Repeat purchase rate vs retention rate: which to use
In short
Repeat purchase rate is the share of your customers who have bought more than once. Retention rate is the share of one cohort still buying after a set time. The difference matters because repeat rate moves when your acquisition moves, even if nothing about loyalty changed. Cohort retention does not.
What is the difference between repeat purchase rate and retention rate?
Repeat purchase rate counts people. Cohort retention rate counts a *specific* group of people over time. The gap between them is time: repeat rate throws it away, cohort retention makes it the axis.
| Repeat purchase rate | Cohort retention rate | |
|---|---|---|
| The question it answers | Of everyone who bought, how many bought more than once? | Of the people whose first order was in month X, how many had come back by month Y? |
| Denominator | Every customer active in the period | One cohort, fixed for ever |
| Handles time by | Ignoring it | Making it the axis |
| Moves when acquisition moves | Yes | No |
| Good for | One number on a dashboard | Deciding whether a change worked |
Most guides stop at the definitions. The consequence is the part worth your attention, and it is in the fourth row: one of these two numbers moves when your media buying changes shape, and it moves in the direction that looks like a loyalty problem.
How to measure repeat rate in ecommerce
Repeat purchase rate is customers with two or more orders, divided by all customers, inside a window you choose. Three of those words hide a decision, and a repeat rate quoted without all three is not comparable to anyone else's.
- The window. Twelve months is the common default. A shorter window mechanically produces a smaller number, because fewer customers have had time to come back.
- The denominator. "All customers who ordered in the window" and "all customers whose first order was in the window" are different populations and give different answers.
- What counts as an order. Exchanges processed as new orders inflate the numerator, which is why apparel repeat rates read high. Refunded and cancelled orders should come out.
Write the three choices down next to the number. Changing any one of them silently is how a metric stops meaning anything across quarters.
Why your repeat rate fell while loyalty improved
Here is the failure mode nobody puts on the page. Take a store whose customers behave identically in two consecutive years — same product, same emails, same cohort curve:
| Months since first order | Share of the cohort that has ordered again |
|---|---|
| 3 months | 12% |
| 6 months | 20% |
| 9 months | 26% |
| 12 months | 30% |
Now acquire 12,000 customers in a year, two ways. Pattern A spreads them evenly: 6,000 in the first half, 6,000 in the second. Pattern B is a store scaling up: 3,000 in the first half, 9,000 in the second.
At the end of December, the first-half customers have had roughly nine months to come back and the second-half customers roughly three.
| First half | Second half | Repeaters | Repeat rate | |
|---|---|---|---|---|
| Pattern A | 6,000 × 26% = 1,560 | 6,000 × 12% = 720 | 2,280 | 19.0% |
| Pattern B | 3,000 × 26% = 780 | 9,000 × 12% = 1,080 | 1,860 | 15.5% |
Same customers, same loyalty, same year, same total. Pattern B reports 15.5% against Pattern A's 19.0%, a fall of 3.5 points, and the only thing that changed is *when* the customers arrived. A store that grows its acquisition through the year will watch its repeat rate drop while every underlying cohort performs exactly as before.
The cohort view is immune to this. Both patterns reach 30% at twelve months, because a cohort is always compared with itself at the same age.
This is also why a repeat rate is close to useless for judging a retention campaign. If the campaign ran alongside a scaling quarter, the acquisition shape and the campaign are pulling the same number in opposite directions and you cannot separate them afterwards. Measuring customer retention properly means holding maturity constant, then comparing.
What is a good repeat purchase rate?
There is one benchmark in this category worth quoting, because it states its sample. Across 156,110+ customers in one agency's DTC portfolio, measured over a 365-day lookback, the repeat purchase rate was 18.8%. The same report breaks it down by category:
| Category | Reported range |
|---|---|
| Consumables | 22–44% |
| Fashion | 10–17% |
| Durables | 7–18% |
A consumables brand and a furniture brand are three to six times apart, which tells you that a cross-vertical average is a number with no use. Compare with your own last four quarters instead, on an identical window and denominator.
Treat any benchmark without an *n* as decoration. LoyaltyLion's retention guide, currently the top result for the formula, asserts that e-commerce brands average around 30% and attaches no sample size to it. That figure may well be right; there is simply no way to check it, and no way to know whether the stores behind it look anything like yours.
Retention rate formula for ecommerce
The formula in general circulation is:
Retention rate = ((E − N) / S) × 100
where E is customers at the end of the period, N is customers acquired during it, and S is customers at the start. It comes from subscription businesses and it carries an assumption that does not survive the move to a store: that "customer at the end of the period" means something.
In a subscription, it does — the subscription is live or it is cancelled. In a transactional store nobody cancels anything. They just stop buying, and they do not announce it. To use the formula you have to invent the missing definition by choosing an active window: a customer is "active" if they ordered in the last 90 days, or 180, or 365. The number you publish then depends on that choice as much as on your customers.
That is not a reason to avoid the formula. It is a reason to state the window every time you quote the output, and to keep the window fixed once you have picked it. If you never want to make that choice, use cohort retention, where the question "is this person still a customer?" never has to be asked.
Which metric to use, and when
| If you want to | Use | Because |
|---|---|---|
| Report one loyalty number to a board | Repeat purchase rate, window stated | It is one figure and it is easy to reproduce |
| Know whether last quarter's retention work did anything | Cohort retention | Acquisition shape cannot contaminate it |
| Forecast lifetime value | Cohort retention | The curve, not one point on it, is what compounds |
| Diagnose a sudden drop | Both | If cohorts are flat and repeat rate fell, you have an acquisition-mix story, not a loyalty one |
The practical setup is to keep repeat rate for the dashboard and cohort curves for decisions, and never let the first one trigger a project on its own. Before it does, check whether the customer retention curve moved at all. If it did not, the money is better spent on how to increase average order value than on a loyalty programme that is fixing a problem the cohorts say you do not have.
Common questions
Is repeat purchase rate the same as repeat customer rate?
In practice yes — both name the share of customers with two or more orders. The label is not the problem; the undeclared window and denominator underneath it are. Two stores quoting 'repeat customer rate' are usually not measuring the same thing.
Why did my repeat rate drop after a good quarter?
Most often because you acquired more customers late in the period and they have not had time to come back yet. Check the cohort curves before concluding anything about loyalty. If every cohort sits where it did last year, the drop is arithmetic, not behaviour.
What window should I use for repeat purchase rate?
Twelve months is the common default and the one most published benchmarks use. Match the window to your purchase cycle: a coffee subscription and a mattress brand should not share one. Whatever you pick, keep it fixed, because changing it moves the number on its own.
Should exchanges count as repeat purchases?
No, and this matters most in apparel, where exchanges are often processed as fresh orders. Left in, they turn a returns problem into an apparent loyalty win. Exclude them along with refunded and cancelled orders.
Do I need a cohort tool to measure retention properly?
No. A cohort table is a pivot of order history by first-order month against order month, which a spreadsheet handles for most catalogue sizes. A tool saves time and removes copy-paste errors; it does not change the definition.
See it on your data
On a twenty-minute call we connect one ad account and one month of your orders, and work out the gap between what the platform claims and what your shop took. Profit and POAS stay blank until you enter your margin and costs per order.
Book a demo