Win-Back

Win-back flows are a standard fixture in retail and e-commerce marketing, where every transaction carries an account and the last-purchase date is sitting right there in the platform. The restaurant version is harder for one reason: the event that matters is a visit, and most visits are recorded in a point-of-sale system that never learned the guest’s name. So a win-back in hospitality depends on a restaurant CRM having already joined checks to bookings and both to a profile. The message is the easy part.

How does a win-back work in a restaurant?

A win-back has four moving pieces: a definition of lapse, the guests who meet it, a message, and a way to tell whether it worked. It gets built in one of two shapes.

As a campaign, an operator pulls the lapsed segment and sends it something on a chosen date. This is quick, it suits a seasonal hook such as a menu change, and every guest receives it on the restaurant’s calendar rather than their own.

As a journey, the message fires per guest at the moment they cross the threshold. A guest who goes quiet in March gets the note in March, and a guest who goes quiet in October gets an identical note in October. The journey shape is what makes the outreach read as attentive, because the timing lines up with the guest’s own absence instead of with a marketing schedule. It also runs unattended, which matters more than it sounds: win-back is exactly the work that gets skipped in a busy quarter.

When should a win-back go out?

The right moment is a little after the guest’s normal interval has clearly passed, and finding it means looking at the venue’s own return-interval distribution rather than at a benchmark. A room where regulars come monthly and a room where they come twice a year are answering completely different questions with the same word.

Two mistakes are more common than getting the number slightly wrong. Sending too early catches guests who were never lapsed, and a note about how much you have missed somebody who ate there three weeks ago reads as either careless or automated. Sending too late means reintroducing the restaurant rather than continuing a relationship, and by then the guest has found somewhere else for that occasion. Seasonality deserves one check as well, since a guest who never comes in August is not lapsing in August.

Should a win-back include an offer?

A discount is the reflex, and it carries two costs beyond the margin. It teaches a segment of guests that going quiet earns a comp, and it converts the guests who were going to come back anyway, which inflates the apparent result. Three alternatives usually perform better with a restaurant’s better guests.

• A reason to return. A new menu, the start of a season, the return of a dish that guest ordered. Specific beats generous.
• An invitation. A seat at a counter dinner, a wine event, a table held on a night the room is quiet. Access reads as recognition rather than as a promotion.
• A note from a person. Short, signed by a name the guest saw on their last visit, asking nothing.

Whichever route is chosen, the only way to know what the outreach earned is to hold back a comparable group and compare their return rate against the group that received it. Without a holdout, a win-back report is measuring guests who were returning regardless. Anything sent by text also needs prior consent under US telemarketing rules, and consent to be texted about a reservation is not consent to be marketed to.

Can an email platform run a restaurant win-back?

Not on the signal that matters. Mailchimp and Klaviyo both ship win-back flows, and both key them on last purchase or last email engagement. In a restaurant, most spend never passes through either system, so “lapsed” inside an email tool usually means the guest stopped opening messages. Those are two unrelated populations. A guest can read every email and never book, or dine weekly and never open anything.

The result is a win-back aimed at inbox behaviour instead of at absence from the dining room, which puts the note in front of guests who were there on Tuesday. Fixing it requires a last-visit date derived from reservations and checks, and neither platform has a native connection to a reservation book. Bridging one is custom integration work, and the connectors that exist mostly import bookings as marketing contacts rather than reconciling them against what the table spent. The message quality was never the constraint. The trigger was.