Restaurant repeat-visit programs: the numbers before and after

The 19% nobody measured and the 5.3 million it cost to ignore

A 180-cover restaurant in Bogotá closed January 2026 with 11,400 identified guests and a 30-day second-visit rate of 19%, while the general manager estimated from memory «about half». With an average check of 68,000 pesos and a 68% contribution margin, each percentage point of that rate was worth 5.3 million pesos a year, so the gap between the measured 19% and the imagined 50% was not a perception error but a hidden line of the P&L. The team had spent eighteen months debating a 4% price increase —when ACODRES had already reported that Colombian restaurants raised menu prices 9.8% since February 2025 to sustain 98,000 jobs— and had never once discussed the 19%. These figures together trigger one decision: before you touch the menu, measure repeat business. Because bringing someone back a second time costs a fraction of bringing them in the first time, and that gap widened brutally over the past decade.

Why is the second visit the cheap lever?

Customer acquisition cost rose 222% in the eight years through 2025, according to Marqii, which means every peso spent at the top of the funnel now buys less than a third of the traffic it bought in 2017.

That number rewrites the arithmetic of any restaurant marketing budget. If acquiring costs three times more and contribution margin still hovers around 65-70%, the only variable an owner moves week to week without fighting suppliers or the landlord is frequency: taking a guest from two visits a year to three across a base of 11,400 records generates more cash than any reach campaign. And the operating conclusion is uncomfortable to accept: the acquisition budget gets cut and that money moves into the repeat program. A guest who returns within the first month is three to four times more likely to become a regular than one reactivated at 90 days, according to aggregated data published by hospitality loyalty platforms.

The 30-day window rules everything else

The first decision in a repeat program, then, is not what to give away but WHEN to speak, and that hierarchy escapes almost everyone because the discount is visible and the calendar is not. Turn it around: if you delay the first contact to week five so as not to «bother» anyone, you hand that guest to the restaurant across the street precisely when the memory of your kitchen still weighed on the decision. Channels handle that rhythm fine —Omnisend measured a 25.1% average email open rate through 2023— so customer fatigue was never the bottleneck. The rule that follows: day 7, day 18, day 27. After that you are no longer retaining, you are reactivating, and reactivating costs what acquiring costs. Check identification is the real bottleneck of repeat business and almost no operator treats it as such, because it demands floor discipline instead of a software purchase.

Without a name on the check you are running ads, not retention

In the Bogotá operation, going from 14% to 57% of identified checks moved sales more than any change of message: with the same base of 11,400 records, the actionable universe went from 1,596 to 6,498 reachable guests, and not a single word of the email had changed. Capture mechanisms are already cheap. The QR menu, which by now most places installed for other reasons —QR Code puts the average annual saving for a restaurant that adopts it at US$3,600—, works as a doorway into registration if you wire it to the CRM instead of letting it dead-end in a PDF. My judgment here is blunt: until identification clears 50%, any investment in message automation is premature. When a program hands 30% off to the entire base, what it confesses is that it cannot estimate return probability per guest, and it pays for that ignorance in margin.

A big discount gives away a model that doesn't know who it's talking to

On a 68,000-peso check with 68% contribution, a flat 30% eats nearly half the margin of that visit; the same money, aimed only at the third with low probability of returning, costs a third as much and converts better because it lands where the incentive actually changes the decision. The operation also carries enough volume to justify segmentation: more than 40% of adults order delivery or takeout three to five times a month, according to UpMenu, which gives you a near-weekly habit signal to read. I got this wrong for years, recommending uniform points programs, elegant to explain and expensive to sustain. The right decision is the ugly one: surgical discounting, small, aimed at whoever is genuinely leaving. Segmenting 11,400 records by consumption habit, writing different messages per segment and deciding who NOT to write to was, three years ago, agency work on a monthly retainer; today it is a flow that runs overnight and leaves the manager a list of 300 names with the reason each one is on it.

AI does not create repeat business, it makes it cheap to run

The same technology already proved its return on the other side of the business: TimeForge documented labor cost reductions of 8-12% with AI-driven scheduling in 2025, plus forecast accuracy above 90%. Diego F. Parra keeps pressing the Masterestaurant teams on the distinction almost nobody makes when buying tools: technology does not generate the visit, it lowers the cost of administering it. An owner who buys the platform without having measured their second-visit rate bought a dashboard, not a program. First the number, then the flow. A well-built repeat program moves guest LTV between 15% and 30% within two quarters, and that range comes from arithmetic rather than magic: lifting the second-visit rate from 19% to 26% across 11,400 records adds roughly 800 visits a year, which at 68,000 pesos with 68% contribution leaves close to 37 million pesos of margin without one extra seat or one new dish on the menu.

How much LTV moves in two quarters and why frequency wins?

Frequency beats check size because it is the only variable you control week to week without asking the market for permission:

raise prices and the guest punishes you, cut costs and the supplier limits you, but the contact calendar is yours to decide on Monday. The best QSRs enroll around 110 new loyalty members per store each month, according to Paytronix's Annual Loyalty Report 2024, and that steady cadence —not one heroic campaign— is what compounds. Set the quarterly target in points of second visit, not in pesos of sales. Three numbers and the action each one triggers. First, your 30-day second-visit rate: if it sits at 19% as in the Bogotá case, every point is worth 5.3 million pesos a year on that base, so set the quarterly target in points of repeat business and review it every Monday. Second, 222% growth in customer acquisition cost over eight years through 2025 (Marqii): shift a fifth of the acquisition budget into the retention program this quarter and measure the result at 90 days, not at 30.

The 3 figures you should tattoo on yourself

Third, your share of identified checks: below 50% do not buy software, fix the floor —a capture script at the check, a QR wired to the CRM, a small incentive on the first visit— because without a name there is no funnel. Start tomorrow with number one; the other two cannot be calculated without it. The 30-day window governs everything else. A guest who repeats within the first month is, according to aggregate data published by hospitality loyalty platforms, three to four times likelier to become a regular than one reactivated at 90 days. So the first decision is not what to give away but WHEN to speak. Ticket identification is the real bottleneck, and almost nobody treats it as one. Without a name on the ticket there is no sales funnel to manage: you are running advertising, not retention. Moving from 14% to 57% of identified tickets did more for revenue than any change of message.

The differences that actually move cash

A large discount signals that the model does not know who it is talking to. Once the system estimates return probability per guest, 62% of the base receives an invitation with no discount and responds anyway; the incentive stays reserved for the cold segment. The margin saved there funds the whole program. Customer acquisition cost and recovery cost are different accounts and should not be blended. Recovering a visit cost 4,800 pesos against 22,000 for a new guest, but the new guest brings future base and the recovered one does not. A repeat-visit program does not replace acquisition: it lowers the pressure on it. Automation contributes consistency, not ideas. The difference between a program that pays and one that dies in month four is rarely the strategy; it is that somebody stopped executing on Tuesday. That is exactly what an automated routine solves well. Measuring by cohort instead of by campaign changes the boardroom conversation. A coupon can «work» and still cannibalize visits that would have happened anyway; only treated-versus-control comparison reveals how much revenue is truly incremental.