A stamp card rewards every guest the same — the one who comes in every week anyway, just like the one who is quietly starting to forget you. Predictive retention flips the question around: it spots who is about to slip away and gives them a timely reason to come back — before the guest disappears for good.
The difference is timing. Winning back a wavering regular is far cheaper than acquiring a lost one from scratch. Read the signs early, and you turn reaction into prevention.
The signals that matter
It’s not raw revenue that betrays churn, but the pattern. A guest who used to order every two weeks and has now gone quiet for six is a clearer warning sign than any point total. Watch for:
- the gap between visits stretching out,
- a shrinking variety in the basket (always just the same thing),
- a reliance on discounts (only comes in for deals),
- an open complaint that was never resolved.
Especially valuable is the group of once-good guests whose frequency is fading. They’re not gone yet — but right now they need a reason.
Ethics guests understand
Foresight must never tip into pressure. Explain, when asked, why an offer appeared, and skip the artificial urgency. And after a bad experience, hold a quiet period — someone who just had a cold delivery doesn’t want a points promotion, they want an apology. Loyalty runs on trust; every manipulative message costs more than it earns.
Always tied to operations
The smartest model becomes a problem when it triggers a rush promotion while the kitchen is already at its limit. Tie every outreach to capacity: offers stand down when staff or stock are tight. Otherwise the prediction produces exactly the bad reviews it was meant to prevent.
The 7 most common mistakes
- Steering by revenue alone instead of the visit pattern.
- Reacting too late — only once the guest is already gone.
- Blanket discounts instead of a fitting offer.
- Artificial urgency that reads as manipulation.
- No quiet period after a bad experience.
- Firing offers during a rush, when the kitchen has no room.
- Starting too big instead of with one audience and one success metric.
How to start — small and honest
Frequently asked questions
Do I need artificial intelligence for this?+
How do I tell that a guest is about to slip away?+
How do I keep predictions from feeling pushy?+
Why the link to kitchen capacity?+
Timing beats the fire hose
Predictive retention isn’t a magic trick, it’s attention to the pattern: it brings the right guests back at the right moment, with the right offer, and only when the kitchen can keep up. That’s cheaper than any new acquisition — and more respectful than any fire hose.
