Loyalty programs people don't resent

Most loyalty programs fail in a quietly expensive way. They launch with a burst of sign-ups — because "join and get 10% off" would sign anyone — and then settle into their real function: an automated discount machine for customers who were going to buy anyway, funded by margin, reported on with vanity metrics like "member count". The program doesn't create loyalty; it taxes revenue from the already-loyal and calls it retention.

A loyalty program worth building starts from a harder premise: loyalty is behaviour change you can measure, not enrolment you can count. Everything below follows from that.

Start with the behaviour, not the rewards

Before any points mechanic exists, answer one question with numbers: which customer behaviour, increased by what amount, pays for this program? The candidates are few:

Frequency — the classic. Moving a twice-a-year customer to three times a year compounds beautifully.
Second-purchase rate — for most stores, the highest-value gap in the entire business is between first and second order. A program aimed here can be startlingly cheap because the prize is so large.
Category breadth — customers who buy across two or three categories churn at a fraction of single-category buyers.
Advocacy — referrals and reviews, where genuine.

Pick one primary behaviour. Programs that try to reward everything reward nothing legibly. For the Hearthbrew subscription club, the target was explicit: convert occasional buyers into subscribers and keep subscribers past month three — which shaped everything from the reward cadence to the skip-and-swap mechanics. Points had to serve that behaviour, not decorate an account page.

Points design: clarity is the mechanic

The single most common loyalty UX failure is that customers cannot answer "what do I get and when?" If the answer requires a calculator, the program is an accounting product, not a loyalty product.

The earn rate stays visible at the point of decision. Points earned this order shown in the cart and on the PDP — not buried in a dashboard. "This roast earns you 48 points" does more behavioural work than a balance page, because it attaches the program to the moment of desire.

One currency, one metaphor. Points that convert to dollars at an unstable or unexplained rate breed low-grade resentment. Fixed, memorable maths — 100 points = $5 — or skip points entirely and use punch-card mechanics (buy eight, the ninth is on us), which remain the most legible loyalty design ever shipped.

Expiry policy is where programs confess their values. Points that expire in 90 days are a pressure device wearing a reward costume; customers learn it and recall it as a small betrayal at exactly the moment you wanted warmth. Long or no expiry, generous warnings before anything lapses, and never a silent forfeiture.

The balance is everywhere or nowhere. A points balance should follow the customer — header, cart, checkout, emails — or don't build the program. A balance only visible in an account sub-page is a program that's already over.

Tiers without the trap mechanics

Tiers work because status is motivating; they fail when the motivation tips into anxiety. The honest version:

Progress is always visible and fairly framed. "You're $40 from Gold" is motivation. "You're $960 from Platinum" — technically the same design, psychologically a taunt — is how programs teach customers the tier isn't for them. Chunk the gap into attainable-feeling steps, and celebrate the current tier's benefits genuinely rather than holding them hostage to the next one.

Status is earned, never teased. Showing a customer the Gold lounge they can't enter is an airline move, and even airlines are quietly regretting it. Tier benefits you don't qualify for shouldn't be on your page as a locked-preview catalogue of exclusion. Describe them; don't wave the key under the customer's nose.

Devaluation is a breach of contract, treat it that way. Every program eventually faces the margin meeting where someone proposes making points worth less. If you must devalue, do it with notice, grandfathering, and an honest explanation — the same candour standard as subscription price rises. A silent devaluation converts your most engaged customers into your most credible critics, and they will do the maths, in public, with screenshots.

The top tier should be short on maths and long on meaning. The benefits whales actually value are access and recognition — early releases, a human who knows their name, input into what's next — which are cheaper than discounts and can't be comparison-shopped. A top tier that is merely "more cashback" has no defence against a competitor offering more cashback.

Discovery UX: the program nobody finds can't work

Program enrolment UX is usually a checkout checkbox and a buried footer link. The touchpoints that actually grow membership with the right members:

Post-purchase, not pre-purchase. The moment after order one — confirmation page and email — is the honest high-intent moment: "You just earned 62 points. Want them?" Enrolment that retroactively credits the purchase just made converts better and starts the relationship generously instead of transactionally.
Order tracking and delivery emails. The highest-open-rate messages a store sends. A points balance line in the delivery email costs nothing and reminds at a warm moment.
The return flow. A customer mid-return offered bonus credit instead of refund — clearly optional, refund untouched, as discussed in our returns UX piece — is a loyalty-building offer, not a dark pattern, because the alternative remains fully available.
In-box and packaging. Physical touchpoints for physical goods: the card in the parcel is read at the highest-attention moment of the entire relationship.

Checkout enrolment checkboxes still exist, but pre-ticked boxes and forced account creation are exactly the resentments that make people hate programs. Opt-in, plainly worded, skippable without guilt-tripping copy.

Prove it or kill it: measuring real behaviour change

Here's where most programs get away with murder, because measurement is hard and "member count" is easy. The honest dashboard:

Incrementality, not gross member revenue. Members will always outspend non-members — high-intent customers self-select into programs. That comparison is worthless. What you need is a difference-in-differences view: did members' behaviour change more after joining than a matched cohort of similar non-members changed over the same period? If your tools can't do this, build holdouts: a slice of eligible customers who aren't shown the program, tracked quietly for two quarters. The same discipline we bring to CRO experimentation applies — pre-registered kill criteria included.

Second-purchase rate and time-to-second-order. If this doesn't move for members, the program is a discount scheme; rename it accordingly and check whether the discount would be cheaper as a plain price cut.

Program P&L, in the open. Reward costs, platform fees, and the discount cannibalisation estimate on one side; incremental margin on the other, with the attribution method labelled. If it's not clearly positive after a year, say so and fix or fold it. The programs that survive honest accounting are a minority, and they're worth five times the others.

Churn signals. Redemption rate is a health metric: members who never redeem are accruing a liability and receiving no emotional payoff — the worst of both sides. Prompt redemption; a redeemed reward creates a visit, and visits create orders.

When not to build one

Honesty requires the negative case. Skip the program when: purchase frequency is naturally annual (a mattress store punch card is a joke the customer is in on); margins can't fund real rewards without devaluation tricks; or the product experience still has fixable friction — a loyalty program layered over a painful checkout or an unreliable delivery promise is lipstick on operational debt. Fix the fundamentals; loyalty is a multiplier of an experience already worth repeating.

Key takeaways

Define the one behaviour the program exists to change — usually frequency or second-purchase rate — and let it shape every mechanic.
Keep the maths legible: visible earn rates at decision moments, one stable currency, generous expiry, balance everywhere.
Tiers motivate with visible attainable progress, earned-not-teased status, and top-tier benefits built on access and meaning, not more cashback.
Grow membership at warm moments — post-purchase, delivery emails, returns, packaging — always opt-in, never pre-ticked.
Measure incrementally with holdouts and cohorts, publish the program P&L internally, and be willing to fold a program that fails its own audit.

FAQ

Points or tiers or subscriptions — which model? They solve different problems. Points reward accumulated behaviour; tiers reward status level; paid memberships (subscribe-and-save, Prime-alikes) monetise commitment directly. Many good programs layer two; few survive all three without confusing their customers.

How generous should rewards be? Model it backwards from target behaviour-cost: what is moving a customer to one extra order a year worth in margin? Fund the program at a sensible fraction of that, and keep the earn rate honest — 2–5% effective value is the range where maths stays legible and finance stays calm.

When should we launch a program — at what size? When you have enough repeat purchase to observe the behaviour you want to improve, and enough order volume to measure change against a holdout. Below that, invest in retention fundamentals: post-purchase emails, service quality, and product breadth.

Should loyalty points integrate with referrals and reviews? Carefully. Rewarding referrals works; rewarding reviews contaminates review honesty and can breach platform guidelines. If you must incentivise reviews, reward the act, never the stars.

What's the fastest loyalty win without building a full program? The post-purchase retroactive enrolment email — "your order earned you credit" — pointed at store credit rather than a formal program. It tests the appetite for the mechanic at a hundredth of the build cost.

Retention economics run through every e-commerce engagement we take on — programs included, illusions optional.