Segmentation: stop sending one message to one list

Most pizza shops have one list and one message. Everyone gets the same Tuesday email at the same time with the same $3 off. That single blast is why open rates sag, why unsubscribes climb, and why the discount goes to the person who was going to order anyway. Segmentation is the fix, and it is not complicated. It is six cuts through a file you already own.

The six cuts that matter

A pizza customer file is richer than most operators realise. Every ticket carries a timestamp, a channel, an order value, an item list and, if online ordering or loyalty is switched on, an identity. Those five fields give you every segment worth having.

• Recency: days since last order. The single strongest predictor of whether someone orders again. A 14-day customer and a 140-day customer are different people and should never see the same message.
• Frequency: orders in the trailing twelve months. Separates the once-a-year birthday party from the Friday-night regular.
• Monetary: average ticket and trailing spend. A $58 family order and a $17 single are different economics and deserve different offers.
• Channel: delivery, carryout, dine-in, or third-party marketplace. Marketplace-only customers are effectively rented; converting them to first-party is a campaign in its own right.
• Daypart and day: the lunch crowd and the 10pm crowd are not the same audience and rarely respond to the same creative.
• Menu affinity: specialty versus cheese, wings attach, salad buyers, gluten-free, vegan. Affinity is what makes a message feel written for one person.

Build the grid, not the list

Do not build twenty ad-hoc lists. Build a grid: recency bands down one axis, frequency bands across. Nine to twelve cells is plenty. Every customer sits in exactly one cell on any given day, and every cell has one job.

The discipline is that a customer moves. A regular who slips from 0–21 days to 22–45 days has told you something before they have consciously decided anything. The grid catches drift while it is still cheap to fix: a nudge at day 30 costs a fraction of a win-back at day 120.

Segments that pay for themselves fastest

If you can only build three segments this month, build these.

1. One-and-done, last 30 days. A customer who ordered once and has not come back is the cheapest incremental order in the building. The second order roughly doubles the probability of a third.
2. Frequent, slipping. Someone who ordered nine times in ten months and nothing in six weeks. High value, clear signal, and a message that lands as attentiveness rather than desperation.
3. Marketplace-only. Customers who only ever come through a third-party app. Every one you convert to direct ordering is 15–30 points of margin you keep permanently.

What segmentation is not

It is not personalisation theatre. Putting a first name in a subject line is not segmentation and no customer has ever been moved by it. Segmentation is sending a different offer, at a different time, through a different channel, because the person is genuinely in a different situation.

It is also not an excuse to send more. A well-segmented programme usually sends fewer total messages than a blast programme, because each segment is contacted on its own clock rather than everyone being contacted on the shop’s clock.

The test of a segment is simple: if two segments would get the same message, they are one segment.

Keeping it honest

Segments decay. A grid built in March describes March. Recompute nightly, not quarterly, and hold a control group out of every segment so you can tell whether the segment is producing incremental orders or just taking credit for orders that were coming anyway.

Questions