Guide
Not every customer is the same customer.
Four groups is enough. Sending everybody the same message wastes most of it.
In short
Split your list by how recently somebody ordered and how often. That gives four practical groups: regulars, occasionals, at-risk and lapsed. Each needs a different message, and the at-risk group is where a small intervention produces the most return, because they are still reachable and have not yet gone.
Updated 28 August 2026
Two dimensions are enough
Recency and frequency. How long since their last order, and how many orders they have placed. You can add spend later, but these two carry most of the value and both are available from your order history without any extra work.
The four groups
Each behaves differently and wants a different message. Treating them identically is what makes generic marketing feel generic.
- Regulars: recent and frequent. Protect. Do not discount them.
- Occasionals: recent, infrequent. Nudge to a second habit.
- At-risk: was frequent, gone quiet. Highest-value intervention.
- Lapsed: long gone. One or two attempts, then stop.
Why at-risk is where the money is
A regular does not need an offer, and giving one costs you margin on orders you already had. A lapsed customer often will not come back whatever you send. The at-risk group is the one where a message genuinely changes the outcome, and it is small enough that you can afford a decent incentive.
Do not discount your regulars
The most common segmentation error is sending the whole list the same offer. Your regulars redeem it happily, and you have paid margin for behaviour you already had. If an offer is going out, exclude the people who ordered in the last fortnight.
The honest limitation
Segmentation needs order history attached to customers, which means your own ordering channel. If most of your orders come through marketplaces, you cannot segment, because you do not have the data. That is a channel problem, not an analysis problem.
Questions, answered straight.
Not covered here? Just ask us.
How many segments should I have?
Four is plenty for a single restaurant. More segments mean more messages to write and diminishing differences between them.
What counts as at-risk?
Roughly twice their own normal gap between orders. It is a relative measure, not a fixed number of days, because a weekly customer and a monthly one drift at different speeds.
Should I segment by dish preference?
It is useful for one thing specifically: referencing somebody's usual in a win-back, which lifts response noticeably. Beyond that it is rarely worth the complexity.
Can I do this in a spreadsheet?
Yes, if you can export orders with a customer identifier and a date. It is a sort and two columns. The hard part is having the data, not the analysis.
See it run your restaurant.
Twenty minutes on your menu and your numbers. We'll show you what actually changes in the first month and exactly what it costs. If it's not right for your restaurant, we'll tell you that instead.
- Live in days, not months
- We build your menu and train your staff
- No contract, and thirty days to change your mind
Rather just ring us? +353 87 438 8032
Tell us about your restaurant
We reply the same day, usually inside a couple of hours.