Loyalty programs are one of the most common categories of restaurant marketing spend and one of the least honestly measured. Every operator I know has one or wants one. Almost none of them can tell you what the program actually returns against a matched control group. The reason is that most programs are designed the wrong way from the first meeting.
The wrong-way design goes like this: someone in leadership says "we should have a loyalty program." The team looks at what competitors do. Something with points. Something with a birthday freebie. Something with tiers. Sign up on the tablet or through an app. Ten dollars off after your tenth visit. Launch. Report enrollment numbers. Feel good.
Twelve months later the program has 8,000 enrolled members, has given away roughly $60K in discounts, and nobody can honestly say whether any of it changed guest behavior. The enrolled members were probably going to visit anyway. The discount just made the visit cheaper for them and for you.
Below is the design framework for a program that pays back. It came out of the Zareen's rebuild where the group had a functioning bar, a healthy corporate catering channel, and a growing dine-in base that needed the visit frequency to increase to justify continued investment in the model.
Start with a behavior question, not a program question
The first mistake is designing the program before deciding what behavior you want to change. There are four possible behavior targets:
- Visit frequency. Get guests who come once a month to come twice.
- Share of wallet. Get guests who come but do not order beverages to attach beverages. Get takeout-only guests to try dine-in.
- Referral. Get existing guests to bring new guests.
- Occasion capture. Get guests to book their birthday, anniversary, or work dinner at your restaurant instead of a competitor.
Each target implies a different program design. A visit frequency program looks nothing like a share-of-wallet program. A referral program looks nothing like an occasion-capture program. Trying to do all four in one program almost always produces a program that does none of them well.
Pick one target. Design for that target. Measure against that target. Add the second target only after the first is working.
The three levers, in order of ROI
Fig. 1 · Recognition costs almost nothing. Discounts cost you your margin.
Recognition
The cheapest and most powerful loyalty lever is recognition. A hostess who greets a returning guest by name. A server who remembers a wine preference. A general manager who comes by the table on a birthday and says thank you. These cost almost nothing and produce more emotional loyalty than any discount ever will.
The system for this is a light CRM (Toast, OpenTable, or something purpose-built) that surfaces guest history at the moment of interaction. The hostess sees "Sarah, second visit, birthday next week" when the reservation is checked in. The server sees "prefers table 12, ordered the biryani last two times" when the seat is assigned. This is not stalking. It is professional hospitality at scale.
Experience
The second lever is aspirational experience. Early access to a new seasonal menu. An invitation-only chef's tasting. A wine pairing dinner for members only. These reward loyalty with something guests cannot buy elsewhere and cannot get without membership.
The cost is modest (one evening per quarter, maybe four staff, food cost at wholesale) and the emotional return is disproportionately high. Members feel like they are inside something. That feeling is what drives visit frequency, referral, and lifetime value.
Discount
The lever most operators reach for first is the lever with the lowest ROI. A 10 percent off next visit does two things: it trains the guest to expect the discount (which they now feel entitled to), and it hands you back your own margin on a visit that would have happened anyway.
Discounts are not useless. They work in narrow situations: winning back lapsed guests who have not visited in 6+ months, driving off-peak visits (Tuesday lunch discount to a Friday-dinner regular), or converting takeout-only guests to dine-in. Used in those specific contexts, discounts move behavior. Used broadly across every enrolled member, discounts subsidize.
The loyal guest does not need 10 percent off. They need to feel seen. The lapsed guest does not care about being seen. They need a reason to come back. Different guests, different levers.
Measure against a matched control
The single most important discipline in loyalty program measurement is the matched control cohort. Compare enrolled members not to the average of all guests, but to a matched cohort of non-enrolled guests who look the same on the dimensions that matter: check average, visit history, geography, cohort tenure.
If the enrolled member visits 6.4 times a year and the matched non-member visits 4.2 times a year, and the incremental margin from those extra 2.2 visits exceeds the cost of rewards delivered, the program is working. If the enrolled member visits 6.4 times and the matched non-member visits 6.1 times, the program is not moving behavior. It is discounting guests who would have come anyway.
Report this quarterly. Not annually. Not in enrollment numbers. In behavior lift. Enrollment numbers are vanity. Behavior lift is the operating truth.
Collect narrow data
Every field you collect that you never act on is friction with no return. The temptation is to collect everything at signup: name, email, phone, birthday, anniversary, favorite dish, dietary restrictions, preferred communication channel, referral source, marital status, kids' ages, favorite table.
Do not. Collect narrow. Name, email, phone, birthday. That is enough to launch. Add fields only when the marketing team has demonstrated they can act on the fields already collected.
The birthday field is worth calling out. A birthday recognition (dessert with a candle, sung to by the team) at the table costs about $3 in food cost and produces one of the highest guest satisfaction moments possible. Cost effective. Emotionally durable. Do it for every enrolled member whose birthday you know.
The technology decision
For most single-brand groups the loyalty technology is a feature of the POS or the reservation platform. Toast, Square, and OpenTable all offer light loyalty modules. For multi-brand groups or groups with corporate catering as a major channel, a dedicated platform (Punchh, Paytronix, Thanx) is worth the additional cost because you get real segmentation and campaign tools.
Do not build a custom loyalty app. Restaurant loyalty apps are one of the most reliably disappointing technology investments in the industry. Guests do not download restaurant apps. Guests use their phone's wallet, their email, and text messages. Design the program to work through those channels.
The redemption experience is the loyalty moment
The moment that determines whether a guest values the program is the moment they redeem a reward. If the redemption is smooth (server acknowledges it, applies it, thanks the guest, moves on) the guest feels rewarded. If the redemption is awkward (server does not know how to apply it, has to call a manager, takes five minutes) the guest feels the program was fake.
Train every server on redemption. Test it at every unit at least quarterly. The redemption experience is the whole loyalty program from the guest's perspective. Everything else is invisible.
What I would tell a first-time operator
- Do not launch a program without a behavior target. "Engage guests" is not a target. "Move dine-in visit frequency from 3.8 to 5.0 for enrolled members within 12 months" is a target.
- Recognition first, experience second, discount third. Reverse the order and you subsidize your own margin.
- Report against a matched control cohort quarterly. If you cannot measure lift, you cannot manage lift.
- Collect narrow data. Start with four fields. Add only when you use.
- The redemption experience is the program. If it is awkward, the program is broken.
The point
Loyalty is a real business capability. Loyalty programs, done as they usually are, are not. The difference is design discipline. Pick a behavior to move. Choose the lever that moves it. Measure against a control cohort. Iterate on redemption experience. Do not subsidize guests who were going to come anyway.
A well-designed program lifts visit frequency by 15 to 30 percent among members, produces measurable share-of-wallet expansion, and generates the data that makes marketing decisions better. It also usually pays for itself inside a year.
A poorly designed program feels productive because enrollment numbers grow, and it costs you your margin quietly month after month. If your program cannot answer "what behavior did we change" against a matched cohort, redesign it. That is not a marketing failure. That is an operator failure.