Your best customers leave silently — how do you notice?

Subscription businesses get a cancellation event. Restaurants get nothing. A customer who came in every Tuesday for two years does not tell you they have stopped; there is simply a Tuesday, and then another one, and by the time the absence is obvious enough to notice, they have been eating somewhere else for six weeks.

The reason this stays invisible is that revenue covers it. A venue losing four regulars a month and gaining five new customers looks flat on the P&L and slightly positive on foot traffic. Underneath, the composition is deteriorating: high-frequency, high-margin regulars are being replaced by first-timers who mostly will not return, and the acquisition cost of that replacement is rising every year. The number stays the same right up until acquisition gets harder, and then it falls quickly.

Detection is straightforward in principle and impossible by hand. Every customer has an implicit rhythm — four days, ten days, three weeks. Departure is that rhythm breaking. What makes it hard is that the rhythm is different for every person, so a single threshold cannot work: at 30 days, the four-day commuter is long gone and the six-week brunch customer has done nothing wrong. Getting this right means computing a personal baseline for each customer and watching the deviation, which is a counting task no operator can perform for 800 people.

The window is short. In the first two weeks past a customer's normal gap, they usually have not replaced you — they have just had a disrupted fortnight. A relevant message in that period converts at a multiple of the same message sent two months later, when a new habit has formed around a competitor. Almost all the recoverable value sits in those two weeks, which is why detection speed matters more than message cleverness.

Two things follow. First, the intervention should be small: acknowledgement, a specific reference, no heavy discount. The customer has not decided against you; they have drifted. Second, this is the clearest case in F&B for automation, because the work is continuous monitoring of many individual baselines — exactly the kind of task that is trivial for software and impossible for a person running a venue.

The uncomfortable implication for anyone running a points programme: a ledger that counts points but does not watch rhythms cannot see silent churn at all. It will happily accumulate points for a customer who has not been seen in two months and report a healthy member base while the base quietly hollows out.