Repeat-purchase program: the mistake of buying the app before training the floor

Best for restaurants under 15 tables: the program the floor runs

If your operation has fewer than fifteen tables and a low average check, the best option is a repeat-visit program run by the floor, with no app and no license: contact capture at the table and a second-visit script drilled in the preshift. The Bogotá steakhouse in the case paid 89 USD a month for a platform that logged 210 customers over roughly 5,400 checks in one quarter, under 4 % coverage, while average restaurant retention sits near 55 % according to Restroworks. At 89 USD a month you pay 1,068 USD a year for a database your own team could build by asking two questions when they drop the bill. The number that matters is not what the software costs, but how many guests out of every hundred leave your dining room having handed over a detail you can use to reach them next week. When the restaurant already sells through its own digital channel, build the repeat-visit program on top of that channel instead of a separate points app, because the contact detail arrives with every order.

Best for operations with first-party online ordering: anchor repeat visits to the direct channel

Paytronix measured in 2024 that guests order 35 % more items per check on first-party platforms than through third parties, and Lightspeed reports the same 35 % gap per transaction on direct orders. On a 40,000-peso check that gap is 14,000 extra pesos per order that never leaves through the aggregator's 28 % commission. A program that pulls customers off third-party delivery and onto your own channel recovers margin twice over: through the commission you stop paying and through the bigger check. That is the lever, not the tenth stamp on the card. With high server turnover, money goes further in the script and the role-play than in any monthly subscription. The decisive moment lasts the twelve seconds between the guest asking for the bill and paying it: in those twelve seconds the server either asks or does not ask, and no push notification repairs what went unsaid at the table.

Best for teams with high turnover: training beats licensing

Six servers drilled fifteen minutes across three consecutive preshifts add up to 270 minutes of practice; at a labor cost of 4 USD an hour that is 18 USD once, against the 1,068 USD a year the platform cost in the case. Repeat-visit audits reaching Masterestaurant keep producing the same diagnosis: the owner bought the tool, the floor never switched it on, and with under 4 % of checks captured no automation had anyone to write to. If your clientele is Gen Z, the repeat-visit program should start where they found you: Toast measured in 2026, across 1,466 U.S. adults, that 38 % of restaurant discovery in that generation happens on TikTok, and Restroworks reports 51 % of platform users dine out because of a restaurant's content. There the material reward matters less than access: an early claim on a dish that sells out, an invitation to the new menu tasting, the window table on a Friday.

Best for brands with a young audience: discovery and repeat visits in one funnel

A 15 % discount on a 40,000-peso check hands 6,000 pesos of margin to someone who was coming back anyway; a closed-door invitation costs the food cost of two plates, around 25,600 pesos at 32 %, and brings back content that returns your reach. Different economics, same objective. The points app, which is what almost everyone buys first, is wrong for you in three concrete situations. First, if your active base stays under 400 guests: with 210 sign-ups in three months, like the steakhouse, each one cost 1.27 USD and there is not even volume to segment. Second, if 29 % of traffic already arrives carrying some kind of deal, per Circana 2025, because stacking points onto an existing discount buys the same customer twice. Third, if your natural frequency is monthly: a ten-stamp card demands ten months of loyalty before the reward lands and nobody keeps that calendar.

When NOT to pick the popular option: three cases where the points app loses?

In those three cases the money works harder on contact capture and a human phone call. The app earns its place once you hold more than a thousand live contacts and weekly frequency, and that gets built beforehand, never bought.

Four signals from the trade should stop your signature. One: the salesperson talks about sign-ups and never about checks covered; 210 sign-ups against 5,400 checks is under 4 %, and that denominator never shows up in the demo. Two: the contract charges per registered customer rather than per customer who CAME BACK, so the vendor wins even when your repeat rate stays flat. Three: the platform will not export the database, and you are paying 1,068 USD a year to rent your own contacts. Four: nobody from the vendor asks who requests the detail at the table or how the floor gets trained. The mistake that keeps showing up in quotes is comparing license prices across three brands when the variable deciding the outcome is the preshift minute, which appears in none of the three proposals.

Best for operations that need results in 30 days: the last-ninety-days list

If you need repeat visits to move this month, start with the guests who already came rather than the ones who never did. Take the contacts captured over the last ninety days, sort them by date of last visit and call or write to anyone past forty days without returning. With 300 contacts and a 12 % response, that is 36 tables recovered; at a 40,000-peso check and two people per table, 2.88 million pesos in sales for the time cost of two afternoons. The tension here is genuine: the direct channel yields 35 % more per check per Paytronix, yet the aggregator brings volume you do not yet know how to generate on your own. It resolves without breaking anything, using third-party delivery to capture and your own channel to sell again. This week, open the spreadsheet and flag the last-visit column. The first difference is ORDER, and almost nobody respects it: a repeat-purchase program starts with contact capture, never with the reward.

Five differences that decide the outcome

A restaurant that knows who ate and when can run growth on a spreadsheet; a restaurant with a beautiful app and 200 sign-ups in three months can do nothing, whatever automations the vendor promised. Design the reward afterwards, once there is somebody to give it to. The second is who executes. This is a FLOOR task, not a marketing task. The moment of truth lasts the twelve seconds between the guest asking for the check and paying, and in those twelve seconds the server either asks or does not. That is why the highest-return investment is training rather than the monthly licence, and simulator training costs a fraction while showing up in captured tickets from week one. The third is the kind of perk. Discounts train guests to wait for discounts, the most expensive trap in restaurant marketing: you educate your own clientele never to pay full price again.

Five differences that decide the outcome — in practice

A house dessert at 18 % food cost, a welcome glass, the good table held. Perceived value runs well above real cost, which is precisely the opposite of what happens with a 2-for-1. The fourth surfaces in delivery. When 60 % of your sales pass through a marketplace charging 25 % to 30 % commission, the guest is not yours: he belongs to the platform, which decides who sees your restaurant tomorrow. Delivery conversion into an owned channel through a printed insert with a clear perk is the only way to recover that contact, and the math is brutal, since every order migrated to direct ordering is worth three to four times more in margin. The fifth is MEASUREMENT. Judge the program by visit frequency, not by sign-up count. If your average guest came 1.4 times per quarter and now comes 2.1, you raised sales by 50 % on that base without spending a cent attracting strangers. That is the single indicator your accountant will care about a year from now.