When is a customer actually lapsed? Cadence beats calendar
Every win-back program stands on one definition: lapsed. Get it wrong in one direction and you pester loyal customers who were simply between orders; get it wrong in the other and you reach out to people who forgot your store exists. The flat-days rule most tools ship — “lapsed = no order in 90 days” — manages to make both mistakes at once.
Why the calendar lies
Ninety days of silence is three missed orders for a monthly coffee buyer and zero missed orders for a quarterly skincare customer. Same number, opposite meanings. A fixed window mails the skincare customer a worried “we miss you” while she’s mid-jar — and waits two extra months before noticing the coffee buyer left. Any definition that ignores how each customer actually shops is guessing.
Cadence: the customer’s own clock
The fix is to measure silence against the customer’s own rhythm — the typical gap between their orders, learned from their history. Then lapse becomes a multiple of that gap: a customer at 2–3× their usual interval has clearly broken pattern, whoever they are. The multiplier is your patience dial:
Lapse threshold calculator
That gives a customer who usually orders every 38 days a 57-day grace period before win-back begins — late enough that most would-be organic orders have happened, early enough that your store is still remembered. A fixed “90 days since last order” rule gets this wrong in both directions; measuring against each customer’s own rhythm is how Forwarding Address decides.
Choosing the multiplier
- ~2× their gap — attentive. Right for consumable-driven stores where a broken rhythm usually means the customer replaced you somewhere else and speed matters.
- ~2.5–3× their gap — patient. Right for discretionary catalogs where gaps naturally stretch, and for stores that would rather never message a customer who was coming back anyway.
- Whatever you choose, watch the honest number. A holdout comparison tells you whether your threshold is early enough to matter and late enough to be incremental — and it will happily contradict your intuition.
From detection to action
Crossing the threshold should start a sequence, not a blast: the escalation ladder — email, then email with substance, then a printed card for customers whose lifetime value earns it. And detection must run continuously: lapses happen every day, and each one is freshest — most winnable — the day it’s detected. This is precisely the kind of per-customer arithmetic that belongs to software rather than to a monthly spreadsheet afternoon.
Common questions
›What about a customer with only one order — what cadence do they have?
None yet, which is honest information: single-purchase customers get a store-level default window instead (your typical reorder gap is a good one). They are also the segment most likely to churn silently, so a slightly earlier, gentler first touch is often right for them.
›Why not just react when revenue drops?
Because aggregate revenue hides individual drift — new customers replace lapsed ones on the top line while the retention leak grows underneath. Lapse detection works per customer, which is the only level where a win-back message makes sense anyway.
›Can a lapse threshold be too early?
Yes, and it is the costlier mistake: message customers who were returning anyway and you pay in goodwill and discounts for orders you already had. The patience multiplier exists exactly for this — win-back should begin after the customer’s own rhythm is clearly broken, not the moment an order is fashionably late.