Designing a D2C loyalty programme that is not a discount
Most loyalty programmes reward people who were going to buy anyway. That is not loyalty, it is a margin transfer with a points balance attached.
- If the reward goes to behaviour the customer already had, you have cut your margin and changed nothing.
- Points are a liability the moment they are issued. Model breakage and the cost of redemption before launch.
- Reward the behaviour you actually want: frequency, basket size, or the second order that predicts retention.
- Non-discount rewards often work better and cost less: early access, samples, service, and status.
Loyalty programmes are easy to launch and hard to design. The machinery is available off the shelf, so the question is rarely how to run one. It is whether the one you are running changes anything.
The uncomfortable test: if you switched it off tomorrow, would your customers buy less. For a large share of programmes the honest answer is no, which means the programme is a margin transfer with a points balance attached.
The core mistake
Most programmes reward volume. Spend more, earn more, redeem later.
That sounds obviously correct and it has a flaw. Your highest volume customers were already your highest volume customers. Giving them points for behaviour they already had reduces your margin on exactly the cohort where margin was healthiest, and changes nothing about what they do.
Meanwhile the customer you actually want to influence, the one who bought once and has not returned, earns almost nothing because they have almost no volume. The programme pays the people who need no persuading and ignores the people who do.
Reward the turning point instead
Look at your own cohort data and find where customers are lost. In most consumer categories the answer is between the first and second purchase. A buyer who reaches a third order behaves very differently from one who stopped at one.
That gap is where a reward changes an outcome rather than subsidising one. A meaningful incentive attached to the second order, delivered soon enough after the first to still be relevant, is doing real work.
Compare that with a tiered programme where the benefit arrives after the sixth order. By then the customer has already demonstrated they are staying, and you are paying for information you already had.
Points are a liability from the moment you issue them
This gets treated as an accounting technicality and it is a commercial one.
Every point issued is a promise of future value. It sits as an obligation until it is redeemed or expires. Brands that only recognise the cost at redemption are understating what they owe, and a growing balance can turn into an unpleasant quarter when a large tranche is claimed at once.
Two things follow. Estimate breakage, the share never redeemed, conservatively rather than optimistically, and revisit it against real behaviour each quarter. And set expiry rules that are clear and fair at issuance, because retrospective changes to a points balance damage exactly the trust the programme was built to create.
The rewards that are not discounts
The strongest programmes lean on things a competitor cannot replicate with a price cut.
Early access to a launch, which costs almost nothing and makes a customer feel like an insider. Samples of an adjacent product, which doubles as merchandising for a line they have not tried. Faster or free shipping, which removes friction rather than reducing price. A named contact or priority support when something goes wrong, which matters more than any discount at the moment it is needed. And visible status, which is free and surprisingly effective in categories with any social component.
These share a useful property: none of them trains the customer to wait for a better price. Discount-based programmes quietly teach people that paying full price is for the uninformed, which is an expensive lesson to have taught once you want to stop.
Decide what you are actually optimising
Before designing anything, name the metric. More frequent orders, larger baskets, longer retention, or higher margin mix. These pull in different directions and a programme that tries to serve all four serves none.
If the answer is frequency, reward the interval. If it is basket size, reward the threshold. If it is retention, reward the second order and the anniversary. If it is margin mix, reward buying the lines you want to sell rather than the ones already selling themselves.
When not to bother
If your product is genuinely bought once every few years, a loyalty programme is machinery with no work to do. The equivalent effort spent on referral mechanics and on a post purchase experience worth talking about will return more, because in that category your growth comes from what customers say rather than what they repeat.