Spotting who is drifting away

A customer who leaves does not resign. There is no cancellation, no final visit marked as final, no moment where anyone notices. The Thursday couple simply come every three weeks instead of every two, then every six, then not at all — and because each individual absence is unremarkable, the loss is invisible right up until the year's takings are down and nobody can say why. This topic is about making that visible, which is the single most valuable thing customer data does for a restaurant.

⏳ Decide what "lost" means, in days

You cannot notice an absence until you have defined one. The definition has to be a number of days, it has to come from your own rhythm as topic 3 described, and it has to be written down — because the whole value lies in applying the same rule every month and watching the list change.

Work it out from the customers you already have. Take your genuine regulars, look at the usual gap between their visits, and set your warning threshold at roughly two to three times that gap. A place whose regulars come weekly might warn at a month; a special-occasion restaurant whose best customers come twice a year cannot sensibly warn before eight or nine months. Then use two lines rather than one, because they call for different responses.

Almost all the value is in the first line. A customer at twice their usual gap has probably not decided anything — they have been busy, or something small went wrong, and a friendly nudge is often enough. By the time someone is three or four gaps out, they have a new habit somewhere else, and habits are much more expensive to break than to maintain.

🔍 Run the list, then look for the pattern

Once a month, produce two lists: who crossed into quiet, and who crossed into gone. Ten minutes of work, and it is the most useful ten minutes in this course. But before acting on individuals, read the lists as a group, because the pattern in them is often worth more than the customers themselves.

• Are the numbers growing month over month? A stable trickle is normal in any restaurant. A rising count is a problem in the business, not with those people.
• Do they share a date? If a cluster of good customers all went quiet after March, something changed in March — a menu, a price, a chef, a shift pattern.
• Do they share a day or a service? Losing your Sunday brunch people and nobody else points straight at Sunday brunch.
• Are they your best ones? Losing a hundred one-timers is weather. Losing five regulars is a fire.
• Did any of them tell you why? Cross-check against the complaints and reviews from the last course; that is where the reason usually is.

That last point is the join between the two courses and the reason they sit next to each other. The numbers can show you exactly who left and exactly when, and they will never tell you why. The words can tell you why and will never tell you how many. Put a departure date next to a complaint from the same week and you frequently get the whole story in one line.

✉️ What to actually send them

Having the list is most of the work; the message is simpler than people expect, and the biggest mistake is making it too commercial. Someone who has drifted has not rejected you, so a warm, human note that acknowledges the gap without guilt-tripping outperforms a discount almost every time — and the email course already covered how to write and send it.

A few things that matter specifically for this list. Do not lead with money: opening with a discount tells a customer who might have simply been busy that visits from them are worth buying, and it trains people to wait for offers. Do not tell them how long it has been in a way that reveals your counting — "we miss you" is warm, "you have not visited in 94 days" is a receipt. Give a reason to return that is about the restaurant rather than the price: something new, something back on the menu, a season that suits them. And send it once, not four times, because a drifting customer who gets pestered stops being a drifting customer and becomes an unsubscribe.

Then measure it honestly, which almost nobody does. Mark who was on the list, and check next month how many came back. That number — not opens, not clicks — is the only one that says whether any of this works, and it is also the number that tells you whether your quiet threshold is set right. If almost everyone you nudge comes back, your line is too tight and you are chasing people who were never leaving. If almost nobody does, it is too loose and you are finding them after the habit has already gone.

🛡️ The cheaper half: not losing them

Win-back is the visible half of retention and the smaller one. The bigger half is that most of the reasons customers drift are ordinary and preventable, and your own data will point at them if you ask it a different question: not "who left" but "what happens just before people leave."

Look at the last visits of your recent departures. Often there is something there — a long wait, a comped dish, a complaint, a first visit to a service that is weaker than your main one, a table that always gets refused. That is the thing to fix, and fixing it protects everyone who has not left yet, which is a far better return than any win-back campaign. It is also where this course and the previous one become one piece of work: the numbers find which visits went wrong, and the words tell you what went wrong in them.

And do the arithmetic once, because it settles priorities for good. If a regular is worth eight times a one-time visitor over a year, then keeping one regular is worth eight new strangers — and keeping them usually costs a phone call, a remade dish, or a manager noticing something on a Tuesday. Nothing else in your marketing has that return.