Only three types of loyalty programs fit one shop

What a loyalty program type actually is, and which ones survive one counter

Strip the vocabulary away and a loyalty program type is just the rule that decides what a customer's next visit earns them. Everything else is packaging.

Size is not the test. What matters is whether the rule needs something from outside your four walls: a partner, an app store, or a pricing analyst.

A stamp card counts visits, which works when every ticket is about the same size

Nick Bombaij and Marnik Dekimpe studied 358 grocery banners across 27 European countries for the International Journal of Research in Marketing in March 2020. Direct, immediate rewards raised sales productivity. The effect disappeared once a retailer ran a more complex progressive-reward system.

A stamp card is the simplest thing on that table. After the result above, that is a reason to choose it rather than an apology for it, and the customer knows the rule after hearing it once.

The limit is arithmetic, not taste. A stamp card pays the same reward to the person buying a $4 drip coffee and the person buying a $9 pour-over, so if your ticket sizes sit far apart, you are quietly running a bigger discount for your cheapest customers. We went through what that reward costs and who ends up collecting it in an earlier piece.

Points are for tickets that swing

A salon is the clearest case. A trim is $45 and a balayage is $220, and no stamp card can treat those as the same event without insulting one of the two customers. A restaurant has the most money riding on the same choice, because the check covers a table rather than a person.

Starbucks hit this wall in public. Until April 2016 the program gave one star per visit, and twelve stars bought a free drink. TIME reported on 22 February 2016 that the company was switching to two stars per dollar, which moved a free drink out to 125 stars, or about $62.50 of spending. Part of the reason was operational: roughly 1% of transactions were customers asking to have their items rung up separately so each one earned its own star.

The lesson is about order. The switch was correct and it still cost them, because the customers who had been earning a free drink every twelve visits worked out very quickly what had just happened to them. Pick the counting unit before you launch, not after your regulars have learned the old one, and settle the rest of what a launch decides at the same time.

Points carry a trap of their own. Joseph Nunes and Xavier Drèze warned in Harvard Business Review in April 2006 that two rewards of identical economic value can produce very different amounts of buying. Name the reward in the thing, not in the currency. "A free cut at 500 points" is worse than "your sixth cut is free", even where the two are the same offer.

One VIP tier works at any size, as long as nobody can fall out of it

Tiers look like a big-company mechanic, and the research says they are not. Drèze and Nunes published a set of experiments in the Journal of Consumer Research in April 2009 showing that status is read relatively: adding a tier below yours makes your tier feel more elite, while enlarging the top tier dilutes it. None of that depends on absolute numbers. A shop with 300 customers can make 30 of them feel picked out as legitimately as Hilton can with a few million, because the mechanic runs on who sits above and below rather than on how many.

There is a version that goes badly. Tillmann Wagner, Thorsten Hennig-Thurau and Thomas Rudolph asked what taking status away costs you, in the Journal of Marketing in May 2009, and found that demoted customers reported lower loyalty intentions than customers who had never been given the status in the first place. Being dropped from gold is worse than never being gold.

So: one tier, earned once, never revoked. Annual requalification is a feature for airlines with revenue-management teams, and at one location it does nothing except give your best customers a date on which to be disappointed. In our own product a tier sits on the paid plans, which is worth knowing before you design around one.

Paid subscriptions break on exactly the customers they attract

This is the type that looks most tempting right now, and the two chains that pushed hardest on it have both pulled back.

Pret A Manger launched a coffee subscription in 2020 at £20 a month for up to five barista-made drinks a day. It went to £25 in February 2022 and to £30 in April 2023. In July 2024 Pret scrapped the free-drink model altogether, replacing it with half-price drinks, and managing director Clare Clough measured the old deal against "the £360 a year people have to pay". Subscribers had bought over a quarter of a billion coffees in four years. I checked today. Club Pret is £5 a month, and the drinks are half price rather than free.

Panera is going the same way. The Unlimited Sip Club launched in 2022 at $10.99 a month, and Nation's Restaurant News reported on 13 July 2026 that it now costs $14.99 and that the word "unlimited" has been dropped from the name. From 19 August 2026 members are capped at 30 drinks a month. Buried in the same reporting is the number that settles this for a small business: Panera says four drinks a month covers the subscription fee.

Everyone past their fourth drink is costing you money against the menu price. At a chain that barely matters. Hundreds of thousands of light subscribers dilute the heavy ones, and a finance team reprices the whole thing whenever the mix drifts, which is exactly what both of these stories are.

The honest counterpoint is that paid loyalty works when you can afford to steer it. Panera capped and repriced its program rather than killing it, and CEO Paul Carbone has said the chain is testing a move to points across its roughly 70 million MyPanera members. That is three corrections in four years. You get none, because the people who sign up for a drinks subscription at one location are your regulars, so your subscriber roll and your heaviest-using customers are the same forty people.

Coalition programs need partners you don't have

Bombaij and Dekimpe found something else in those 358 retailers: the positive sales effect also disappeared when a retailer was part of a multivendor program. Sharing a currency with other businesses did not add to the result. It removed it.

American Express built the biggest one. Plenti launched in 2015 with Macy's, AT&T, Hulu, Expedia, Enterprise, Alamo, Nationwide and Direct Energy, and shut down on 10 July 2018 after the partners left one by one. Retail Dive reported on 18 April 2018 that research by Maritz found half of consumers were not familiar with Plenti at all, that fewer than half of self-identified members had ever redeemed a reward, and that most members had only bought from one or two of the partner brands.

That last finding kills the local version. Three shops on one street sharing a card sounds neighborly, and it still needs somebody to hold the ledger, agree what a point is worth in each business, and settle up when the bakery's customer redeems at the bookshop.

Cash back, charity and arcade games each fail on different arithmetic

Cash back and percentage discounts fail on margin, and the failure is easy to miss because it looks cheaper than giving something away. A free drink costs you its ingredients. A 5% discount comes off the sale price on every visit, including all the visits that were happening anyway, which is the expensive half of the loyalty bill.

Charity programs convert the customer's reward into a donation, so the customer gets nothing back and the gesture has to carry the entire program. That works when the donation is the reason people chose you in the first place. At one location, 2% of a month's coffee sales is a donation nobody can feel, including you.

Gamified programs need somewhere to live. Antavo's examples are Domino's Piece of the Pie Rewards and KFC's Rewards Arcade, which are pieces of software with engineers attached to them. The version you can actually run is a double-stamp Tuesday, which is a stamp card with a calendar.

Referrals are not a loyalty program, and are worth running anyway

Referral schemes appear on every list of program types, and they belong on a different list, because they bring in people who have never bought from you rather than changing what your existing customers do.

The evidence is the strongest here. Philipp Schmitt, Bernd Skiera and Christophe Van den Bulte tracked roughly 10,000 customers of a German bank for nearly three years for the Journal of Marketing in January 2011 and found referred customers were worth at least 16% more than comparable customers acquired another way, with a retention advantage that held up over time. They also found the gap varies widely between segments, so a blanket referral bonus for everybody is not what the paper supports.

Run one alongside your program. Just don't count it as the program, because what it moves is how many customers you have rather than how often the ones you already have come back.