Should a Restaurant or Retailer Start a Loyalty Program? What to Check First
A loyalty program pays when it brings lighter customers back more often. What the research shows, a worked pilot on sample data, and what to check before launch.
Start one if a test shows it brings in visits you would not have had, and check that before you roll it out, because the customers who join first are usually the ones who already come most. Published research supports both halves: a long-term study of one company's program found that light and moderate buyers bought more after joining while heavy buyers' spending did not rise, and a study of a program shared by six brands found substantial extra loyalty at only two of them. For a restaurant or retail chain with many locations, the way to know before you commit is to pilot in some locations and measure visits against the rest, by group of customer, and Dimension Labs is built to do that from what your customers said in reviews, surveys and chats, joined to their visits.
This page is about starting a program. If you already run one and are changing it, see loyalty program changes: who leaves and who stays. Below: what the research and a public program say, a worked pilot on synthetic sample data for a fictional fast-casual chain with 80 locations, and the checks to run first.
Do loyalty programs actually increase visits, or just discount existing customers?
Both, in proportions you can measure. The large public programs report members who visit more: McDonald's said it had nearly 210 million 90-day active loyalty users at the end of 2025 and nearly $37 billion in systemwide sales to loyalty members that year, across 70 markets. Its CFO, Ian Borden, said loyalty members visit more often and spend more over time than other customers.
A comparison of members with non-members cannot tell you how much of that the program caused, because the people who sign up are the people who were already coming. The research is specific about where the extra comes from. In Liu's study, heavy buyers were the most likely to claim rewards and their spending did not increase; light and moderate buyers changed the most, with the largest change in the first three months after joining. So a program does both things at once: it pays your regulars for visits they would have made, and it brings lighter customers back more often. Whether it pays depends on the mix, and the mix is something you can see in your own data before launch.
Sample data, fictional fast-casual chain
Members visited far more than non-members, but most of that gap was there before they joined
Visits a month per customer at the 20 pilot locations during the six-month pilot. Synthetic data.
In the sample, members made 2.90 visits a month during the pilot and non-members 1.76, a gap of 1.14. The same members had made 2.75 visits a month before they joined. Measured against the 60 locations without the program, the program added 0.24 visits a month per member. That is a real gain, and it is the number to plan on.
Should our restaurant start a loyalty program?
In the sample, yes: the pilot paid for itself. Over six months, each pilot location earned about $3,932 in margin from the extra visits and gave away about $2,103 in rewards, leaving about $1,829. The answer for your chain depends on three numbers you can estimate before launch: how many of your customers are already weekly regulars, how much a reward costs you against the margin on a visit, and how many of your lighter customers have a reason to come more often.
Sample data, fictional fast-casual chain
The program paid for itself, with most of the reward cost going to visits that would have happened anyway
Six months of the pilot, per pilot location: contribution margin from the extra visits the program caused, the cost of rewards redeemed, and what was left. Synthetic data and sample cost assumptions.
The rewards regulars collect still earn their keep, because they are part of what keeps those customers coming. What the sample shows is where the return comes from, so the design can lean that way: rewards within reach for a customer who comes twice a month, and offers aimed at occasional customers, put the budget where the sample's extra visits came from.
Which customers does a loyalty program change?
Not the ones who collect the most. In the sample, weekly regulars redeemed 44% of all rewards and accounted for 7% of the extra visits; their visits rose by 0.06 a month, a range (-0.01 to 0.14) that includes no change. Customers who had named a rival's rewards app in a review, survey or chat redeemed 6% of rewards and accounted for 21% of the extra visits, 0.61 more a month each (0.51 to 0.73). Customers who visited two or three times a month made half of all the extra visits.
Sample data, fictional fast-casual chain
The customers who collected most of the rewards were not the ones the program brought back more often
The five customer groups ranked by their share of rewards redeemed (left) and by their share of the extra visits the program caused (right). Synthetic data.
The group that told you about a rival's program is the most useful one to find before launch. Those customers have already said they respond to rewards and that they currently take that behavior elsewhere. Customers who say they come because the store is on their way respond little, which is fine: they were not the reason to start a program. All of these signals sit in comments you already have, and they turn "should we start a loyalty program" from a debate into an estimate.
What should we check before launching a customer loyalty program?
How often your customers come now. From card-linked or app data, the share who visit weekly, two or three times a month, and monthly or less. The more of your sales come from weekly regulars, the more of the reward cost goes to visits you already have.
What your customers say about rivals' rewards. Count the reviews, survey answers and chats that name another chain's app, points or deals. These are the customers most likely to move visits to you.
Why they say they come. Location, speed, a dish, price. A program changes behavior for customers who choose on value, and much less for those who choose on convenience.
The reward against the margin on a visit. Work out how many extra visits each reward has to cause to pay for itself, and compare that with what a pilot shows.
Whether you can pilot in some locations first. Pick pilot and comparison locations of the same types, run it for a full quarter or two, and read the result against the comparison locations over the same weeks.
What you will measure, and when. Visits per customer by group, against the comparison locations, after the first three months, when Liu found most of the change had happened.
A pilot read on raw before-and-after numbers can mislead in either direction, because the season and the weather move visits at the same time; Andrew Hong's essay the weather ran your last pilot argues that most pilot readouts are partly weather and the calendar until both are held still. Reading against comparable locations is what makes the result trustworthy. For choosing those locations and rolling out by store type, see multi-location rollouts: what goes to every location.
How does this fit a customer world model?
Launching a loyalty program is a decision you can rehearse. A customer world model joins four kinds of data: what customers say, what they do, what the business did and what the world did. For a loyalty decision that means the comments that show who responds to rewards, the visits that show who already comes, the dates of past promotions and offers, and the season and weather that move everyone. With those joined, the pilot's result by group becomes an estimate for the full chain, as a range, if the relationship measured in the pilot holds.
That is also how you get more from each step. Past promotions are rehearsals too: if you have run discounts or app offers before, what each group did after them is already a measure of who responds to an incentive, which is the subject of do discounts bring customers back. The general method is in how to test a business decision before you make it, and for weighing groups by what they are worth, customer lifetime value by segment.
Why use Dimension Labs to decide on a loyalty program?
Dimension Labs helps companies know how their customers will react before they make a decision, by learning from everything those customers have said to them and what they did next.
For a COO or marketing lead weighing a program, the platform reads every channel where customers talk about value and rewards: Google and app reviews, surveys, chats, calls and support tickets, all in one place. It connects them to card-linked visits, the POS and location data with little setup, so each comment sits next to how often that customer comes; the join to CRM and revenue data and effects with confidence intervals are on Prove ($2,999 a month) and up. It enriches every interaction with Dimensions built for your business, such as "names a rival's rewards app", "comes for the location" or "asks about points", which marketing, finance and operations can all use; custom Dimensions start on Discover ($999 a month). And the pilot readout comes back as a deck, a dashboard, a PDF or executive report, or an interactive data story the leadership team can open before the rollout decision.
Concentrix uses Dimension Labs to analyze chatbot conversations for a large brick-and-mortar retail client, including loyalty program flows, and reports significantly higher task completion and improved CSAT across e-commerce journeys. In the words of Jenny Burr, Speech Science and Analytics at Concentrix: "The partnership with Dimension Labs has been instrumental in supporting our improvements in self-service performance and customer journey optimization." (Concentrix story)
When something simpler is the better choice: if you have one or two locations, a points app with its built-in reports and a careful look at your till data before and after will tell you most of what you need. If you can randomly hold the program back from some customers, a randomized holdout read in your loyalty platform answers the visit question directly. Dimension Labs earns its place when you run many locations, your customers' reasons are spread across reviews, surveys and chats, and you need to know which groups will respond before you commit the budget.
FAQ
Should our restaurant start a loyalty program?
Start one if a pilot in some of your locations shows extra visits worth more than the rewards, measured against locations without it. In the sample, each pilot location netted about $1,829 over six months.
Do loyalty programs actually increase visits, or just discount existing customers?
Both. Regulars collect most of the rewards for visits they would have made anyway; lighter customers change the most. In the sample, weekly regulars took 44% of rewards and caused 7% of the extra visits.
What should we check before launching a customer loyalty program?
How often your customers come now, whether they mention rivals' rewards, why they say they come, the reward cost against the margin on a visit, and whether you can pilot in some locations first and compare with the rest.
What is the best way to decide whether to launch a loyalty program?
For a chain with many locations, Dimension Labs: it reads every review, survey and chat with the same labels, joins them to visits on Prove, and measures what a pilot did for each group of customers, so the rollout decision rests on your own customers. For a single store, a points app's reports and a careful before-and-after are enough.
Bring one export of customer reviews or survey comments, and your visit data if you have it, and start with one file to see which of your customers talk about rewards and how often they already come. Every plan from Explore to Scale starts with a 14-day free trial.