The loyalty programme playbook
Loyalty programmes fail from the wrong first question. It is not "points or tiers?" — it is "what repeat behaviour is realistic for what we sell, and what would genuinely change it?"
A programme that fits a coffee brand is nonsense for a mattress brand: purchase frequency, margin, and the customer’s natural cycle decide what loyalty can even mean. This playbook starts from those economics, chooses the lightest mechanism that changes real behaviour — sometimes points, often something simpler — and measures the only thing that matters: incremental repeat behaviour against a holdout, not enrolment counts and not redemption theatre.
Stage 1: Let the economics choose the mechanism
From your own data: natural repeat rate and interval per category, margin headroom for rewards, and where the repeat curve breaks (the second-purchase cliff is nearly universal). Then match: frequent-purchase catalogues can support points; considered-purchase catalogues do better with perks, access, and service benefits; the second-purchase cliff often needs only a well-timed post-first-purchase benefit, not a programme at all.
Repeat rate, interval, and margin headroom computed per category
The repeat curve’s break point identified — usually purchase two
Mechanism matched to frequency: points for often, perks and access for seldom
The null option taken seriously: a timed benefit may beat a programme
Stage 2: Design for the behaviour, against the spreadsheet
Rewards priced against margin with honest redemption assumptions — a programme that only works if nobody redeems is a liability, not loyalty. Earning legible ("spend X, get Y" beats point exchange rates needing a calculator), rewards reachable within a realistic cycle (a horizon nobody reaches de-motivates below zero), and tiers only where the gap between levels changes actual behaviour rather than decorating receipts.
Stage 3: Measure incrementality or measure nothing
The programme’s question is: do members repeat more than they would have anyway? Enrolment-versus-non-enrolment comparisons flatter — loyal customers enrol. A holdout of eligible customers, or at minimum cohort repeat curves before and after, isolates the real effect. Judge on incremental repeat revenue against reward cost, review quarterly, and be willing to simplify: the best loyalty change is sometimes removing a mechanism nobody valued.
How it works
Compute the economics
Repeat rates, intervals, margin headroom, and the break point, per category.
Design the lightest mechanism
Matched to frequency, legible to customers, priced honestly against margin.
Measure against holdout
Incremental repeat behaviour isolated, reviewed quarterly, simplified when the data says so.
What you get
A mechanism matched to how your customers actually buy
Reward economics that survive honest redemption assumptions
Proof of incremental behaviour, not enrolment vanity
A programme willing to shrink when shrinking wins
Handled by these agents
Works with
Frequently asked questions
Points, tiers, or paid membership?
Frequency decides. Weekly-to-monthly purchase cycles can carry points; quarterly-and-longer cycles suit perks and early access; paid membership requires benefits strong enough to sell — the hardest to pull off, the strongest lock-in when real. Start with the economics, not the fashion.
What does a good programme cost?
Priced as a margin percentage on member revenue against the incremental repeat revenue it creates. If the holdout shows little incremental behaviour, the honest answer is that you built a discount scheme with extra steps.
Do customers even want another loyalty programme?
They want to feel recognised; they do not want another app and a mental exchange rate. Legibility and reachability beat richness — which is why the lightest sufficient mechanism wins.
Which agents run this?
Analyst computes the economics and runs the holdout measurement; Retention operates the member communications and timed benefits; Builder ships the on-site surfaces. The quarterly review runs on schedule with the incrementality numbers attached.
Keep reading
Subscription retention
Reduce subscriber churn with intervention before the cancel, not a survey after it.
Increase average order value
Raise AOV with co-purchase data, threshold design, and offers that do not train discounting.
Customer segmentation
Segments built from purchase behaviour, not from generic engagement windows.
Reduce cart abandonment
A diagnostic-first programme: find why carts die, fix the causes, then automate recovery.
Put the team to work on your store
Connect your storefront, analytics, and email stack, set a goal, and let the agents run the work end to end. Start on the free plan with your own model key.