The free tenth coffee is the least interesting thing a café can do with a loyalty budget. It is not wrong. It is just the one mechanic that treats every guest identically, which is the one thing no café would ever do in person.
What follows is a list of mechanics you could start running on Monday. For each: what it is good at, what it costs, and the specific way it fails. Almost all of them need something a stamp card cannot give you, which is knowing who is at the counter.
Streaks instead of counts
Reward consecutive periods rather than cumulative visits. Three weeks running, the fourth opens with one on the house. Or tighter: a second visit inside seven days unlocks something small.
Good at: a streak has a clock and a stamp card does not. Ten stamps can sit at four for a year. It also gives you a reason to message a guest that is not a discount, and "you are one visit off" is information rather than an offer.
Costs: you have to see the gap between visits, per guest. And you will pay for frequency you already had, because the guests likeliest to hold a streak are already in rhythm.
Backfires: streaks punish normal life. A guest goes away for a fortnight, returns to a reset counter, and reads it as a fine for taking a holiday. Build in a grace week or a single skip, and keep the streak short enough to picture. Four is a plan; twelve is a chore.
Move the reward to the hour that needs it
The reward is only live between 2pm and 4pm, or before 8am, or on the day you are always dead.
Good at: this is the only idea here that changes the shape of your week rather than its size. The rent is the same at 3pm and the barista is already standing there.
Costs: the discount is real and the margin on those covers is no better, so you are buying traffic rather than selling it. That can be correct, but only if the traffic is new.
Backfires: cannibalisation. If the guest who was going to pay full price at 8:30 now waits until 2pm, you have paid to move a sale you already had. It works when the off-peak visit is an extra visit, not a rescheduled one. The test: watch whether peak covers hold while the trough fills. If peak drops, stop.
Reward spend, not visits
Points or tiers against dollars rather than trips through the door.
Good at: it stops ticket size deciding the reward. On a visit-count card the $4 espresso and the $30 brunch earn the same stamp, so the espresso gets the better deal while the brunch table pays your rent.
Costs: it is slower to feel. A stamp is legible at a glance; a points balance is arithmetic, and nobody does arithmetic at a counter. "You are $12 from your next reward" works. "You have 880 points" does not.
Backfires: a spend tier says we like you more when you spend more. That is true, and it is not always what a neighbourhood café wants said out loud. It also penalises the daily flat white regular, your most frequent guest and smallest ticket, who may be the person you most want to keep.
Reward the usual, not the visit
Build the mechanic around the thing a guest always orders. A prepaid bundle, ten flat whites bought once and drawn down over a month. The usual, ordered ahead, ready at 8:15. A bonus that fires on the item rather than the trip.
Good at: it turns a habit into a commitment. A prepaid bundle is a cash-flow event and a switching cost in one transaction. Order-ahead removes the queue, a common reason a regular skips a morning they meant to have.
Costs: you are taking money now for product later. That is a liability, and if you price the bundle at a discount it is also a discount. Both are fine; confusing them is not.
Backfires: two ways, the second easy to miss. It locks the guest onto one item, so your odds of them trying the seasonal drop. And it shrinks the basket, because someone drawing down a prepaid coffee is not reaching for a wallet, and the pastry was an impulse attached to paying. Watch attachment rate, not just redemption.
Two triggers you set once and leave alone
The date and the gap. Both fire without you touching them, which matters if you have no marketing hour in the week.
Birthday, and better, the first-visit anniversary. The birthday message is the only marketing a guest is pleased to receive. The anniversary is better, because every brand the guest ever gave an email to sends the birthday one in the same week. It costs you the date, which means asking, which is friction at signup: ask day and month, never the year. It backfires when codes get shared and reused, so tie the reward to the account rather than a code and give it a week to live, not a month.
The gap. Nobody cancels a café. They stop, and you find out in the monthly numbers or not at all. A trigger that fires when a weekly guest goes quiet for three weeks catches it while they still remember you. It needs per-guest rhythm rather than an average, because three weeks quiet means something different for a daily regular than a Saturday one. It backfires when the win-back is generous and predictable, because you have then taught your regulars that going quiet is worth money. Wording matters more than the offer: "we noticed you stopped coming" reads as surveillance to some people, while "it has been a while, here is one on us" does the same job without the flinch.
Referral, with the fraud priced in
An existing guest gets something when a new guest they sent completes a first purchase.
Good at: the incentive is only paid once it has already worked, which is not true of most advertising.
Costs: you pay both sides of one transaction, the referrer's reward and the newcomer's reason to try. That only makes sense if you have some idea what a new regular is worth across a few months. Without that number it is a coupon with extra steps.
Backfires: self-referral is the standard exploit, a second email and a second phone. Paying on the referred guest's first transaction rather than their signup handles most of it, because the fraud now has to buy a coffee. The subtler problem is that referral rewards flow to your most social guests, who are not necessarily your most valuable. And a referral scheme with no account behind it, no way to recognise the new guest when they walk in, is a business card with a promise on it.
What all of these actually require
Most of this list is impossible on paper, and the parts that are not are worse on paper. A stamp card cannot tell you a streak broke. It cannot tell a $4 order from a $30 one. It does not know it is anyone's birthday, and it has no idea the guest who came every Tuesday for a year stopped in March. An off-peak sign on a chalkboard needs none of this, but it cannot tell the guest you are moving to 3pm from the one who already comes at 3pm, so you discount both.
The prerequisite under all of it is the same and it is unglamorous: the transaction has to be attached to a person. That is the thing a stamp card cannot do, and it is the reason every café's loyalty program looks like every other café's. Once you can see who ordered what and when, pick two mechanics, one aimed at frequency and one at value, and run them a quarter before adding a third. Two a guest understands beat five they cannot keep track of.
Habitu is one way to get that layer: a café's own branded app running off the POS it already has, with the guest record behind it. The shipped POS integrations are Square and Foodics. The app ships in English and Arabic. Pricing runs $49 to $499 per location per month, plus per-order charges on the lower plans. Pretzel Australia runs its guest app on it.