Repeat Purchase Rate
Date: 2026-08-16
The proportion of customers who buy again. It’s the single most diagnostic number in ecommerce, because it decides whether you have a business or a series of transactions — and it’s the one most sites can’t state.
What it is
Repeat purchase rate (RPR) is the share of customers who place more than one order.
On the running model — 10,000 new customers a year and 15,000 total orders — around 5,000 orders come from returning customers, giving roughly 1.6 orders per customer.
Why it decides everything
Because a customer who buys once is worth their first-order contribution, and one who buys repeatedly is worth a multiple of it — against the same acquisition cost.
CAC £22.50, contribution £15.00 per order
one order only £15.00 − £22.50 = −£7.50 loses money
two orders £30.00 − £22.50 = +£7.50
three orders £45.00 − £22.50 = +£22.50
Every new customer loses money on their first order. The entire business case rests on the second one — which makes RPR the number that decides whether acquisition is investment or expenditure.
It also caps how much you can afford to bid. A business at 15% RPR simply cannot pay what one at 45% can, on identical products.
Defining it properly
Three choices, all of which need stating:
Window. “Repeat within 12 months” is measurable; “ever repeats” is unfalsifiable and always rising. Use a bounded window matched to your repurchase cycle.
Cohort or snapshot. A snapshot (“what share of all customers have ordered twice”) mixes cohorts of every age and rises mechanically as the customer base ages. Cohort-based is the only version that can be compared over time — Cohort Analysis.
Customers or orders. RPR counts customers. “Percentage of orders from returning customers” is a different, also useful metric — and the two move independently.
Reading it
- By acquisition channel. The spread is usually enormous — discount-led acquisition retains far worse — and blended RPR hides which channels are buying customers worth having — Channel Mix
- By first product or category. Entry product predicts repeat behaviour strongly, and it’s actionable: promote the entry points that produce repeaters
- By cohort over time. Improving month-1 repeat rate cohort on cohort is the clearest evidence a lifecycle change worked
- Alongside Time Between Orders. RPR says whether they come back; the interval says when, and the interval is what you can act on
Improving it
The highest-return work in most retail businesses, because nobody is competing for it:
- The post-purchase window. The period immediately after the first order is when repeat propensity is highest and most sites do nothing — Lifecycle Messaging
- Replenishment timing for consumables. Reaching someone as they run out is the single most effective trigger available — Replenishment Timing
- The second-order offer, targeted at first-time buyers only. Unlike blanket discounting there’s no cannibalisation, because they weren’t buying
- Delivery and returns experience. The largest driver of whether someone orders again is whether the first order went well, and that’s operations rather than marketing
- Range depth. Customers repeat if there’s something else to buy
Where it’s measured wrongly
- Identity failures read as one-time customers. A returning buyer with a new device and a guest checkout is a new customer in your data. Guest checkout is the main destroyer of measurable RPR — match on email against the order system, not on cookies — Identity Stitching, User Counting
- Snapshot rather than cohort, so the number drifts upward with base age
- Returns not deducted — a customer who bought twice and returned once
- No window, making it unfalsifiable