# Building loyalty programs that change shopping behavior
In 1994, a British grocer named Terry Leahy stood before the Tesco board and showed them what a small green card had revealed: their customers were not one crowd, but hundreds of distinct segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition →, each shopping for different reasons. One director reportedly said, "You know more about my customers in three months than I know in 30 years." That card was the Tesco Clubcard, and it helped Tesco overtake Sainsbury's to become the UK's largest retailer.
That is the promise of a loyalty program done right. Not a discount you hand to everyone, but a data engine that tells you who to talk to, what to offer, and how to move behavior.
Most people think loyalty programs exist to reward loyal customers. That is the wrong frame.
A loyalty program exists to change behavior you can measure. Specifically, two things:
A program that simply gives 5 percent off to people who were already going to buy is not a loyalty program. It is a margin leak. The goal is to shift the behavior of customers who would otherwise have shopped elsewhere, bought less often, or spent less.
Points and cards are the visible layer. The valuable layer underneath is identified purchase data: knowing that customer #48213 buys nappies every 12 days, switched from your milk to a competitor's last month, and always shops on Thursday evenings.
Tesco Clubcard and Kroger's Plus Card both work this way. The card links a scan at the till to a named household, building a longitudinal record of what that household buys over years.
Tesco famously partnered with an analytics firm, dunnhumby, to turn Clubcard data into segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition →. Kroger did the same, eventually taking majority ownership of dunnhumby's US operation and rebranding it 84.51°. Both use the data for two revenue streams:
1. Personalized marketing to shift customer behavior.
2. Retail media: selling insights and ad space back to the brands (CocaCocaCustomer Acquisition Cost: total sales and marketing spend divided by the number of new customers acquired over the same period.View full definition →-Cola, Unilever, and so on) that want to reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → specific shoppers. This is now a multibillion-dollar business for large grocers.
The lesson: your loyalty data can fund the very rewards you give away.
Points are the currency that gets customers to identify themselves at checkout. Without identification, you have anonymous transactions and no behavioral record.
Design points to encourage the behavior you want, not just spend. Examples:
A common rule of thumb is that a mature grocery program returns value worth roughly 1 percent of spend in points, though this varies widely and should be treated as an estimate, not a benchmark.
Tiers create aspiration and lock in your best customers. Airlines pioneered this (Silver, Gold, Platinum), and retailers borrowed it.
Tiers work when:
Sephora's Beauty Insider program is a widely studied example: its Rouge top tier drives disproportionate spend by combining early product access, events, and status, not merely a bigger percentage off.
Be careful: tiers can demotivate the majority who will never reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition → the top. Segment first. If most of your customers are occasional shoppers, a tier ladder may be the wrong tool.
This is where behavior actually changes. Instead of a blanket offer, you send a specific reward to a specific customer to prompt a specific action.
Four high-value targeting plays:
Tesco's Clubcard mailings were legendary for this: households received coupon booklets tailored to their actual buying patterns, so a curry-loving household got curry-adjacent offers, not random ones.
🎬 [VIDEO: "How Tesco Clubcard Changed Retail Forever" — youtube.com — a short business documentary on how Clubcard data reshaped Tesco's strategy and the wider grocery sector]
Here is the discipline most programs skip. To prove a program changed behavior, you need a control group: customers who did not receive the offer, so you can compare.
The core metric is incrementality: the extra sales caused by the program that would not have happened otherwise. If you give a $5 reward and 90 percent of redeemers would have bought anyway, your incremental return is terrible even if redemption looks high.
A simple incrementality test:
Test group: receives targeted offer
Control group: matched customers, no offer
Incremental lift = (Test group avg spend) - (Control group avg spend)
ROI = (Incremental margin) - (Reward cost) / (Reward cost)Track these over a rolling window:
If you cannot measure incrementality, assume you are subsidizing existing behavior.
Knowledge check
1. According to the lesson, what is the primary purpose of a loyalty program?
2. A grocer gives every shopper 5 percent off, and most of those shoppers were already going to buy. Why does the lesson call this a 'margin leak' rather than a loyalty program?
3. A shopper spends $500 a month on groceries, of which $300 goes to your store. What does the roughly 60 percent figure represent, and why does it matter?
4. Select ALL correct answers. According to the lesson, why is identified purchase data considered the real asset of a loyalty program rather than the points or cards?
Select all the correct answers.
5. Select ALL correct answers. Which customer behaviors would a well-designed loyalty program aim to shift?
Select all the correct answers.
Discounting sales you already had. The single biggest trap. If your offers mostly go to loyal, high-frequency customers, you are cutting margin on guaranteed revenue. Target the wavering and the lapsing instead.
Overloading the customer. Complex point math and confusing tiers reduce participation. The best programs are simple to understand and rich under the hood.
Ignoring privacy expectations. Loyalty programs run on personal data, and rules like the EU's GDPR (General Data Protection Regulation, the law governing how personal data is collected and used) and various US state privacy laws require clear consent and give customers rights over their data. Personalization that feels intrusive backfires. The UK Information Commissioner's Office publishes a plain-language guide to the data protection principles worth reading before you design data capture.
Copying tiers without segmenting. Sephora's model works because its customers aspire to status. A discount grocer's customers may just want lower prices. Match the mechanic to the motivation.
No exit from points liability. Unredeemed points are a balance-sheet liability. A program that issues points faster than customers redeem them stores up a cost. Design breakage (unredeemed points) expectations deliberately, not by accident.
1. Define the behavior you want to change (frequency, basket size, category penetration). Pick one primary goal.
2. Capture identity at checkout with a simple, valuable reason to enroll.
3. Segment by behavior, not demographics. Recency, frequency, and monetary value (the classic RFM model) is a strong starting point.
4. Design targeted offers for the segmentssegmentsDividing a market into distinct groups of customers who share similar needs, characteristics or behaviours, so each group can be served with a tailored approach.View full definition → that can actually move: lapsing, cross-category, trade-up.
5. Always test against a control group. No control, no proof.
6. Fund it partly through retail media, selling shopper insights to your suppliers.