Building loyalty programs that change shopping behavior

# Building loyalty programs that change shopping behavior

In 1994, Tesco's board sat through a read-out of a Clubcard trial analysed by a then tiny consultancy, dunnhumby. The chairman, Ian MacLaurin, reportedly told them: "What scares me about this is that you know more about my customers after three months than I know after 30 years." Clubcard went national in February 1995, and within a year Tesco had passed Sainsbury's to become Britain's largest grocer.

The card did not win because 1 percent back is exciting. It won because it turned anonymous till scans into named households whose behaviour could be nudged, one segment at a time. That mechanism, earn, tiers, identity and the offers they make possible, is what this lesson is about.

The two behaviours a program is paid to move

• Frequency: how often a customer shops with you. A shopper who comes four times a month instead of three is a 33 percent uplift with no change at all in what they put in the trolley.
• Share of wallet: the percentage of a customer's category spend that comes to you rather than a competitor. If a household spends $400 a month on groceries and $250 lands with you, your share is roughly 62 percent.

Frequency is usually the easier of the two to move, because a trip has a purpose and there is a ceiling on how much of a weekly shop you can enlarge in one visit (the lever the traffic and basket lesson sets up). Share of wallet moves through category penetration: getting the household that buys your groceries but never your wine to try the wine.

A program that hands 5 percent to people who were already coming is a margin leak dressed as marketing. The target is the wavering shopper, not the devoted one.

The asset is identity, not points

Points are the price you pay for identification. The valuable layer underneath is the record: customer #48213 buys nappies every 12 days, switched off your milk last month, and shops Thursday evenings.

Two things break that record more often than people expect.

Loyalty accounts are usually households, not people. A couple sharing one Clubcard looks like a single erratic shopper with contradictory tastes, and your trade-up offers land with whichever of them did not choose the product. Grocery programs live with this; specialty retailers with per-person accounts do not, which is one reason Sephora's targeting can be sharper than a supermarket's on the same volume of data.

The second is till-side pollution. Where staff are measured on scan rate, cashiers scan their own card for customers who have none. A handful of employee cards then carry thousands of unrelated baskets, and every segment model built on them is quietly wrong. Audit for accounts with implausible visit counts before you trust any segmentationsegmentationDividing 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 →.

Online, identity comes free with a login. In store it does not, and joining the two is its own project: customer data platforms such as Segment (a vendor that sells exactly this stitching) exist to resolve a web session, an app login and a till scan into one profile. Buy or build, but budget for it, because unstitched channels make every frequency number look worse than reality.

The same identified data can be sold back to suppliers as media inventory. That P&L belongs to the retail media lesson; here it matters only as the reason a loyalty budget can be part self-funding.

Four mechanics, and what each one actually does

Earn

Design the earn rate to buy behaviour, not to thank spend:

• Bonus points weighted to a category you want to grow, if fresh produce penetration is your weak point.
• Threshold mechanics ("double points once you spend $60 this week") to pull the basket up.
• Time-based boosts for slow trading periods: Tuesday mornings, the second week of January.

A mature grocery program tends to return value worth around 1 percent of spend, but treat that as an order of magnitude rather than a benchmark.

Member pricing

Tesco's Clubcard Prices, rolled out in the early 2020s, replaced deferred points with a lower shelf price for members. Enrolment and scan rates jump, because the reward is immediate and visible at the fixture. Two costs come with it. The discount is real margin the day it is taken, with none of the escape valve that unredeemed points provide (the depth question itself belongs to the markdown lesson). And non-members now read a higher shelf price, which can damage price perception among exactly the occasional shoppers you were trying to recruit. The UK's Competition and Markets Authority reviewed supermarket loyalty pricing in 2024 and concluded that in the large majority of cases the member price was a genuine saving, which settled the question of whether the mechanic is legitimate, not whether it is profitable for you.

Tiers

Sephora's Beauty Insider is the reference case: free entry, VIB at $350 of annual spend, Rouge at $1,000. Rouge holds because the benefits are access and status (early product drops, events, gifting) rather than a bigger percentage off. Discount-only tiers train customers to wait for the discount.

Watch the qualification window. Tier resets create December stockpiling from customers a few hundred dollars short, which reads as a spend spike and is mostly purchases pulled forward from January. Strip requalification-driven volume out before you claim the ladder worked.

Tiers also demotivate the majority who will never climb them. If most of your base shops occasionally and on price, the ladder is the wrong tool.

Targeted rewards

Four plays worth the effort:

• Winback for a customer whose frequency has dropped but who has not gone yet.
• Cross-category, to buy a first purchase in a category they have never touched with you.
• Trade-up, from value line to premium.
• Habit reinforcement: give the Thursday shopper a Thursday reason.

Clubcard's coupon booklets were built this way. A household that bought curry ingredients got curry-adjacent offers, and the redemption rates were multiples of untargeted mail.

🎬 [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]

Measuring whether it actually worked

To prove a program changed behaviour you need a control group: matched customers who did not get the offer.

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)

The most common methodological error is using non-members as the control. Members are self-selected heavy shoppers, so that comparison flatters every program ever run. Hold out a random slice of enrolled members instead, and keep holding it out permanently, even when finance asks why you are leaving revenue on the table.

Track redemption rate (low rates mean weak relevance), frequency change test versus control, category penetration as a proxy for share of wallet, and lapse rate. High redemption with zero incremental lift is the signature failure: the offer was popular precisely because it went to people already on their way to the till. Feed the incremental margin, not the gross, into the lifetime valuelifetime valueLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → calculation the LTVLTVLifetime Value: the total revenue (or profit) a customer generates throughout their entire relationship with your business.View full definition → lesson sets out.

Common ways loyalty programs fail

Discounting sales you already had. If most rewards flow to your top decile, you are cutting margin on guaranteed revenue.

Overloading the customer. Complicated point maths suppresses participation. Simple on the front, rich underneath.

No plan for points liability. Under IFRS 15, points issued are a separate performance obligation: revenue sits deferred on the balance sheet until they are redeemed or expire. Issue faster than customers burn and you build a liability whose release date you do not control. Breakage should be an assumption you set and monitor, not a surprise. The correction, a sudden devaluation of the earn rate, is the single most reliable way to make loyal members feel cheated.

Copying tiers without segmenting. Sephora's ladder works because its customers want status. A discount grocer's customers want a lower price, and will read a tier as a hurdle.

Underestimating the ratchet. Member pricing is easy to launch and painful to withdraw: taking it away reads as a price rise to your best customers.

Consent treated as paperwork. Personalisation that feels intrusive backfires, and the comms rules are covered in the fair treatment lesson. Before you design data capture, read the ICO's plain-language guide to the data protection principles.

A practical design sequence

1. Define the behaviour you want to change. Pick one primary goal.

2. Capture identity at checkout with a reason to enrol that is worth the friction.

3. Segment by behaviour, not demographics. Recency, frequency and monetary value (RFM) is a strong start.

4. Design offers for the segments that can move: lapsing, cross-category, trade-up.

5. Hold out a control group permanently. No control, no proof.

6. Fund part of the reward pool from the media value of the data you are collecting.

Key takeaways

• The job is to move frequency and share of wallet, not to thank people for purchases they were making anyway.
• Points buy identification. The identified record is the asset, and it is only as good as your household resolution and your till-side data hygiene.
• Earn rates steer category mix, member pricing buys immediate scan rate at immediate margin cost, tiers work only where status motivates, and targeted offers do the actual behaviour change.
• Measure against a held-out group of members, never against non-members.
• Points are a liability with an accounting treatment. Set breakage deliberately, and never fix a runaway earn rate with a surprise devaluation.