Segmentation projects fail in a predictable way. Somebody opens the email tool, discovers it can build segments, and builds nineteen of them. Six months later there are twenty-three, four of them contradict each other, and every campaign still goes to All Contacts because nobody can remember which is which.
The fix is to start from the other end. Not "what can this tool filter on," but "which guests would I genuinely treat differently if I could." That question has a short answer, and the shortness is the point.
Four audiences. That is the entire recommendation. You may add a fifth once you have run the first four for a year and can say out loud what each one produced.
Why more segments make things worse
Every segment is a standing commitment. Somebody has to keep its definition true, remember it exists, and write something for it that they would not write for anyone else. A segment nobody writes for is not a segment, it is a saved filter with ambitions.
The real cost of over-segmenting is not confusion, it is silence. When choosing an audience takes twenty minutes of deliberation, the campaign quietly does not go out at all, and the reason given at the next meeting is that everyone has been busy.
The four audiences worth having
These four are chosen because each has an obvious and different action attached, and because you can define all of them from data you already hold. Nothing here requires a survey, a loyalty scheme, or a tag somebody has to remember to apply.
- Lapsed regulars. They visited often and have stopped. The highest-value list you will ever build, and the one most groups have never actually assembled.
- One-visit guests. They came once and never returned. Large, low-context, and worth a cheap message rather than a personal one.
- Occasion guests. People with a date attached — an anniversary, a birthday, the same week every year. The smallest list and the one that responds best.
- Active high spenders. Currently visiting, spending in your own upper band. This is not really a marketing list. It is a service list, and it should mostly be read by managers rather than mailed to.
Definitions you can put in a spreadsheet
Written as rules, because a segment that lives in someone's head gets redefined every time it is rebuilt, and then nobody can explain why the list shrank.
The thresholds below are placeholders. Set yours from your own visit distribution — four visits a year is devotion at one concept and indifference at another.
LAPSED REGULAR 4+ completed visits in the last 24 months AND no visit in the last 90 days ACTION a personal message from a named manager ONE-VISIT GUEST exactly 1 completed visit AND that visit was 3 to 18 months ago ACTION one short message about the restaurant, not about them OCCASION GUEST a date on record — anniversary, birthday, or the same week booked two years running ACTION ask 3 to 4 weeks before the date, never on it ACTIVE HIGH SPENDER visited in the last 90 days AND spend in your own top band ACTION brief the floor. Do not email this list.
What you actually send each one
Four audiences, four registers. The register is doing more work than the offer, and three of these four should not contain an offer at all.
- Lapsed regulars — something only you could know about them, from a named person, with no discount attached. Send this group a code and you have told them they were a transaction.
- One-visit guests — a reason that is about the restaurant rather than about them: a new menu, a season, a chef. Accept a lower response and do not pretend to remember someone you do not.
- Occasion guests — an early, low-pressure offer to hold something, three to four weeks out.
- Active high spenders — nothing by email. A note at the host stand, a recognised face, a table you kept back. Emailing this group is the most common way to make a good relationship feel automated.
The cadence that makes this survivable
Without a marketing team the constraint is not ideas, it is hours. One hour a month covers all four, and one hour a month is a thing a GM can actually defend on a calendar.
:00 Rebuild the four lists from last month's exports. :15 Read the lapsed regular list properly. Cross off anyone a manager knows has moved away, fallen out with us, or died. This step is not optional and it is why a human does it. :25 Write one lapsed-regular message. Send it from your own address, signed with your own name. :40 Send the one-visit message. Same copy as last month with the menu changed. :50 Print the occasion list for the next six weeks and hand it to the GMs. :60 Stop. The high spenders were handled at the briefing.
Where this stops being a spreadsheet
One restaurant, one reservation platform, one POS: all of the above is a spreadsheet and an afternoon, and you should do it that way until it hurts. Buying software to solve a problem you can solve with a sort and a filter is how operators end up with three tools and no lists.
It starts hurting when the definitions have to hold across systems that cannot see each other, and when the lists go stale within a week of being built. That is why most operators build them once, feel good about it, and never do it again.
Keeping those four lists continuously true across locations is the specific thing Compliments OS does. The writing stays yours, and honestly you will do it better than we would.