Repeat Purchase Rate in Ecommerce: Formula, Benchmarks, and the Definition Nobody Agrees On
TL;DR
- Repeat purchase rate measures the share of your customers who bought more than once inside a time window. It is the earliest reliable read on retention, since customer lifetime value takes months to resolve and this number moves in weeks.
- The trap is the denominator. Some sources divide repeat customers by total customers, others divide repeat orders by total orders, and on identical data those answers range from 22% to 51%. Add immature cohorts, samples counted as purchases, and promotion inflated cohorts, and a blended figure stops meaning anything.
- That gap is exactly what Polar closes. The Retention dashboard groups customers by first order period and toggles between calendar and rolling basis, Cohort Evolution breaks repurchase rate down by first order product and channel, and average time to second purchase sits next to the rate instead of in a separate report. Ask Polar runs the analysis conversationally and segments push into Klaviyo so replenishment timing follows the measured cycle.
Repeat purchase rate in ecommerce measures the share of your customers who came back and bought again. It is the cheapest retention number to calculate, it sits in every dashboard, and it is quietly one of the least reliable metrics an operator will quote in a board meeting.
The reason is not complicated. The four pages currently ranking highest for this metric do not use the same formula. Two of them divide by customers. One divides by orders. They land on numbers that can differ by more than twenty points for the same store in the same quarter, and none of them say which window they measured.
This guide gives you the formula, the arithmetic that shows why the denominator matters, what the published benchmarks actually claim, and the four ways a healthy looking repeat purchase rate hides a retention problem. By the end you will be able to state your number in a way another operator can check.
What repeat purchase rate measures in ecommerce
Repeat purchase rate tells you what proportion of your buyers are not one time buyers. That is the whole idea. If 1,000 people bought from you last quarter and 220 of them placed more than one order, your repeat purchase rate for that quarter is 22 percent.
The metric earns its place because acquisition gets more expensive every year and repeat orders do not carry an acquisition cost. A store with a 30 percent repeat purchase rate and a store with a 12 percent repeat purchase rate can run identical ad accounts and end up with completely different contribution margins. The second store has to buy every order it gets.
It also acts as an early warning. Customer lifetime value takes months to resolve, because you cannot know a cohort's lifetime value until the cohort has lived a while. Repeat purchase rate moves earlier. A cohort that is not producing second orders will not produce a good lifetime value later, and you can see that within weeks rather than quarters.
What it does not measure is loyalty, satisfaction, or how fast anyone came back. A customer who bought twice in eight days and a customer who bought twice in eight months both count as one repeat customer. That flattening is the source of most of the trouble later in this article.
The repeat purchase rate formula, and the denominator that changes the answer
There are two families of formula in common use, and they answer different questions. Both are defensible. Reporting one while your colleague reports the other is where the damage happens.
The customer-level formula
This is the standard, and it is what most sources mean when they say repeat purchase rate.
Repeat purchase rate = customers with more than one order / total customers, times 100.
Wall Street Prep defines it as the number of repeat purchase customers divided by the total number of customers. Klaviyo gives the same two step calculation: divide return customers by total customers, then multiply by 100. Yotpo writes it as customers with more than one purchase divided by total unique customers, and is the only one of the three to say out loud that the count sits inside a defined time frame.
The customer-level version answers a question about people: what fraction of the humans who bought from us came back?
The order-level formula
The second family divides by orders instead. CCBill instructs readers to calculate the repeat purchase rate by dividing the number of purchases from repeat customers by the total number of purchases in the period. That is a different metric with the same name.
The order-level version answers a question about revenue mix: what fraction of our order volume comes from people who were already customers? That is a useful number. Finance teams often want exactly that. It is simply not the same as the customer-level rate, and it is almost always larger, because repeat customers place more orders each than one time buyers do.
Worse, the order-level phrasing is itself ambiguous. "Purchases from repeat customers" can mean every order those customers placed, including their first, or only the orders after the first. Those two readings produce different numbers again.
One store, three answers
Take a single quarter at a single store. 1,000 unique customers placed orders. 780 of them ordered once. 220 ordered more than once, and between them those 220 placed 820 orders. Total orders for the quarter: 1,600.
| Definition | Formula | This store | Answers |
|---|---|---|---|
| Customer levelThe standard reading | 220 repeat customers / 1,000 customers | 22% | What share of our buyers came back? |
| Repeat order shareOrders after the first | 600 repeat orders / 1,600 orders | 38% | How much of our volume is incremental to acquisition? |
| Order levelAll orders from repeat buyers | 820 orders / 1,600 orders | 51% | What share of volume comes from known customers? |
Same store, same quarter, same underlying data, and the answer ranges from 22 percent to 51 percent depending on which definition you picked up. Nobody in this example is lying. They are answering different questions and using one name for all three.
The practical fix is boring and it works: write the formula down next to the number, every time it appears. A repeat purchase rate quoted without its denominator and its window is not a measurement, it is a rumour.
What is a good repeat purchase rate?
This is the most searched question on the topic and it has the least satisfying answer, because the published figures disagree and because category drives the number harder than performance does.
What the published benchmarks actually say
Klaviyo states that a good repeat purchase rate is typically around 20 to 30 percent. Yotpo publishes bands by category instead, and its bands sit higher for consumables and lower for durable goods.
| Category | Published band | Source |
|---|---|---|
| Consumables, food and beverage, cosmetics | 30% to 40% | Yotpo |
| Fashion and apparel | 25% to 35% | Yotpo |
| Cross category, general guidance | 20% to 30% | Klaviyo |
| Electronics and high ticket items | 10% to 20% | Yotpo |
Read those bands with the previous section in mind. None of the sources publishing them states its denominator in the same breath as the figure, and at least one of the pages ranking alongside them is calculating on orders rather than customers. Comparing your 22 percent against someone else's 35 percent is only meaningful if both were built the same way.
Category moves this number more than performance does
Klaviyo makes the point directly: rates run higher for affordable or perishable goods and lower for high value goods such as technology and luxury. That is a statement about replenishment cycles, not about marketing quality.
A coffee brand and a mattress brand can both be excellent operators. The coffee brand will post a repeat purchase rate several times higher, because coffee runs out. If you sell something people buy once every four years, a 12 percent quarterly repeat purchase rate might be outstanding, and chasing an industry average would push you into discounting that destroys margin without changing behaviour.
The only benchmark that reliably means something is your own trend. Cohort over cohort, is the number rising? That comparison holds the category constant automatically.
The time window nobody states
A repeat purchase rate is meaningless without a window, and almost every published figure omits it. Measured over a customer's entire history, a store's rate climbs forever, because customers keep having chances to come back. Measured over 30 days, the same store looks terrible.
Pick a window that matches your replenishment cycle, state it, and hold it constant. Ninety days works for most consumables. Twelve months is more honest for considered purchases. What matters is that the window does not move between reports.
Repeat purchase rate vs retention rate vs returning customer rate
Three names circulate for closely related ideas, and the confusion is common enough that it is one of the more active threads on the topic in ecommerce communities.
Repeat purchase rate counts customers who bought more than once, out of all customers, inside a window. It is a property of a group of buyers.
Customer retention rate asks whether the customers you had at the start of a period are still buying at the end of it. It starts from an existing base and measures survival. A store can have a high repeat purchase rate and a falling retention rate at the same time, if it is acquiring well and losing its older customers quietly.
Returning customer rate usually describes order mix in a period: how many of this month's orders came from people who had bought before. It is closer to the order-level formula above, and it is the one most likely to be shown to you by a storefront analytics dashboard without a definition attached.
There is a fourth distinction worth holding onto, because it changes what you do about the number. A repeat customer is someone on their second, third or fourth order. A returning customer may be someone who bought two years ago, disappeared, and came back. Both count identically in a naive calculation, and they need completely different treatment: one is a lifecycle to accelerate, the other is a reactivation.
Four ways repeat purchase rate lies to you
The formula is arithmetic, so it cannot be wrong. What goes wrong is the data it runs on. These four are the ones that come up again and again with operators.
1. Immature cohorts read as decline
This is the most common misreading and it causes real panic. Recent cohorts always look worse than older ones, because they have had less time to produce a second order.
If you are looking at a 90 day repeat purchase rate on a cohort acquired 50 days ago, that cohort is not underperforming. It is unfinished. A customer acquired at the start of the month has had far longer to come back than one acquired on the last day of it, and lumping them into one bucket makes the newest cohort look like a problem to solve.
The fix is to grey out or exclude any cohort that has not fully aged past your window. A cohort table that shows immature cells with the same visual weight as mature ones will generate a false alarm roughly once a month.
2. Transactions that are not really purchases
Plenty of stores sell things that are not products. Sample kits, fabric swatches, five dollar accessories, gift card top ups, warranty registrations. Every one of them creates an order and a customer record.
A brand that sells swatches before a large considered purchase will show an inflated repeat purchase rate that is really just a two step buying process. The swatch order and the real order are one purchase decision. Counting them as two makes retention look strong while the actual repeat rate on finished goods might be a third of the reported figure.
Tag those transactions and exclude them explicitly. This one is easy to fix and very easy to miss, because the number it produces is flattering rather than alarming.
3. Promotion inflated cohorts
A heavy discount month brings in a lot of customers who came for the discount. They convert well, they inflate the cohort size, and most of them never return at full price.
The effect is delayed and it is easy to misattribute. Three months after a large sale the repeat purchase rate drops, and the natural conclusion is that something broke in the lifecycle emails. Usually nothing broke. The denominator was filled with people who were never going to buy again, and acquisition quality is showing up on the retention curve exactly where you would expect it.
Segmenting cohorts by whether the first order carried a discount code separates the two stories cleanly. If discounted first orders repeat at half the rate of full price ones, that is a paid acquisition decision, not an email problem.
4. Mix shift in a growing store
A store that doubles its new customer count will see its blended repeat purchase rate fall, even if every cohort is performing identically to last year. New customers enter the denominator immediately and can only enter the numerator later.
Growth suppresses the blended number mechanically. This is the single strongest argument for reading repeat purchase rate by cohort rather than as one store wide figure. The blended number tells you as much about your growth rate as it does about your retention.
The number repeat purchase rate hides: time to second order
Repeat purchase rate is a yes or no question. Did this customer come back inside the window? A customer who reordered in three weeks and one who reordered in seven months are recorded identically, and they are not the same customer at all.
Time to second order, sometimes called purchase latency, is the paired metric that restores what the rate throws away. It answers how fast, not whether. Two stores with the same 25 percent repeat purchase rate can have completely different economics if one gets its second orders in 40 days and the other takes 200. The faster store recovers acquisition cost sooner, compounds within the same reporting year, and can run a lifecycle programme that actually fits the gap.
It is also the more actionable of the two. Repeat purchase rate is a lagging outcome of many things. The gap between first and second order is a specific, targetable interval: you can time a replenishment reminder to it, size a post purchase offer against it, and measure whether the gap moved. Compressing the first to second order gap is usually the largest available lever on customer lifetime value, and it is invisible if the rate is the only number you track.
Track the two together. A rate that holds steady while the gap stretches is a retention problem that has not surfaced yet.
How to measure repeat purchase rate properly with cohorts
Every problem above is solved by the same move: stop reporting one store wide number and start reporting the rate by acquisition cohort. A cohort is simply a group of customers bucketed by when they placed their first order.
Calendar basis versus rolling basis
There are two ways to lay out the periods and they answer slightly different questions.
Calendar basis buckets by the actual calendar month, quarter or year. Period zero is the period in which the customer placed a first order. This is the traditional cohort triangle, and it lines up neatly with financial reporting. Its weakness is the one described earlier: a customer acquired on the 30th of the month gets credited with a full period they did not have.
Rolling basis measures from each customer's own first order date. Period one is that customer's first repeat window, whenever they happened to arrive. This removes the distortion for late period acquirers and gives a much fairer read on recent cohorts. It is the better basis for judging whether a cohort is performing, and the harder one to reconcile against a calendar P&L.
Use rolling to judge performance, calendar to report to finance, and do not mix them inside one chart.
Order-rank cohorts
The most useful view is the one almost nobody builds. Instead of asking what percentage came back inside a window, ask what percentage reached each order number: what share of the cohort placed a second order, a third, a fifth.
This turns retention into a funnel with named steps, and the steps behave very differently. The drop from first to second order is nearly always the steepest in the whole sequence, and it is where the money is. Customers who reach a third order tend to convert to a fourth at much higher rates, because by then buying from you is a habit rather than a decision.
Order-rank cohorts also tell you where to spend. If the first to second step is your cliff, lifecycle work and replenishment timing are the levers. If second to third is where you lose people, the problem is more likely assortment or the product itself. A blended repeat purchase rate cannot distinguish those two situations, and they call for entirely different budgets.
The same logic underpins how to think about ecommerce customer retention generally: the aggregate is a symptom, the cohort is the diagnosis.
How to improve repeat purchase rate
Once the number is measured properly, the interventions are reasonably well understood. Ranked roughly by how much they move the metric for most stores:
Time the second order prompt to your actual replenishment cycle. Not to a generic 30 day flow. If your median gap to second order is 63 days, a reminder at day 21 arrives before anyone has run out and a reminder at day 90 arrives after they have bought elsewhere. Measure the gap first, then build the flow around it.
Work on the first to second order step specifically. It is the steepest drop in the sequence, so a point gained there is worth more than a point gained anywhere else. Post purchase sequences, a genuinely useful onboarding email, and making reordering take two clicks all attack this step directly.
Find your gateway products. Some first order products produce customers who come back at much higher rates than others. Break repeat purchase rate down by first order product and the pattern is usually obvious within a minute. Then push acquisition budget toward the products that create repeat buyers rather than the products with the best first order conversion rate. Those are rarely the same list.
Fix acquisition quality before blaming retention. If discounted first orders repeat at half the rate of full price ones, the retention programme is not the problem. Look at the cohorts a promotion produced before you rebuild the email flows.
Segment and act on the segments. A customer who has bought twice and a customer who has bought once need different messages, and a lapsed customer past their normal cycle needs a third. Pushing those segments into your email and ads platforms so the timing follows the data is where measurement turns into revenue.
All of this compounds into customer lifetime value on Shopify, which is the number these efforts are ultimately paying into.
How Polar tracks repeat purchase rate
Polar's Retention dashboard is built around the cohort view rather than the blended figure. Customers are grouped by first order period, and the cohort table shows one metric at a time across periods, with a toggle between calendar and rolling basis so you can switch between the finance read and the performance read without rebuilding anything.
Alongside the table, Cohort Evolution charts repurchase rate and lifetime value trends across cohorts, and breaks them down by first order product, marketing channel, geography and campaign. That is what turns the gateway product question from a data project into a dropdown.
Average time to second purchase sits next to repeat customer rate rather than in a separate report, which is the pairing this article argues for. Ask Polar will run the analysis conversationally and point at the largest gap, and customer segments can be pushed into Klaviyo so the replenishment timing follows the measured cycle instead of a guess.
For the metrics where we do publish aggregate figures, conversion rate, ROAS, customer acquisition cost, average order value and cart behaviour, our DTC benchmarks drawn from 4,000+ Shopify brands are free and refreshed weekly. Repeat purchase rate is not currently one of them, which is part of why the published bands in this article come from other vendors rather than from us.
State your denominator, state your window, read it by cohort, and track the gap alongside the rate. That is most of the distance between a repeat purchase rate that starts arguments and one that settles them.
FAQ
Table of contents
Frequently asked questions
