What counts as a high churn rate for a cafe?

Churn only means something once you pin down the window. The workable definition for F&B is 90 days: take everyone who transacted at least once in a 90-day period, then check how many made no transaction in the following 90 days. Shorter windows overstate churn by counting normal gaps as departures; annual windows are too slow to act on.

The bands, for an independent cafe on that definition: under 30% is excellent and usually indicates a captive location such as an office lobby; 30–40% is good; 40–55% is where most independents sit; 55–65% is concerning; above 65% is structural — location, product consistency or pricing — and no messaging programme will fix it. Chains with strong programmes typically run five to ten points better than independents in the same category, mostly because they identify more customers and act on the data.

The trap in the number is the denominator. Churn is computed on identified customers, so if you only identify 8% of transactions, you are measuring the churn of the small group willing to hand over a phone number at the counter — a group that skews loyal. Venues frequently discover their 'improving' churn was an artefact of who was being captured. Raise identification to 40%+ and expect the number to get worse before it gets honest.

Split the number in two. New-customer churn — people who visited once or twice and vanished — is an onboarding and second-visit problem. Regular churn — people with an established rhythm who stopped — is a service, competition or attention problem. They have different causes and different fixes, and averaging them produces a number you cannot act on. In most cafes, new-customer churn is far higher, but regular churn is far more expensive per head.

Read churn against revenue. Flat revenue with rising churn is the most dangerous pattern in F&B, because acquisition is quietly funding the leak: you are paying to replace customers you already had. It looks stable on the P&L right up to the point where acquisition costs rise or a competitor opens, and then it does not.

Practical cadence: compute it monthly on a rolling basis, track the trend rather than the absolute, and treat a three-month directional move as signal. Single-month movements in a small venue are mostly noise from weather, holidays and school terms.