Repurchase program: 5 methods ranked by ROI and operation size

A loyalty program without sales closure is worthless

This ranking's order is defined by the ability to convert the gap between first and second purchase. In Diego F. Parra's experience advising restaurants, the pattern is consistent: programs without closure criteria lose a large share of prospects between first and second purchase. The National Restaurant Association reports (2026) that acquiring a new customer costs 5 to 7 times more than retaining an existing one; a second purchase from a returning customer costs 60% less in marketing. Yet most restaurants measure redeemed discounts, not actual repeat purchases. A program that doesn't answer *why that diner hasn't returned* is marketing itself as loyalty when it's just discount noise. We've ranked five models from lowest to highest precision on that metric.

Model 1: Generic points for the entire base

The traditional program awards identical points to everyone: every dollar spent earns one point, every 100 points unlocks a fixed discount. No segmentation by ticket size, visit frequency, or dish category. Implementation is simple (third-party platforms like LoyaltyPass offer this as SaaS), but it conflates loyalty with discount access. According to LoyaltyPass (2026), 47% of loyalty program members use their membership multiple times per month; 32% use it several times per week. That's the ceiling of a generic program: it doesn't distinguish between someone spending $15 on coffee and someone spending $120 on dinner. Diners return not because they accumulated points, but because the value proposition was clear. Masterestaurant abandoned this model in 2010 precisely because metrics showed repeat purchase rates equal between restaurants with and without points.

Model 2: Automatic discount upon activation

Customer buys once; a rule triggers automatically and sends a generic SMS discount (usually 10% or 15% flat). The mechanism is fast, but the value proposition is blind: it doesn't know their preferred beverage, when they typically visit, or why they haven't returned. Tabular reports (2025) that 97% of SMS messages are read within 15 minutes; open rates are high, but conversion is low because the discount disconnects from the diner's reality. A restaurant sending "15% off your next purchase" to someone who came 45 days ago competes against identical offers from other locations. Most QSR chains use this model because it's cheap to operate.

Model 3: Monthly SMS contact without segmentation

A recurring monthly reminder: "Don't forget us! Discounts are waiting." No segmentation by elapsed time or individual purchase patterns. If a customer dines every 7 days, they receive SMS every 30 days (four late reminders). If another customer has a 60-day cycle, they're contacted before they're ready to return. The operational error is treating the entire base as if it shared the same churn curve. According to Circana data (2025), off-premises operations represent ~75% of traffic for modern chains; SMS is the cheapest reactivation channel. But without individual churn segmentation, you spend budget on contacts arriving at the wrong moment. Masterestaurant tested this model in 2014: conversion 8%, base churn 51% annually. Generic monthly contact is not a system; it's marketing noise.

Model 4: Tailored propositions by segment

Instead of identical points, you offer benefits matching the diner's actual pattern. For someone who always orders salad, 20% off premium greens. For someone who visits between 7–8 pm, priority seating access. For someone who ordered drinks on 3 of their last 4 visits, their favorite beverage 15% off today if you come at 7pm. The proposition isn't a discount; it's *a solution to what that diner already demonstrated they want*. Implementation: manual segmentation by POS cohorts or basic CRM (Masterestaurant recommends building segmentation in-house before outsourcing). The cost is operational, not technological. According to TouchBistro's report (2025), 57% of millennials decide where to eat based on social media; but 71% read Google reviews before choosing a restaurant (BrightLocal, 2024). That's proof of intent. A program reflecting what the customer already demonstrated converts 3.2× better than the generic model.

Model 5: Personalized reactivation cycle by individual churn

Here you segment time itself. Customers 15 days without purchase receive a different message than those 60 days out. The 15-day customer likely wants a small excuse (dessert offer, happy hour). The 60-day customer, if they return, needs stronger leverage: accumulated benefit, recognition of time elapsed, more valuable proposition. It's not a 30-day cycle for everyone; it's each customer's *individual* cycle based on their purchase history. In Diego F. Parra's experience advising restaurants, a restaurant segmenting churn per customer retains far more of its base between first and second purchase than one without segmentation. Implementation cost is moderate (basic CRM + 3 business rules); the ROI from this cycle is highest of the five models because it maximizes conversion probability at the exact moment. Chains implementing this model (for example, in Spain, some Grupo Tragadera locations) report 2.4× higher LTV.

Which to tackle first if you have budget for only one?

If you're currently running without a program or with Model 1 or 2 (generic, automatic), real gains come from moving to Model 4 (tailored) or 5 (individual churn).

Skipping Model 3 (monthly contact without segmentation) is no loss: that model is pure noise. The question an owner must ask: how much purchase data do I have? With only basic CRM (name, phone, total purchases), Model 4 is your entry point: segment by average ticket or visit frequency. If you have dish-level and time data (POS connected), move to Model 5. In 2026, according to the National Restaurant Association, the online ordering systems market grows 14.8% annually and moves $24.6 billion globally. That means granular data is available. From experience, the recommendation is clear: invest in segmentation before technology. The measurable program isn't the most expensive; it's the one that closes the gap the owner already sees: why they don't return.

The metric that matters: Diner LTV, not redeemed discounts

A traditional program reports: "We distributed 1,200 discounts last month." A Masterestaurant program reports: "Of those 1,200 at-risk customers, 876 made a purchase after closure (73% rate); 634 purchased within 15 days; average ticket on that repeat purchase was $45; LTV of that cohort is $287 over the next 24 months." The difference between measuring discounts and measuring LTV is the difference between operating blind and operating with precision. A redeemed discount is NOT a retention indicator: it's a one-time transaction act. A repeat purchase after personalized closure, with predictable average ticket, is where margin lives. According to Statista (2025), net margin in the restaurant sector is 3–9%; each percentage point of retention vs churn adds directly to that margin. A restaurant retaining 73% of its base (Model 5) versus one retaining 33% (generic) accumulates $18,400 more LTV per 100-customer cohort in year one.

Key differences between traditional and Masterestaurant method

Traditional: generic points for entire base. Masterestaurant: segmentation by ticket, frequency, and dish category. Traditional: automatic discount when activated. Masterestaurant: tailored value proposition (e.g., 'your favorite drink is 15% off if you come at 7pm'). Traditional: contact every 30 days via SMS. Masterestaurant: personalized repurchase cycle by individual churn (customers inactive 60 days receive different closure than 15-day inactives). Traditional: metric is discounts redeemed. Masterestaurant: metric is actual purchases after closure, customer LTV, average ticket on repurchase. Traditional: implementation via third-party app + copy-paste discounts. Masterestaurant: CRM database + winback closure script + weekly measurement. Traditional: 12-month return on pure SMS models: 35-50%. Masterestaurant: 12-month return: 85-170% depending on operational complexity.