong people at the wrong time.

Key takeaways

• Replace the flat 90-day rule with a per-product threshold calculated from your own order history: the median reorder gap and the 90th percentile.

• A win-back email and a replenishment reminder are different messages. One fires when a customer is on time, the other when they are late. Mixing them into one trigger wastes both.

• Win back the recently late first-time buyers first, then late repeat customers, and only then the people who have been gone for a year.

• Two or three touches over two weeks, no discount on the first, a small incentive on the second, and a clear stop.

• Measure against a hold-out group and report reactivation by lapse band, not opens.

What a win-back email is, and how it differs from a replenishment reminder

A win-back email goes to a customer who is overdue for a purchase. Its job is to bring them back before they drift away for good.

The distinction that matters:

• A replenishment reminder is an on-time nudge inside the normal reorder window. A pet food customer who reorders every 40 days gets "you might be running low" around day 35. No discount, because the order is expected anyway. There is a full walkthrough in the replenishment email guide.

• A win-back email fires once the expected window has clearly passed. The same customer at day 55 with no reorder gets a message that acknowledges the gap. This one may carry an incentive, because something in their behaviour has changed.

Sending the same email to someone who is on time and someone who is late means discounting orders that would have arrived on their own.

Define "lapsed" from your order export, not from a guess

This is the practical core. You can do it from a CSV of orders with a spreadsheet.

What you need: one row per order or line item, with customer email or id, product SKU, order date, and order value.

Steps, per product:

1. Filter to customers who bought the same product more than once. You need real reorder gaps.

2. Sort by customer, then by order date. For each second and later order of that product, calculate the days since the previous order of the same product.

3. Take the median of those gaps. That is the typical reorder interval.

4. Take the 90th percentile. Only one in ten reorders happens later than this. A customer who passes it without reordering is behaving differently from ninety percent of your repeat buyers for that product.

5. Define lapsed for that product as: days since last order exceeds the 90th percentile, and no new order has arrived.

The threshold is per product, not per store. Coffee beans and a skincare kit should not share one.

Illustration, with round numbers: a coffee brand sells a 1 kg Brazil blend. Ten repeat customers bought it two or more times, with gaps of 25, 28, 29, 30, 31, 32, 34, 36, 40 and 55 days. The median is about 31 days. The 90th percentile is around 52 to 55 days. A customer is lapsed on this product after day 55 with no reorder.

That gives you the timing directly: replenishment reminder around day 25 to 30, first win-back shortly after day 55, second touch around day 62 to 65 if nothing has happened. Run the same exercise on a pet food SKU and a serum and you will get different thresholds inside the same store.

Who to win back first

Not all lapsed customers are worth the same effort. Order matters.

Recently late first-time buyers. One purchase, now 5 to 15 days past the 90th percentile for their product. This is the hardest conversion in the store, first order to second, and these people are closest to it. A plain reminder with the product image and a one-click reorder button often works without a discount. They have not lost interest. They needed a prompt.

Late repeat customers. Two or more orders, now 10 to 20 percent past their own usual gap. They know the product and the brand. A contextual reminder, a companion product, or a bundle can bring them back. If you run RFM segmentation, this is where the high-value At-Risk group lives, and they deserve the more personal version.

Long-gone customers. Twelve months or more, past every product's 90th percentile. They need stronger incentives and convert less often. Third priority, not first.

Useful segment labels for an agency deck: "first-time buyers 0 to 30 days late", "repeat buyers 0 to 45 days late", "hibernating 365+ days".

The sequence: timing, touches, incentives

Two or three touches over two weeks beats both a single send and a long drip.

Email 1, on the day the product crosses its 90th percentile. "We noticed you might be out." No discount. A specific button: "Reorder your 1 kg Brazil blend" beats "Shop now".

Email 2, five to seven days later, if no order. More direct. Remind them what they bought and when. A low-friction incentive such as free shipping, or a modest percentage off if margins allow.

Email 3, ten to fourteen days after the first, if still nothing. A last-chance frame with a time-limited offer, and a promise to send less if they do not respond. Keep the unsubscribe link visible.

Then stop. Move non-responders to a low-frequency list and try again in about three months. A six or eight email drip to lapsed customers damages sender reputation and raises complaints for very little return.

When not to discount: customers only a few days late, thin-margin fast-cadence products, and any case where a helpful reminder (brewing tips, a feeding guide) will do the work.

Five win-back emails, each tied to a data signal

1. "Out of beans already?" Trigger: first-time buyer, 5 to 10 days past the 90th percentile on a coffee blend. Body: their last purchase date, the exact product image, one-click reorder. No discount.

2. Pet food for a repeat customer. Trigger: multi-order buyer, 10 percent past their personal average gap on a dog food SKU. Subject: "Is Max running low on dinner?" Use the pet's name if you have it, show the usual bag size, and offer a small loyalty perk rather than a blanket code. If you add social proof, use a figure you can count from your own orders.

3. Skincare routine reset. Trigger: first-time serum buyer at day 70, where the median is 40 and the 90th percentile is 60. Subject: "Still using your serum daily?" Explain that most customers reorder every five to seven weeks, add a short note on consistency, and a "rebuild your routine" button with a modest offer.

4. One last hello. Trigger: no orders in 365 days across all products. Acknowledge the gap without drama, show what has changed since they last bought, make one stronger time-limited offer, and keep the unsubscribe link obvious.

5. The companion product. Trigger: customer late on a product that, in your own order data, usually leads to a second order of a specific companion item within 30 days. Subject: "Most people who loved your collagen also tried this." Show one or two products that actually follow it in your orders, not a catalogue grid. Market basket analysis across orders is how you find which product follows which.

Mistakes that quietly drain win-back campaigns

Flat 90-day rules. Fast-cadence products are already lost by day 60. Slow-cadence products are not late yet at 90. Per-product thresholds fix both.

Leading with a discount. A 20 percent code in email one trains lapsed and active customers alike to wait for the next coupon. Keep incentives for the second and third touch, and only for people who are clearly overdue.

Treating paused subscribers as lapsed. A subscriber who skipped a shipment decided to pause, not to leave. Send them a subscription check-in, not a win-back.

Never stopping. A monthly "we miss you" forever hurts deliverability for the customers who do buy. Cap it at two or three attempts, then suppress.

Measure it honestly

Define reactivated: a customer who was in a lapsed segment and ordered within 14 to 30 days after the last email in the sequence.

Track by lapse band: 0 to 30 days late, 31 to 90, 91 to 365, over 365. Reactivation falls as lapse grows, and the table shows where the effort stops paying.

Hold out a group: withhold 10 to 20 percent of eligible lapsed customers from the sequence. Their natural reorder rate is the baseline. The difference is your real lift. Without it you are counting orders that would have happened anyway.

For agencies: report "customers reactivated by lapse band" and "incremental revenue versus hold-out" every month. Present it as recovered second orders and recovered repeat customers, because that is what the client is paying for.

Doing it yourself versus using a tool

A spreadsheet handles one product well. Export 12 to 24 months of orders, clean the customer ids, compute the gaps per customer per product, and take the median and 90th percentile for your top ten products by revenue.

A good result is a tight window: most coffee buyers reordering between 25 and 40 days. A bad result is a wide, noisy window, which usually means inconsistent supply, promotional buying, or a product bought as a gift. Wide windows are a sign the product is not a strong replenishment candidate.

It gets hard at scale: dozens of products, patterns across orders, and ranking lapsed customers across many SKUs at once. Affinsy computes the reorder cadence per product from the order history, identifies which products tend to earn the second order, and lists the customers who are late for each, as an exportable audience sized for a campaign.

Next steps

1. Pull the last 12 to 24 months of orders and pick your top five replenishable products by revenue.

2. Calculate the median and 90th percentile reorder gap for each.

3. Build two segments per product: recently late first-time buyers and recently late repeat customers.

4. Draft a two or three touch sequence per segment with product-specific subject lines and incentive timing, and set product-level triggers in your email platform.

5. Launch with a 10 to 20 percent hold-out and read reactivation by lapse band after 30 days.

Replacing the generic 90-day trigger with product-level thresholds is the single cheapest improvement most stores can make to win-back. If you want the thresholds and the late lists worked out for you, the 48-hour analysis returns them for every product in the catalogue.

FAQ

How often should I refresh the lapsed definition?

Every three to six months, and after a product launch, a pricing change or a season that shifts buying. Agencies can tie it to the quarterly review.

Same timing for subscribers and one-time buyers?

No. A missed subscription renewal should trigger a "subscription at risk" or failed-payment sequence, not the standard win-back. One-time buyers get the product-level thresholds above.

Can I run win-back without any discount?

Yes, and for customers who are only slightly late it usually works better. Test discount against no-discount on the same lapse band, and watch whether customers start delaying reorders to wait for the coupon.

What attribution window should I use?

Seven to fourteen days after the last email in the sequence. Long enough for a normal decision, short enough to avoid counting unrelated seasonal orders. Keep it constant so periods are comparable.

How many products belong in a win-back email?

One primary product tied to the last purchase, plus at most one or two companions that your order data says actually follow it. A crowded grid dilutes the single action you are asking for.