“How will I know the programme makes me sell more? And not just that I’ve moved shoppers from my partners’ shelves to my own online store, where my margin is higher?” The man asking runs a distributor that represents several global beauty brands in the region and sells them through drugstores, pharmacies, retail chains, salons and hairdressers, as well as an online store of its own. I had come to sell him a loyalty programme. He let me finish. Then came the part that mattered: “Because that’s when a partner finds another brand for the shelf.”
I had a programme ready. I didn’t have an answer. Those aren’t the same thing. We agreed I’d come back once I could tell him two things: which customer behaviour the programme would change, and how I’d show it. Put that question to most loyalty programmes and few survive it.
The problem isn’t beauty. It’s how we measure loyalty.
The metric that grows while the company dies
Nearly every programme reports the same number: the share of sales that comes from members. Sixty percent, seventy. The board nods. It’s a useful number - it tells you how many customers you’ve identified and how far the programme reaches. It tells you nothing about what the programme caused.
The US retailer Sears is the expensive proof. In 2012, members of its Shop Your Way programme generated more than half the revenue of Sears Domestic and Kmart. By spring 2014 it was 74% of sales. The programme grew. Management called it proof the strategy was working. Revenue, meanwhile, fell from more than $53 billion in 2006 to $3.3 billion in 2020, with a bankruptcy along the way. CEO Eddie Lampert admitted that running two promotional models at once, traditional promotions and points side by side, was eating into margins.
The points were a second discount stacked on the first. A discount shifts when people buy; it rarely changes how they buy. And the member share kept climbing for the simplest of reasons: the company was shrinking around its members. Whoever was still there was a member. That isn’t a loyalty metric. It’s a survival metric.
What the programme is actually paying for
The other favourite number: members spend twice as much as non-members. True. And just as silent on what the programme did.
The first people to sign up are the ones already buying a lot. Yuping Liu followed a retail chain’s customers for several years to see who actually changed after joining. Heavy buyers redeemed the most rewards and changed nothing about how they bought. Light and moderate buyers gradually bought more, and started buying other categories too.
The uncomfortable part: the more a customer already buys, the more likely you are paying for purchases that would have happened anyway. The untapped potential often sits with the light and moderate buyers - the segment nobody looks at in meetings, because individually it’s small.
Some numbers. Say you sell €10 million a year and members account for €6 million. Points cost 2% of that; running the programme costs another €60,000. Total: €180,000 a year. At a 40% contribution margin, the programme has to generate around €450,000 in sales that wouldn’t have happened without it - a little more in practice, since the extra sales carry their own point costs. That’s just under 8% more than the same people would have bought with no programme at all. Not 8% more sales from members. You get that number simply by signing up the people who already buy.
So the problem isn’t rewarding loyalty. The problem is rewarding without first deciding which behaviour is supposed to change.
Design changes behaviour
The neatest demonstration that design alone can shift behaviour came from a car wash. Nunes and Drèze handed out 300 loyalty cards. Half had eight empty stamp slots. The other half had ten slots, two already stamped. Same effort, same reward. Over nine months, 19% of customers completed the first card. For the second, it was 34%.
The experiment doesn’t say whether the car wash made more money. It does say that a programme has to know which behaviour it’s buying before it issues its first point. A first purchase that otherwise wouldn’t happen. A repeat purchase that comes sooner. A second category from the same customer. A customer who would otherwise leave. Four goals, four different ways of measuring. A programme that doesn’t know which one it’s pursuing pays for all four and proves none.
What I’ll propose to the distributor
The distributor’s real problem is bigger than cannibalisation. It doesn’t know who uses its brands. It knows the partners who send it orders. The shopper who picks a serum off a drugstore shelf is invisible to it. A loyalty programme is the first way that shopper gets a name.
I’ll propose three hypotheses, not a solution.
First: points are earned everywhere, on equal terms. No channel gets an advantage. Since the distributor can’t plug into its partners’ checkout systems, the product itself carries the proof of purchase. The local-language label it already applies to every unit in its warehouse gets a unique code under a peel-off layer. The customer scans it at home when opening the product. The code tells you which shipment the unit came from and roughly when the customer started using it. It doesn’t tell you everything: engaged customers scan more than others, and the shipment isn’t always the point of sale. A signal, not a census.
Second: rewards are redeemed at partners and paid for by the distributor. The programme brings partners customers rather than taking them away. A salon gets a share of the take-home purchases recorded for the customers it enrolled. Open questions remain: how long a salon should be credited, and whether vouchers simply shuffle shoppers from one partner to another.
Third: the programme changes behaviour, and that shows up against a group that didn’t get it. Randomly selected, comparable groups of salons or markets launch first; the rest wait. Among enrolled customers, some get no reminders - which measures the reminders, not the programme. After a period fixed in advance and long enough for at least one normal repurchase, we compare contribution margin after all programme costs, not revenue. Sales to partners must not fall relative to the groups still waiting, and incremental units are valued at the average margin across all channels. That takes channel shift out of the calculation entirely.
Until a pilot shows this, all three remain promises. I have to hold myself to the standard I’m demanding of everyone else here. That’s what I’ll tell the director. Not a solution. An experiment.
Two questions for your programme
If you already run a loyalty programme, test it with two questions. Which behaviour is it buying? And which comparable group never got it?
If there’s no answer, you know how much your members buy. You don’t know how much of that the programme produced.
Share of sales from members isn’t proof. A loyalty programme is proven by the customers who never got it.
igor.pauletic@frodx.com
P.S. Sears didn’t collapse because of its loyalty programme. It collapsed because of everything else. The programme just made sure the collapse looked like success right up to the end.
