Not long ago, an ad from a well-known Slovenian retailer caught my eye. It celebrated the successful launch of their loyalty programme. Lovely numbers: members buy more, their share keeps growing, revenue is up. The numbers kept nagging at me, so I opened the company’s published accounts. Revenue had indeed grown. Value added had fallen - as a share of revenue, and in absolute terms too.
The accounts don’t say the loyalty programme is to blame. Margins also get eaten by purchase prices, energy and wages. But they say something that should sting any director. The programme’s success story and the income statement speak two different languages. The programme counts revenue and members. The company lives on margin. And nothing in the publicly presented results shows how much discount hides in €100 of member revenue. Or in €100 of non-member revenue.
Here’s the thesis: the discount is not the problem. The problem is the unselective, unmeasured, non-incremental discount. A programme that offers everyone the same “10% off” can’t tell two buyers apart. The one who would have come anyway, and the one who needs convincing. It gifts margin to the first and never reaches the second. And it counts both as success.
Three times more: the most quoted, least useful number in the industry
By Nike’s own numbers, members who shop through its apps spend three times as much as guests on Nike.com. The figure circulates through the industry as proof that membership creates spending.
It doesn’t. Or rather: this figure can’t tell us that. The people who join are precisely the customers already attached to the brand. Comparing members with guests therefore also captures that self-selection. When Leenheer and colleagues statistically accounted for self-selection on a panel of Dutch households, the membership effect stayed positive. But it came out seven times smaller than the naive member-versus-non-member comparison.
Seven times - in their sample. That ratio won’t transfer to your business. The lesson will. If your report shows a 30% gap, you don’t know how much of it the programme created. Not without a control group. That’s not a measurement. It’s a conclusion the data can’t carry. Every “members versus non-members” chart sells this mistake - and I’ve made it myself in the past.
The “discount or relationship” dichotomy is lazy
A meta-analysis published this year in the Journal of Retailing worked through 434 effect-size estimates from 78 independent samples. Membership correlates positively with outcomes at every stage of the customer journey. Most strongly at the purchase stage itself. The effect doesn’t hinge on a simple choice between discount and relationship. It hinges on programme design and product category.
Gabel and Guhl compared points and individually targeted coupons at a large German grocery retailer. A redeemed points reward lifts the probability of a store visit by 5.0 percentage points, a coupon by 1.3. But customers redeem coupons on 18.5% of their trips, points on just 2.5%. Frequency and smaller discounts made coupons considerably more profitable. The authors’ conclusion: the two instruments complement each other.
A targeted coupon, after all, is personalisation - not its opposite. A discount that knows what you buy and a discount that doesn’t share a name. Their economics are entirely different.
A recipe, a wardrobe, and their limits
So what can a programme say to a customer a week after the purchase? At a grocery retailer, something like this. Friday’s bag holds zucchini, feta and pasta. On Sunday morning, a recipe built from those items arrives - with a targeted coupon for the one missing ingredient. The next step is an invitation to contribute a recipe of their own. The discount recipient becomes a co-author.
A fashion retailer holds an even stronger card: the virtual wardrobe. The customer’s account shows everything they’ve bought, with the option to add pieces themselves. The retailer knows at least the size and the style. SAP Engagement Cloud, which we work with at FrodX, can find similar products by comparing images. If a customer clearly loves a floral skirt, the system finds a matching piece in the new collection. For the trousers and sweater she bought three months ago, it suggests the shoes and the belt. The retailer becomes a personal stylist. And with virtual try-on, the customer gets a reason to visit that no percentage can buy. The same question awaits car brands: when did you last offer your owners something the service department can’t invoice?
None of this is free, though, or automatically good. A household’s shopping is not one person’s diet, so the recipe can simply be wrong. Baskets invite inferences about health and religion. The European Data Protection Board explicitly lists the purchase of certain foods among its examples. A virtual wardrobe, too, needs a clear purpose, transparency and restraint with data. A clothing size linked to an identifiable member is personal data. It just usually isn’t a special category. A message that shows the customer you know too much doesn’t build the relationship. It spends it.
What the winners actually do
Marriott Bonvoy is a perfectly ordinary points programme: points, free nights, member rates and status tiers. Its 271 million members generate 68% of all room nights across the chain. Points are clearly not the problem; recognition and late checkout build on them, they don’t replace them.
Bank of America took the same logic all the way. The reward isn’t a coupon but better terms across the whole relationship: lower interest rates, waived fees and higher card bonuses. The benefit level depends on the client’s total balances with the bank. The bank reports eleven million members and says around 30% of members add another product within the first 30 days. That second number is the more interesting one. It’s tied to an event, not a status.
And a lesson from a neighbouring industry, honestly fenced: ResMed gave sleep-apnea patients the myAir app, with a daily score, reminders and tips. In an observational study of 128,000 patients, 87% of app users stayed on therapy, against 70% of the rest. That is not brand loyalty; it’s adherence to prescribed therapy. The transferable hint: feedback on your own behaviour can sustain a habit even without a reward. Whether it works better than a reward, the study didn’t test.
The loyalty premium: a one-afternoon metric
Now for the metric the retailer from my opening should calculate - along with everyone who already runs a programme. Call it the loyalty premium. Take the realised margin rate on member revenue. Subtract the same rate on non-member revenue - after all discounts, redeemed points and reward costs. That’s the premium. Every POS and every ERP holds the data. A first approximation takes one afternoon.
The reading is unforgiving. A positive premium means members generate margin-friendlier revenue. On its own it doesn’t say why: willingness to pay more, a different basket, or the programme itself. A premium around zero means the programme pays for itself and earns its keep in data and frequency. A negative premium is an alarm with a number attached. Take an imaginary retailer: €20 million in revenue, 60% of it from members. Non-members leave €42 of margin per €100; members, after all the benefits, €38. The premium is minus 4 percentage points. The difference: €12,000,000 × 0.04 = €480,000 a year - against a hypothetical world where member revenue carries non-member margins. That is not yet the programme’s cost - without the programme, some of that revenue might not exist at all. It is, however, an alarm loud enough to finally measure the uplift with a control group.
The premium, then, is a diagnostic, not proof. The same goes for LTV: a lifetime value summed from revenue is a compliment. Summed from margin, it’s information. The most active members are often simply the best accountants of their own benefits. And then the programme hasn’t built loyalty - it has built a discount subscription.
Don’t blame the POS
A loyalty module on the POS is not a mistake in principle. For a retailer with one market and three stores, it’s often entirely sufficient. At FrodX we’re moving one such programme onto a separate engine right now. It runs across three markets for a single retailer. Every rule change used to mean three projects - one per market and its POS system. After the move, one will do. That’s an architectural decision. It follows from the number of markets and the pace of change, not from a belief that POS systems are bad.
The mistake is in the order of the questions. Start with “which module do we buy” and you get rules the module can execute. Start with “what will we tell the customer a week after the purchase, and what will our premium be” - and you get a programme worth measuring.
Next time an ad tries to convince you how beautifully a loyalty programme is growing, do what I did: open the accounts. Revenue is the compliment the programme writes for itself. The premium is the grade the customer writes.
