How to build a cafe loyalty program that actually pays back

Most cafe loyalty programs are designed backwards. Operators start with the reward ('buy 9, get 1 free'), then bolt on a way to track it, then discover six months later that they cannot say whether any of it changed behaviour. Build it in the opposite order: decide what you will measure, then choose the enrolment mechanic, then set the earn rate last — because the earn rate is the only part you can safely tune afterwards.

Earn rate. The workable range for cafes is a 4–6% effective reward cost. At an average ticket of $4, 1 point per $0.40 with a free drink at 10 points means a customer spends around $40 to earn a $4 drink — 10% headline, but only about 4–5% real cost once you account for redemption rates below 100% and the food cost of the reward rather than its menu price. Anything above 8% effective cost and you are running a permanent discount with extra steps.

Enrolment. This is the part that decides whether the program works, and it is almost never the part operators worry about. A program that requires the cashier to ask for a phone number, type it in, and explain the mechanic adds 20–40 seconds to a transaction. During the morning rush that conversation simply does not happen, and your program quietly captures only the customers who were already loyal. Self-enrolment — a table or counter QR the customer scans on their own time, or receipt capture that lets them claim a visit after the fact — removes the cashier from the critical path entirely. Venues that switch typically see enrolment rates move from single digits to 40–60% of transactions.

Reward design. Two rewards with identical cost perform very differently. A named item ('a free flat white') converts 1.3–1.6× better than an equivalent percentage discount, because customers price the gift at menu value and the discount at cash value. Percentages also teach customers to wait for the next offer, which is exactly the habit you do not want to build.

Expiry and liability. Points are a liability on your books until they are redeemed or expire. A 6–12 month expiry, clearly disclosed at sign-up, keeps that liability bounded and gives you a legitimate reason to message lapsed members ('your 8 points expire in 14 days') — one of the highest-converting messages in any loyalty stack.

Measurement. Total points issued is a vanity number. The honest test is incremental lift: take members and non-members matched on first-visit week and ticket size, then compare 90-day revenue per customer. Thirty-day numbers are usually a sign-up sugar high. If the 90-day gap is under 8–10%, the mechanic — not the software — needs changing.

Where AI changes the shape of this: the manual version of a good program is roughly ten hours a month of exporting, segmenting and writing messages. An AI agent that watches each customer's own visit cadence, flags the ones drifting past their personal gap, and drafts the message in your venue's voice turns that into a review-and-approve task. That is the difference between a program that runs for three months and one that runs for three years.