Loyalty without discounting
Most loyalty programmes are a discount with extra steps and a dashboard. The lever that actually compounds is cheaper, and it starts with not asking anyone to sign up.
If you strip the branding off most points programmes, what remains is: give up some margin, in arrears, to people who were going to buy anyway, in exchange for an email address you could have got another way. That can still be a good trade. It's just rarely the trade the deck said it was, and it's almost never the highest-return thing available.
What loyalty is actually buying
Start with the effect that's real and well-measured. Contentsquare's benchmark, across 99 billion sessions and more than 6,000 sites, finds returning visitors convert at 2.9% against 1.7% for new ones — and that repeat visits make up 52.8% of all traffic.[2]
- 2.9%Conversion rate, returning visitors[2]
- 1.7%Conversion rate, new visitors[2]
- 52.8%Share of traffic that is a repeat visit[2]
Note carefully what that does and doesn't say. It says familiarity converts. It does not say points cause familiarity. It doesn't even isolate why. A returning visitor already trusts you and already knows the product — and some meaningful share of them also already have an account, an address on file and a stored payment method. Contentsquare counts a returning visitor at the session level, not the purchase level, so we can't say from this figure how big that share is. But it is part of the "loyalty premium", and the part that is saved keystrokes rather than sentiment is the part you can hand to a stranger.
That reframing has a practical consequence. If a meaningful part of what makes repeat customers valuable is that buying again is easy for them, then the cheapest way to manufacture loyalty economics is to make the second purchase easy for everybody — including the people who never enrolled in anything.
A returning customer converts better partly because they don't have to type their address again. You can give that to a first-time buyer for free.
The three jobs a points programme does
Be honest about which one you're actually buying, because they have very different economics.
| Job | What it really is | Honest assessment |
|---|---|---|
| Price discrimination | A discount, delivered later, to people who cared enough to join | Backwards. You are systematically discounting to your least price-sensitive customers — the enthusiasts who joined. This is the job most programmes are unknowingly buying. |
| A switching cost | A balance the customer would forfeit by leaving | Genuinely works, but only above a threshold. A balance worth £3 deters nobody. This is why airline and coffee schemes work and most CPG schemes don't. |
| An identity pretext | A reason for the customer to tell you who they are | The most valuable job, and the one you can now get far more cheaply by other means — which is most of the argument in this piece. |
The third job is why loyalty programmes exist at all for brands that sell through retail. If a supermarket owns the transaction, a points scheme is often the only mechanism a brand has for learning that a specific human bought its product. That's a real need. But paying for it in permanent margin is an expensive way to solve an identity problem, especially now that a code on the pack can solve the same problem without touching price.
Why your attach rate is so low
Ask what a customer has to do before they earn anything. In most programmes the honest list is: notice the programme exists, choose to join, enter an email, choose a password, leave the site to verify it, come back, and only then start accruing value they can't spend yet.
Every one of those is a place people leave. It's the same failure the checkout has — among shoppers who abandoned during checkout, 18% cite required account creation as a reason and 17% cite a checkout that was too long, with multiple answers permitted[1] — except a loyalty sign-up has none of the urgency of a purchase to carry it through. If a required account is one of the leading things people name when they walk away from a purchase they actively wanted to make, consider what it costs you when the reward on offer is theoretical points.
The inversion
Stop asking people to enrol before they can earn. Credit the value at the moment of purchase, to whoever just bought, with no account. Tell them it's there. Ask for the account later, when they come back to spend a balance that already exists. The enrolment step doesn't disappear — it moves to the point where it has obvious value to the customer instead of obvious value to you.
What earn-first looks like
-
Purchase completes with a wallet, no account
The wallet has already supplied a name, an email and a shipping address — the address and card are verified by the payment network; the email is a field the customer sets themselves, so treat it as supplied, not proven. You have an identity, and you don't have a password to store, which you never wanted anyway. That identity is a side effect of one-tap checkout, not something you asked the customer to fill in.
-
Value is credited immediately and shown
On the confirmation screen, before they've left. Not "you could have earned"; "you have earned". The difference between those two sentences is most of the programme.
-
You get one short window to say something useful
If the purchase happened in an App Clip, you can schedule notifications for up to 8 hours after each launch without a permission prompt, via the
NSAppClipRequestEphemeralUserNotificationkey — and the customer can switch it off on the App Clip card.[6] Eight hours is short. That's a design constraint worth respecting rather than fighting: it's enough for "on its way" and "here's what you earned", and not enough to become a nuisance. -
The balance is what brings them back
The reorder prompt isn't "buy more". It's "you have £4 sitting here". A balance already earned is psychologically a possession, and people are markedly more reluctant to abandon a possession than to forgo a discount.
-
Ask for the account at the moment of redemption
Now it's obviously in their interest, they've already transacted with you, and the ask converts several times better than it would have at step one.
Mechanics, ranked by what they cost you
Not all rewards damage the brand equally. Roughly in order of how much they teach a customer that your list price is a fiction:
| Mechanic | Margin cost | What it teaches the customer |
|---|---|---|
| Standing % off for members | High, permanent | The real price is lower. Never buy at full price. |
| Points redeemable as cash off | Moderate, deferred | Same lesson, delayed. Plus a liability on your balance sheet. |
| Free product at your marginal cost | Low | The product is worth having. Costs you COGS, not price integrity. |
| Early or exclusive access | Near zero | Being a customer has status. Works best where supply is genuinely limited. |
| Free or upgraded delivery | Low to moderate | Attacks the single biggest abandonment cause — unexpected cost, cited by 40%[1] — without touching the price on the shelf. |
| Unexpected gift with a repeat order | Low, controllable | Nothing, because it's a surprise. Unpredictable rewards don't set expectations. |
The last row is underused and it's the one we'd start with. A surprise can't be gamed, doesn't create a liability, doesn't get compared on a deal site, and — because it isn't promised — doesn't devalue the price when you stop.
The strongest objection
"Our LTV modelling says the points programme pays for itself." It might. But almost every model of this kind has the same flaw, and it's worth checking yours for it: members are compared to non-members. Members self-selected. They joined because they were already your best customers, so of course their lifetime value is higher — and attributing that gap to the programme is the same error as attributing an app's conversion rate to the app.
The only clean read is a holdout: withhold the programme from a random slice of otherwise-identical customers and compare. Most brands never run one, because it feels like leaving money on the table and because nobody wants the answer. If yours has never been tested that way, the honest position is that you don't know what it's worth — which is different from knowing it doesn't work. Designing that holdout, and reading it honestly, is a piece of its own.
The counter-objection to us: earn-first has a real weakness. Without an account, identity is anchored to whatever the wallet gave you, and a customer who buys on one device with one email and returns on another may not be recognised. Programmes with a hard sign-up wall don't have that problem. Our view is that a slightly leaky identity across a much larger population beats a perfect identity across a tiny one — but it is a trade, and if your category has genuinely high-value, low-frequency purchases, the maths can go the other way.
What to do about it
- Compute your attach rate honestly: enrolled members as a share of customers, not of email subscribers. If it's in single digits, enrolment friction is your programme, not your rewards.
- Count the steps between purchase and first earned value. Anything above zero is a design choice you should be able to justify.
- Move one mechanic down the table — swap a standing percentage for free delivery or a surprise gift on the second order, and watch what happens to full-price sales, not just to redemption.
- Run a holdout, even a small one. It's the only way you'll ever know, and it's much easier to start before the programme is politically load-bearing.
- Ask what identity is worth to you separately from what discounting is worth. If the programme exists mostly to learn who your retail customers are, there are cheaper ways to buy that now.
Sources
- Baymard Institute, Cart Abandonment Rate Statistics — among shoppers who abandoned during checkout, 40% cite unexpected extra costs, 18% required account creation and 17% an over-long checkout; the survey excludes the "just browsing" segment and permits multiple answers, so these are reason-weights among abandoners, not shares of all buyers. Primary ↩
- Contentsquare, 2026 Digital Experience Benchmark — conversions — 99bn sessions, 6,000+ sites. Primary ↩
- Apple, Enabling notifications in App Clips — the 8-hour ephemeral notification window and how users disable it. Primary ↩