RFM is forty years old, came out of catalogue retail, and remains the highest ratio of insight to effort in restaurant marketing. Three numbers per customer. No model to train, no data science hire, no black box. The catch is that the textbook cut-offs were written for people who buy a sofa once a decade, and pizza does not work like that.
Standard RFM splits each dimension into quintiles: five equal buckets of customers. That is fine when purchase intervals are measured in years. Pizza has a natural weekly-to-monthly rhythm, and quintiles will happily put a 12-day customer and a 40-day customer in the same bucket while splitting hairs between day 3 and day 5.
Use fixed, behaviour-anchored bands instead of quintiles. The bands should map to the actual decision rhythm of the category: this week, this month, this quarter, this year, gone.
These are the cut-offs we default to. Adjust them if your shop has an unusual rhythm (a campus store or a stadium-adjacent store will run tighter, a rural store wider) but start here.
| Score | Recency (days) | Frequency (12mo orders) | Monetary (avg ticket) |
|---|---|---|---|
| 5 | 0–14 | 20+ | $45+ |
| 4 | 15–30 | 10–19 | $34–44 |
| 3 | 31–60 | 5–9 | $26–33 |
| 2 | 61–120 | 2–4 | $19–25 |
| 1 | 121+ | 1 | under $19 |
A three-digit score like 5-5-4 is precise but unusable in a marketing calendar. Collapse the 125 combinations into eight named groups that a shop manager can hold in their head.
Average ticket is the least useful of the three and the one operators fixate on. A $60 average ticket that appears twice a year is worth less than a $24 ticket that appears weekly. Weight frequency above monetary in every decision unless you are specifically building a catering programme.
If you want one number instead of three, use trailing-twelve-month spend and let frequency and ticket resolve themselves. But keep recency separate and visible. It is the only one of the three that tells you about the future rather than the past.
Recency predicts. Frequency confirms. Monetary sizes the prize. In that order.
Scores are recomputed nightly against a rolling twelve-month window, not a calendar year. A calendar-year window makes every customer look worse every January and produces a fake win-back surge that wastes a month of budget.
Count them, but tag them. A customer with eight marketplace orders and zero direct orders has a real habit, just not with your ordering channel. They belong in a conversion campaign, not a win-back, and the offer should be about ordering direct rather than about ordering at all.
They sit outside RFM entirely. Put them in a separate prospect track with its own short journey. Folding them into R1/F1 pollutes your lapsed segments and inflates your list size in a way that flatters the reporting and helps nobody.
Yes, and arguably better. A single store has one trade area, one menu and one rhythm, so the bands are cleaner. What changes at multi-unit scale is that you must score within each store, because a 30-day gap means something different in a college town than in a suburb.
Keep reading
How to cut a pizza customer file into segments that actually change behaviour: recency, frequency, ticket, channel, daypart and menu affinity.
Audience & dataHow 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.
MeasurementReading retention curves by acquisition month, spotting the second-order cliff, and why total revenue hides everything that matters.
Loyalty & retentionPoints, punch cards, tiers and surprise rewards, which structure suits which shop, and how to avoid paying regulars for behaviour they were already giving you.
marketing.pizza runs all of it, every night, across every store you have.