The 10-stamp loyalty card is the most common loyalty mechanic in fast-casual restaurants worldwide. It is also the worst-designed mechanic in fast-casual loyalty, which is two facts that ought to alarm anyone running one. The mechanic survives not because it works but because it is legible to the customer, easy to print, and never gets audited against the unit economics it actually produces. Audit it once and it becomes hard to keep running.
This post does the audit. Three frequency tiers, one mechanic, AED 45 average ticket, AED 29 gross margin per visit, AED 15 reward cost. The numbers are the operator's, not the platform vendor's. The question is what the punch card actually does to each cohort and which cohort the mechanic is pretending to incentivise versus which cohort it is actually paying.
The Pareto Concentration Most Operators Have Not Run
Loyalty redemption follows a Pareto distribution that is sharper than most operators expect. In a typical fast-casual guest base, the top 20% of guests by visit frequency account for 55-65% of all redemptions. The mechanic is structurally biased to reward the customers who needed the least convincing to come back, because they are the ones who hit the 10-stamp threshold fastest and most reliably.
That bias is not a bug at small scale. A neighbourhood café where the owner knows every regular by name does not need a loyalty programme to do customer relationship work, the punch card is a small thank-you gesture for guests the owner already values. At 5 locations, 30 staff, and 8,000 active guests, the same mechanic stops being a thank-you and starts being a structural inefficiency: the operator is paying a flat-rate discount to the cohort least likely to need it and ignoring the cohort where the same spend could have produced behaviour change.
The Three-Tier Worked Example
High-frequency cohort: 17 visits per month. A committed regular who hits 10 stamps every 18 days. Annual visits at this cadence: 204. Annual reward redemptions: roughly 11.3 (the guest claims a free item every 10 visits, with a fractional tail). Reward cost per year: AED 169.50. Annual revenue from this guest: 204 × AED 45 = AED 9,180. Effective margin hit per cycle: AED 29 of forgone gross margin plus AED 15 of reward COGS divided by AED 290 of cycle gross margin = roughly 5.3% of margin paid out as reward.
Mid-frequency cohort: 4 visits per month. The moderate regular who completes a card every 2.5 months. Annual visits: 48. Annual reward redemptions: 4.8. Reward cost per year: AED 72. Effective margin hit per cycle: identical 5.4% as the high-frequency cohort, because the 10-stamp ratio is invariant across visit cadence.
Low-frequency cohort: 1 visit per month. The infrequent guest who would need 10 months to complete a card. Annual visits: 12. Annual reward redemptions: 1.2 (assuming card retention across the year, which is generous). Reward cost per year: AED 18. Effective margin hit: same 5.4% per cycle on cards completed, though the realistic redemption rate for low-frequency guests is substantially lower because the card gets lost or forgotten before completion.
The damning observation: the effective discount rate is the same across all three cohorts. The mechanic does not differentiate between a guest who needs no incentive and a guest who could be moved by one. Worse, in dirhams the high-frequency cohort captures 9× the absolute reward spend of the low-frequency cohort (AED 169.50 vs AED 18), and the high-frequency cohort is the cohort least likely to have changed their behaviour because of the programme. The low-frequency cohort is the cohort the programme could in principle have moved, and it captures the smallest share of the reward budget.
The Behaviour Change Question
The right unit of analysis is incremental visits attributable to the programme, not total visits redeemed against. A well-run loyalty audit asks: of the 17 visits per month from the high-frequency guest, how many would have happened anyway and how many were caused by the punch card? The honest answer for an established regular is that nearly all of them would have happened anyway. The punch card is not buying you new visits from the high-frequency cohort; it is paying for visits you already had.
For the mid-frequency cohort, the question is more interesting. A guest visiting 4 times a month is in the moveable zone, close enough to "regular" that a meaningful incentive could push them toward 5 or 6 visits, but far enough from regular that the visit count is not locked in by habit. This is the cohort where the punch card could in principle do work. It does not, because the 10-stamp threshold is too far away to feel achievable, and the reward at the end is not time-locked to recent behaviour.
For the low-frequency cohort, the punch card is irrelevant. A guest visiting once a month is unlikely to retain a paper card across the ten-month redemption cycle, and even if they do, the incentive arrives so far in the future that it has no behavioural pull on the next visit. The same insight in the redemption-tier framing is in the Pareto-optimal redemption tier piece.
Three Mechanics That Beat the 10-Stamp Card
Threshold-based early unlocks. Instead of "10 stamps for a free item," the mechanic is "second visit within 21 days unlocks an instant reward." The incentive ties the reward to the behaviour the operator wants, repeat visits in close succession, rather than to cumulative count. The mid-frequency cohort, where most of the moveable margin lives, responds strongly to this framing because the threshold is achievable and the reward arrives within the timeframe of an active habit.
Low first-tier thresholds in a tiered points programme. A tiered points system with a first-tier reward at visit 3 or 4 · a small but real incentive, produces the same psychological pull at the moment that matters most: the visit where a guest is deciding whether they have a habit yet. The 10th-visit reward is too far away to do this work; the 3rd-visit reward is exactly where the behavioural inflection happens. The full tier design framework is in the redemption tier playbook.
Segmented incentives by behavioural group. The single largest mechanical improvement is to stop running one incentive against the entire guest base and start running different incentives against different cohorts. Champions get recognition rewards (early access to new menu items, named tier status, no discount necessary). Mid-frequency guests get frequency-bonus incentives. Promising newcomers get a first-tier reward at visit 3. At-risk regulars get win-back offers. The same loyalty budget, deployed against four cohorts with four different mechanics, produces a substantially higher return than a uniform 5.4% discount across the whole base. Habitu SmartSegments runs the cohort segmentation automatically against POS-integrated guest data and surfaces the four cohort actions on a single dashboard. The full ROI structure of segmented incentives versus uniform discounts is in the loyalty programme ROI piece.
The Operator's Choice
Running a 10-stamp card is the loyalty equivalent of paying a flat sales-tax on every visit your most valuable guests make. The cost is real, the behaviour change is mostly imaginary, and the cohort the mechanic could actually move sits outside the mechanic's operational reach. The fix is not to redesign the punch card, every redesign of the punch card produces a punch card with a slightly different threshold. The fix is to replace the mechanic with one that distinguishes between guests whose behaviour the operator wants to reward and guests whose behaviour the operator wants to change. Those are different jobs, and they need different incentives.