Win-back is the campaign type most likely to be run badly, because it is emotionally satisfying to send and hard to measure honestly. A large share of the customers who respond to a win-back were going to return anyway. The job is to build a ladder that finds the ones who were not, without paying full price for the ones who were.
Value escalates. Cost per contact descends. That combination is deliberate: it concentrates spend on the customers who prove hardest to recover, while keeping the cheap channels doing the early work.
| Rung | Timing | Channel | Offer | Framing |
|---|---|---|---|---|
| 1 | At lapse threshold | None or small attach ("free garlic bread") | "Here is what is new": a reason, not a bribe | |
| 2 | +10–14 days | Email + SMS | Meaningful deferred value | "We saved you something": specific, time-boxed |
| 3 | +21–30 days | SMS, or mail for high-value | Strongest offer in the programme | "Last one from us": honest finality |
| Exit | +45 days | : | : | Drop cadence hard. Move to dormant. |
This is the counterintuitive part and the part operators skip. A meaningful share of lapsed customers have not made a decision to leave. They have drifted, changed a routine, or simply forgotten. For those people a reminder is sufficient, and giving them a discount is pure margin donation.
Rung one is a reason: a new item, a changed delivery time, a seasonal special, a genuinely useful piece of news. Only when a customer ignores a reason do they become worth paying for.
A flat "20% off, come back" applies the same cost to a customer worth $80 a year and a customer worth $900 a year. Scale the value to trailing spend, capped at a defensible fraction of expected recovered value.
A workable rule: the maximum total win-back spend on a customer should not exceed roughly 15–20% of the gross margin you would expect to recover over the following twelve months if the win-back succeeds. That produces meaningfully bigger offers for lost regulars, and small or no offers for one-time customers, which is correct.
The most expensive win-back offer should go to the customer you can least afford to have lost, not to everyone.
Win-back response decays steeply with time since last order. Past a certain point, usually somewhere between nine and eighteen months depending on the market, the response rate stops being distinguishable from the background rate at which lapsed customers spontaneously return.
Find your own knee by plotting response rate against days since last order, and set the exit point just past it. Continuing to mail beyond that point does two kinds of damage: it wastes budget, and on email it wastes deliverability, because a large unengaged sending population drags inbox placement down for the customers who are still listening.
Of all campaign types, win-backs are the most likely to take credit for other people’s work. Hold out 10% of every lapsed cohort (permanently, not per-campaign) and compare return rates.
Expect the honest incremental number to be a fraction of the gross number. That is not a failure. A win-back programme with 22% gross return and 9 points of incremental lift is a genuinely good programme, and knowing the 9 is what lets you spend confidently.
Gross response of 8–25% across a full three-rung ladder is typical, depending heavily on how the lapse threshold is set. Incremental lift over a holdout is the number that matters and is usually between a third and a half of the gross figure.
A single optional question in rung two costs nothing and occasionally surfaces something operationally important: a bad delivery, a price change, a driver problem. Do not make it a survey, and do not gate the offer behind answering it.
For high-value lapsed customers, yes. It is one of the few cases where the cost per contact is justified, because the recoverable value is large and mail reaches people whose email you may have lost. For one-order customers it never pays.
Evaluate eligibility at send time, not at build time, and exit on conversion immediately. This sounds obvious and is the most common failure in restaurant win-back programmes: a list built on Monday and sent on Thursday will contain people who ordered on Tuesday.
Keep reading
How to set a lapse threshold per customer rather than per shop, and why a fixed 60-day rule sends win-backs to people who ordered last week.
Campaigns & offersWhy issuing value forward, redeemable on the next order rather than this one, protects margin today, buys a second occasion, and self-selects the customers worth paying for.
Campaigns & offersThe seven automated journeys every pizza location should have running (welcome, second order, habit, drift, win-back, birthday and post-catering) and the order to build them in.
MeasurementBuilding holdout groups into every campaign, reading the difference honestly, and accepting that most reported marketing revenue was going to happen anyway.
marketing.pizza runs all of it, every night, across every store you have.