Customer Repurchase Program in Restaurants: A Real Before vs After Case With Masterestaurant

The problem nobody was measuring: 82% of customers never came back

Marta Restrepo had spent years behind the register at El Fogón de Marta without suspecting the real hole wasn't in Saturday night sales but in Tuesday's silence: 82 out of 100 diners never came back within 60 days of trying the menu. With 38 tables and a check that hovered around $42,000 COP, the restaurant rang up a strong weekend and quietly lost, with nobody logging it, the guest who'd already returned once. I pulled 1,200 transactions from the prior quarter and real repurchase never cleared 18%. Marta placed her regulars by face, never by data: POS and CRM ran on separate tracks, each blind to the other. That blindness cost real money: $3 to $5 million pesos a month in social coupons, with redemption barely brushing 4%. In 14 months the cost of adding a new customer rose from $18,000 to $31,000, while operating margin bled 2 points quarter after quarter.

The problem nobody was measuring: 82% of customers never came back — in practice

Foot traffic was never the problem. Retaining someone who already paid costs 5 to 7 times less than winning someone new. The industry knows the number by heart; it rarely turns into cash policy. I found exactly that gap at El Fogón de Marta, the same one I run into across dozens of restaurants every year: the whole budget chased new faces, and nobody gave the guest who'd already paid once a structured reason to return. We audited 3,600 transactions from the prior 90 days over three weeks of diagnosis, and the first NUMBER that stung surfaced fast: 8% of the base generated 29% of recurring sales with zero differentiated treatment. The second finding hit harder: mass coupons pulled in single-visit customers, not high-value ones. Spending $4 million pesos a month to retain the wrong customer turns out worse than spending nothing at all.

The method: three customer tiers, distinct rewards, funded at 2.8% of the average check

We skipped any generic points system and built three tiers off each customer's last 90 POS transactions, every one with its own reward priced at 2.8% of average ticket — never a discount off the sale price. The distinction isn't cosmetic: a 15% discount eats margin, while a 2.8% point, granted only on a real repurchase, costs less and still changes behavior. The top 8% of the base got no bigger markdown, just early access to the new menu and priority seating without a reservation. No discount, just different treatment. The occasional tier got a win-back incentive triggered at 21 days of absence, and the frequent tier earned points faster. The whole scheme came together in six weeks, with weekly tracking of four metrics: frequency, ticket, redemption, cost per point granted. The fixed-date campaign — Mother's Day, payday, month-end, same offer for everyone — rarely clears 12% open rate, and that's exactly what El Fogón de Marta ran before the program, with redemption barely touching 4%.

WhatsApp automation: from a 12% open rate to 58% in eight weeks

We wired the POS into WhatsApp and campaigns left the calendar behind: the message fires when a customer hits 21 days without a logged purchase, not when a date on a flyer says so. That switch alone did it. Open rate climbed to 58% in the first eight weeks, and redemption jumped from 4% to 37%. No generic coupon here: the text named the last dish the customer ordered and offered back a welcome point worth 2.8% of their next check. By day 90 of the program, monthly visit frequency per active customer nearly doubled: from 1.3 to 2.7, without adding a single table or stretching service hours. Not one extra seat. Repurchase rose from 18% to 41% on the same transaction base, and average check grew from $42,000 to $51,000 COP because frequent customers spent more each time they came back carrying an accumulated point.

Results at 90 days: visit frequency from 1.3 to 2.7 monthly visits per active customer

CAC fell from $31,000 to $14,500 because ad spend stopped chasing cold audiences and got reinvested in customers with real history. And food cost, inside the hard 32% ceiling Masterestaurant sets, held at 31% even as redemption climbed from 4% to 37%: point cost was priced against margin, never against sale price. Operating margin, which had been sliding 2 points a quarter, stabilized by week 10. Of everything the diagnosis turned up, the finding that mattered most wasn't the repurchase rate — it was discovering El Fogón de Marta already had ambassadors, customers visiting 4 or more times a month, and that 8% of the active base generated 29% of recurring sales. That segment wasn't asking for a discount. It was asking for RECOGNITION. We built a zero-food-cost VIP protocol: seasonal-menu access 72 hours ahead of public launch, peak-hour seating with no reservation, and a birthday call placed by the chef, not a system.

The ambassador: 8% of customers generating 29% of recurring sales

Running it cost $180,000 COP a month for 94 customers, and the payback beat any paid campaign: those same 94 referred an average of 2.1 new customers each that quarter, 197 first visits without a dollar spent on ads. Treating every customer the same wastes the most profitable lever any restaurant has. Why food cost didn't spike as redemption jumped from 4% to 37% is the technical question I get asked most when I present this case, and the answer sits in how we built the point, not in luck. In the classic model — 10% off your next visit — the cost comes straight out of the sale price and compresses margin with no way around it. In the one we ran at El Fogón de Marta, the point is worth 2.8% of the check and gets funded from contribution margin, never from price. No real transaction, no cost.

Why food cost stayed at 31% even as redemption climbed to 37%?

On top of that, the reward activates on the FOURTH visit, not the second, so the customer has already logged three full checks before collecting anything back.

Food cost held at 31% because margin was the control variable by design, not an uncomfortable discovery at month-end close. If your restaurant runs more than 400 transactions a month and the POS identifies customers, the El Fogón de Marta case transfers almost untouched. With one catch: the exact 2.8% funding figure does NOT copy over as-is, because that number came from a contribution-margin analysis specific to that one business. In my diagnostics the right range runs 2% to 4% of ticket depending on starting food cost and sales mix — restaurants above 30% food cost should stay near 2%, and those under 28% have room to push to 3.5%. What does transfer untouched is the 21-day trigger, the three-tier segmentation, and WhatsApp over a fixed date on the calendar.

What to replicate and what to adjust if you are not El Fogón de Marta?

The mistake I see most when restaurants copy this model is funding the point from sale price instead of margin, and that slip alone raises food cost 1.5 to 2.5 points in the first quarter.

The POS now sorts every customer by their last 90 transactions — occasional, frequent, ambassador — instead of whoever's memory is at the register, and each tier gets a perk funded at 2.8% of average ticket. At 21 days without a visit, WhatsApp fires the message on its own; no fixed date, no coupon blast for everyone, and open rate climbed from 12% to 58%. The point is priced against margin, never against sale price, which is why food cost held at 31% right as redemption climbed from 4% to 37%. The top tier — 8% of the base, source of nearly a third of recurring sales — gets no bigger discount now, just early entry to the new menu.

The 5 differences that explain the jump

Customer acquisition cost fell from $31,000 to $14,500 pesos because ad spend shifted from chasing new faces to keeping the ones who already buy: the ratio flipped from 80/20 to 35/65.