The second order is the hardest
A customer who has ordered twice behaves very differently from one who has ordered once. The first order is an experiment: the shopper does not know whether the product will be as described, whether delivery will arrive, whether returns will be painful. The second order is a decision made with that uncertainty resolved. This is why, in most stores, the share of customers who go from one order to two is far lower than the share who go from two to three, and why the second-order rate is the retention metric most worth owning.
Designing for the second order starts before the first one is delivered. Three things decide it. The first experience — delivery as promised, packaging that does not disappoint, a product that matches the page — is the foundation; no retention tactic repairs a bad first order. The next purchase has to be obvious: a customer who bought a razor needs to know blades exist and when they will run out, a customer who bought a dress needs to see what goes with it. And the store needs permission and a reason to make contact after delivery, which is the moment most stores waste on a single request to rate the order. The e-commerce email and SMS flows guide covers the post-purchase sequence; the design question here is what that sequence should be pointing at. The tutorial on building a post-purchase email flow builds it, with a holdout of first-time buyers that shows what the flow adds to second orders.
The honest measure is the second-order rate within a window matched to the category's natural cycle, read by first product bought and by acquisition channel. Both cuts tend to show large differences, and both suggest actions: lead acquisition with the products that bring customers back, and judge channels on the customers they deliver rather than the first orders they produce. Whether the next unit of budget should go to earning second orders or to finding new customers is the question the comparison of customer acquisition and retention answers from the store's unit economics.
Replenishment and cross-category paths
Every catalogue has a few natural paths from a first purchase to a second. Finding them is the core of repeat-purchase design, and they fall into a handful of types.
| Path | How it works | Catalogues where it applies | Design task |
|---|---|---|---|
| Replenishment | The product runs out and is re-bought | Consumables: coffee, skincare, pet food, supplements, printer ink | Learn the re-order cycle; prompt just before it ends; offer a subscription |
| Consumable for a durable | A durable item needs ongoing supplies | Appliances with filters, pods or blades; printers; tools | Make the consumable visible at first purchase and at the natural interval |
| Completion | The first item is part of a set or system | Furniture ranges, cosmetics lines, sports kit, collectables | Show the set; make the next piece the obvious add |
| Cross-category | A customer in one category is a good prospect for another | Multi-category retailers; fashion to footwear; home to garden | Find the categories that follow each other in order data; recommend across them |
| Occasion and season | The need recurs on a calendar | Gifts, seasonal clothing, school and holiday purchases | Remember the occasion; prompt before it comes round |
| Upgrade and replacement | The item wears out or is superseded | Electronics, footwear, luggage, mattresses | Know the lifespan; return at the right time with the successor |
The order data already contains the answer to which paths matter. A simple analysis of what customers who bought a given product bought next, and how long they took, produces the replenishment cycles and the cross-category pairs the store should build around. The customer analytics section covers that analysis; the output for a store is a short list of first-purchase products with their most common next purchase and the median gap, which becomes the brief for both merchandising and the flows. The tutorial on finding what customers buy after their first order builds that list from the order history step by step, with the customers who never came back kept in every rate.
One trap in replenishment: stock-up promotions and volume discounts delay the next order as surely as they raise the current one. A path that depends on a regular cycle is damaged by promotions that break the cycle. The damage shows up later as lower purchase frequencyAverage number of orders each customer placed in the period. for the customers who stocked up. Where a product's re-order cycle is steady, the guide to subscribe-and-save programmes covers turning the replenishment path into a subscription, with its own offer, cadence controls and cancellation terms.
Loyalty programme types and economics
A loyalty programme is a structured promise: keep buying here and you will get something for it. The promise can take several forms, and the economics differ enough that the choice of type should come from the store's margin and purchase cadence rather than from what competitors run.
Points programmes award points per unit of spend, redeemable against future orders. They are flexible and familiar, they generate a liability the store has to account for, and their effect on behaviour depends on whether customers can see a reward within reach. Points that take a year to become useful do not change behaviour this month.
Tiered programmes give status and benefits that improve with spend or frequency over a period. They work where customers care about recognition and where the top tier's benefits are real: early access, free delivery, dedicated support. They suit stores with a wide spread of customer value.
Paid memberships charge a fee for benefits such as free delivery, discounts or exclusive products. The fee itself selects for committed customers, and the benefits need to be worth clearly more than the fee to those customers; the risk is a programme that only the already-loyal join.
Cashback and store credit return a share of spend as credit for the next order, which pulls the next order forward more directly than points and is simpler to understand.
Perks without points — early access, free returns, members' pricing on selected lines — avoid the liability and the arithmetic of points and suit stores whose customers buy infrequently but at high value.
The economics come down to one question: does the programme change behaviour, or does it reward behaviour that would have happened anyway? A programme whose rewards go mostly to customers who were already buying at full frequency is a discount by another name. Incremental orders, higher order value (AOVAverage revenue per order.) among members, and reduced lapse rates are the gains to look for; the costs are the reward value, the operational overhead and the margin given away on orders that needed no incentive. The retention and payback guide sets this against acquisition cost so that the programme is judged on payback rather than on enrolment. For a consumable bought again on a steady cycle, the comparison of a loyalty programme with subscribe-and-save weighs a reward after each order against a standing order agreed in advance.
Tiers, rewards and breakage
The details of a programme decide whether it works. Three deserve particular care.
Tiers need to be reachable. A middle tier that most engaged customers can attain within a few orders creates the status effect; a top tier that only a tiny share will ever reach is aspiration for the few and irrelevance for the rest. Tier periods — the window over which status is earned and the point at which it is reviewed — should match the category's buying cycle, and the drop from a tier should be softened with a warning and a grace period, because losing status is a strong trigger to leave.
Rewards should be things customers want and the store can supply at low marginal cost. Free delivery, early access to new lines, a members' sale, a free sample with orders: these cost the store less than their perceived value. Cash-equivalent rewards are the most expensive and the least differentiated. The reward's visibility matters as much as its value; a customer who can see they are a small step from a reward behaves differently from one who has to log in to find out.
Breakage is the share of earned rewards that are never redeemed. Some breakage is normal and it reduces the programme's cost, which is why finance departments like it. Too much breakage means the programme is not working: customers who do not redeem are not being motivated by the rewards. Expiry rules, reminder messages and low redemption thresholds all reduce breakage, and a programme should be judged on the behaviour change among redeeming members rather than on the savings from those who never engage. Consumer protection rules in some markets also govern expiry and the clarity of terms, so the terms should be plain and the expiry visible.
A pet-supplies store redesigns its points programme
A pet-supplies store runs a points programme with 42,000 members: one point per £1, with a £5 voucher at 500 points. Analysis shows that only 11% of members have ever redeemed, breakage is 68%, and members' purchase frequency is identical to non-members'. The programme is a discount for a small group and noise for everyone else. The store redesigns it: the first reward is now a free bag of treats at 150 points, reachable on the second order; a two-tier structure gives free delivery on all orders above £25 at the upper tier, earned with three orders in six months; and the post-purchase flow shows the points balance and the distance to the next reward. After six months, redemption among new members is 38%, members order 1.6 times more often than non-members in the same cohort, and the programme's cost per incremental order, including rewards and operations, is £3.40 against an acquisition cost of £22 for a new customer.
Win-back
Customers lapse. Some move away from the category, some were never going to return, and some simply forgot or drifted to a competitor. A win-back effort targets the last group, and the design question is when a customer counts as lapsed and what will bring them back.
Lapsed should be defined relative to the category's cycle, not an arbitrary date. A coffee customer who has not ordered in twice their usual gap is lapsing; a mattress customer who has not ordered in a year is normal. The point at which the probability of a further order drops sharply — visible in the time-between-orders distribution — is where win-back should begin, and it is usually earlier than stores assume.
The offer should be sized to the customer's value, not the same for everyone. High-value lapsed customers justify a personal message, a meaningful incentive, or a call; low-value lapsed customers justify an automated sequence and no more. Scoring customers on recency, frequency and spend with RFM segmentation is a simple way to separate the two groups before the sequence is built. A win-back message that simply announces a discount trains customers to lapse deliberately; one that returns with something new — a product the customer has not seen, a reminder of what they bought and what has changed — gives a reason that survives the discount. Beyond a certain point, win-back stops paying: customers who have ignored a sequence are better removed from active marketing than messaged indefinitely, which also protects deliverability and consent.
The lifecycle marketing guide places win-back in the full customer lifecycle alongside onboarding and nurture; the mechanics of the win-back flow itself belong to the flows guide. The same lapsed definition decides whom paid prospecting should still exclude: the guide to excluding existing customers in Google Ads keeps active customers out of acquisition campaigns while leaving lapsed ones open to reactivation.
Whether lapsed customers need a discount or only a reminder is settled by the win-back offer test, which compares offer, message-only and holdout arms on incremental gross profit.
Measuring loyalty without flattering yourself
Loyalty programmes are easy to make look good. Members buy more than non-members, which proves nothing, because the customers who join a programme were already the ones buying more. Measurement has to separate the programme's effect from the selection effect, and a few disciplines do that.
Compare like with like. Members against non-members matched on their behaviour before joining, or the same customers before and after joining, rather than members against everyone else.
Hold out. Where the programme allows, keep a small share of eligible customers out of a reward or a tier benefit and compare. This is the only clean measure of incrementality and it is worth the discomfort of withholding a benefit from a few. Before a programme launches at all, the loyalty programme pilot test sets out how to hold out part of the eligible base from the invitation and read the result after every reward cost. A programme that has been live for a year without moving retention is worked through cause by cause, from rewards on habitual purchases to a reading on a blended rate, in the troubleshooting page for a loyalty programme not increasing retention.
Track the behavioural indicators (KPIA metric tied directly to an objective.) rather than enrolment. Enrolment is a vanity figure. Second-order rate among members, time between orders, lapse rate, redemption rate and margin per member after rewards are the ones that reveal whether the programme is doing its job.
Count the full cost. Reward value at redemption, the liability on the balance sheet for unredeemed points, platform fees, and the staff time to run it. Set that against incremental margin, not incremental revenue. Incremental marginThe margin an activity added: margin per customer assigned to it minus margin per comparable customer held back, before the activity's own costs are taken off. comes from the holdout described above: the margin per customer the programme added over customers kept out of it, before its costs.
Read it by cohort. A programme that improves retention for customers acquired this year while last year's customers lapse at the old rate is working on new customers only, which is useful to know. The retention hub covers the cohort view and the metrics that frame all of this.
To separate repeat purchases a programme causes from those it merely records, see the guide to mapping loyalty rewards onto a flywheel as force or friction.
For a stage-by-stage view of the store, AARRR for ecommerce sales treats the second order within the category's buying cycle as the retention stage and finds the constraint there.
The retention objective hub sets the cohort KPI set for repeat purchase and compares the channels that can deliver it.
Where the goal is customers who also recommend the brand, the loyalty and advocacy objective hub separates advocacy from repeat behaviour and from points schemes.
Discounts aimed at existing customers can pull purchases forward from full price, and the discount depth test measures that cost with a follow-up window after the promotion.
Common mistakes
- Designing for the hundredth order and ignoring the second. The second-order rate is where the economics turn; most programmes reward customers who are already past that point.
- Copying a competitor's programme. Programme type should follow the store's margin and buying cycle, not the category's habit.
- Points nobody can reach. A first reward that takes a year to earn changes nothing this month.
- Celebrating breakage. Unredeemed rewards save money and signal that the programme is not motivating anyone.
- Discount-only win-back. It trains customers to lapse on purpose; return with something new as well as a reason.
- Measuring members against everyone else. The customers who join were already the better customers; compare like with like or hold out.
- Promotions that break replenishment cycles. Stock-up offers raise this order and delay the next one.
- Letting service undermine the programme. No points balance compensates for a late delivery or a hard return.
