In brief
For a cafe, stamps beat points because the ticket is small and uniform, so points add arithmetic without adding fairness. The exception is a wide menu, and there is a number that tells you when you have one.
The short answer
Use stamps. A cafe sells a small number of similar things at similar prices, so points add a division the customer has to do without making the program any fairer. Stamps are legible from across the counter, complete in a timeframe a daily customer can picture, and cost you a fixed, knowable amount per completion.
The exception is a menu with a wide price spread, and there is a specific ratio below that tells you whether you have one.
For the general comparison across all business types, stamp cards versus points is the reference page. This is the cafe case, worked.
What points are actually for
Points exist to make a program fair when customers spend very different amounts. A customer buying £40 of groceries and one buying £4 should not earn the same, so you count pounds instead of visits.
That fairness has a price: the customer now has to convert. “You have 340 points” means nothing until they divide by something. Every points program is asking its customers to do arithmetic in exchange for fairness they may not have needed.
A cafe usually did not need it.
The spread test
One number decides it. Take the price of the cheapest thing a regular customer buys and the most expensive thing a regular customer buys. Divide.
- Espresso £2.20, flat white £3.40. Ratio 1.5. The spread is narrow. Stamps are fair enough, and the unfairness, which is the espresso drinker earning the same reward as the flat white drinker, is small and arguably the right way round: they came in just as often, and frequency is what a cafe is buying.
- Espresso £2.20, brunch £9.00. Ratio 4.1. Now a stamp card either overpays the espresso drinker or underpays the brunch customer, and both notice.
Under about 2, use stamps. Over about 3, points start to earn their complexity. Between the two, use stamps anyway, because the arithmetic cost is real and the fairness gain is small.
Most cafes selling drinks and a pastry sit under 2. Cafes that are really all-day restaurants sit over 3, and should read restaurant loyalty programs instead.
The discount you are actually giving
This is the arithmetic most people get wrong, and it is worth getting right before you set the threshold.
“Buy ten, get the eleventh free” means the customer pays for ten and walks out with eleven. One free in eleven received is 1 ÷ 11, or 9.1%. Not 10%.
The card that gives exactly ten per cent is the shorter one: pay for nine, receive ten, one free in ten.
Both numbers can describe the same ten-stamp card, because they use different denominators, so say which you mean. Against the eleven drinks received at menu price it is 9.1%. Against the £34.00 the customer handed over it is 10.0%. Neither is what the card costs you.
Here is the whole thing on the flat white from the spread test above, at £3.40 with £0.85 of beans, milk and cup in it:
| Card | Paid drinks | Customer pays | Discount they see | Your cost of goods | Share of your takings |
|---|---|---|---|---|---|
| Nine paid, tenth free | 9 | £30.60 | 10.0% | £0.85 | 2.8% |
| Ten paid, eleventh free | 10 | £34.00 | 9.1% | £0.85 | 2.5% |
| Twelve paid, thirteenth free | 12 | £40.80 | 7.7% | £0.85 | 2.1% |
Read the last two columns together. The customer perceives a discount around nine per cent; your books lose about two and a half. That gap is why loyalty rewards are affordable at all, and the full budget for it, with software and staff time on the same page, is in what a loyalty program actually costs.
Use the discount column when deciding whether the offer is worth putting on a poster. Use the last column when deciding whether you can afford it. Lengthening the card moves both, and points move neither: the reward is one flat white and one flat white is £0.85 whichever mechanic issued it.
What each one costs the customer to understand
| Stamps | Points | |
|---|---|---|
| Question the customer asks | “How many more?” | “How many is that worth?” |
| Answer visible at a glance | Yes | Only if the app converts it for them |
| Staff can explain in one sentence | Yes | Rarely |
| Fair across a wide price spread | No | Yes |
| Reward cost to you | Fixed and knowable | Varies with what they redeem against |
| Breakage | Cards abandoned mid-way | Balances quietly expiring |
The last row matters more than it looks. A stamp card’s cost is bounded at one free coffee per completion, and you know what a coffee costs. A points balance is a liability whose cost depends on what the customer eventually chooses to spend it on, and they will choose the thing with the best value to them, which is the thing with the worst margin for you.
Setting the cafe threshold
Stamps decided, the only remaining question is how many.
A daily customer completes a ten-stamp card in two working weeks. That is fast, arguably too fast, because you are giving away nine per cent to someone who was already coming every day.
A twice-weekly customer completes it in five weeks. That is close to ideal: near enough to stay interesting, far enough that the visits are worth something.
A weekly customer takes ten weeks, which is the outer edge of what people hold onto.
So the threshold follows your actual customer, not a convention:
| Typical customer | Stamps | Time to reward |
|---|---|---|
| Daily commuter | 12 | About two and a half weeks |
| Two or three times a week | 10 | Four to five weeks |
| Weekly | 8 | Two months |
| Less than weekly | Stamps are the wrong tool | Longer than anyone holds a card |
When a cafe should use points after all
Three cases, and only three:
The menu genuinely spans a lot. Ratio over 3 by the test above.
Retail alongside the counter. A cafe selling bags of beans at £14 next to £3 drinks has two businesses on one till, and a visit-based card treats a bag of beans as one coffee.
A till that already does it. If your point of sale runs a points scheme natively, at no extra cost, and staff already use it, the marginal cost of points is close to zero. Fairness for free is worth taking. It is still worth asking whether the customer can see their balance without asking someone. How loyalty points work covers what running a rate involves once you own one.
Outside those three, stamps, and the simplicity is the point rather than a compromise. A stamp is one action at the counter, which is the only thing that survives a morning rush; the rest of the counter rules that survive one are on the cafes and restaurants page.
Do this before you decide
Sort last week’s till report by transaction value and read the top line and the bottom line. That is the spread test with your own numbers in it, and it takes about a minute.
Under 2, set a card at the threshold your typical customer’s frequency earns from the table above, scope the reward to a category rather than an item, and stop reading about mechanics. Over 3, try the scoped reward first, because narrowing “anything on the menu” to “any hot drink” costs one word and caps your exposure at the price of a flat white. Buy points only if that fails and your till can already pass the basket total without anyone typing it.
Figures in this post are a worked model with the prices stated where they are used, not a survey. The espresso and the brunch are placeholders for whatever the cheapest and dearest things on your own board actually are.