"Our return rate is 12%" is a sentence with at least two meanings, and the difference between them decides whether your margin analysis is right. Twelve percent of units returned and twelve percent of revenue refunded are different measurements that rarely produce the same number.
This guide separates the two, shows how returns reach profit, and explains why the cost of a return is larger than the sale you lost.
Two measurements, both called "return rate"
Unit return rate = Returned units ÷ Units sold
Refund ratio = Refund value ÷ Realized revenue
They diverge whenever returns are not evenly distributed across your price range, which is to say almost always. A store returning 8% of units but 15% of value has a problem concentrated in its expensive products. The reverse — 15% of units, 8% of value — points at a cheap item coming back constantly, which is an operational cost problem rather than a margin one.
| Product | Units sold | Price | Units returned | Value returned |
|---|---|---|---|---|
| Coat | 300 | $180 | 66 | $11,880 |
| T-shirt | 1,400 | $28 | 70 | $1,960 |
| Total | 1,700 | — | 136 | $13,840 |
Unit return rate: 136 ÷ 1,700 = 8.0%. Refund ratio against gross sales of $93,200: 13,840 ÷ 93,200 = 14.8%. Same month, same store, both correct, nearly double the figure depending on which you quote.
The coat is the problem: 22% of coats come back, and each one takes $180 with it. The t-shirt's 5% unit rate is almost irrelevant to margin. A single blended "return rate" would have hidden that entirely.
How returns reach revenue
There are two conventions, and mixing them double counts the loss.
The practical consequence when reading your own reports: if your revenue line already excludes returns, do not subtract a refunds line beneath it. And when comparing your numbers against a benchmark or a previous system, check which convention each used before concluding anything changed.
Why the timing matters too
A return usually lands in a later period than the sale. If revenue is recognized on the order date and refunds land six weeks later, a growing month looks better than it is and a shrinking one looks worse — the returns arriving now belong to sales you already counted.
Recognizing revenue on delivery narrows that gap considerably, because the sale and its outcome sit much closer together.
What a return actually costs
Removing the sale from revenue captures the margin you did not earn. It does not capture what the attempt cost, and that is usually the larger surprise.
| Cost of one returned $180 coat | Amount | Recovered? |
|---|---|---|
| Gross profit not earned | $99.00 | Never existed |
| Outbound shipping | $9.50 | No |
| Return shipping | $11.00 | No |
| Inspection & restocking labor | $4.00 | No |
| Payment processing on the original sale | $5.31 | Usually partly |
| Unit value lost if not resellable | $81.00 | Sometimes |
| Hard cost, resellable unit | $24.50 – $29.81 | |
| Hard cost, unsellable unit | $105.50 – $110.81 |
Illustrative figures. The point is the spread: whether the unit can be resold changes the cost of a return by a factor of four. Apparel returned in sellable condition is a shipping-and-labor cost; the same item returned worn is the whole landed cost as well.
This is why return rate belongs alongside product profitability rather than in a customer-service report. A product with 22% returns and a $4 handling cost is a different business from one with 22% returns and an $81 write-off. See product profitability.
The effect on advertising economics
Attribution is recorded at purchase. Returns arrive afterward and generally never flow back into reported ROAS, so a campaign selling a high-return product reports revenue that partly does not exist.
If a campaign reports 3.5x ROAS on a product with a 22% return rate, the revenue that survived is roughly 2.73x — before any cost is considered. Against a 55% gross margin, that moves the campaign from comfortably above a 1.82x break-even to barely clearing it. Break-even ROAS works that through properly, and ROAS vs profit covers why attributed revenue and business revenue must stay separate.
On industry benchmarks
You will find widely quoted figures for "average ecommerce return rate". Most trace back to survey data with undisclosed methodology, mix categories with wildly different behavior, and are republished until the original source is unrecoverable.
We are not going to quote one. A number you cannot trace is not evidence, and the comparison that matters is narrower anyway: this product against that product, this month against last month, this variant against the size above it. Your own data answers the question a benchmark cannot, which is what changed.
Reducing returns, in order of leverage
- Sizing information, for apparel. Measurements per size, on the product page, in the units your customers use. Most apparel returns are fit.
- Photography that matches the product. Colour accuracy and scale. A return driven by "not as pictured" is a photograph problem, and it is cheap to fix.
- Return reasons, captured and counted. You cannot fix what you cannot group. Three reason codes used consistently beats twelve used inconsistently.
- Per-variant analysis. Returns usually concentrate in specific sizes or colours. An overall product return rate hides that.
- Reconsidering the promotion mix. Discount-driven and impulse-driven orders return at different rates from full-price ones in many categories — worth measuring in your own data before assuming either direction.
Common mistakes
Quoting one return rate without saying which
Units or value. They are different numbers and they support different decisions.
Subtracting refunds from revenue that already excludes them
Double counting. Check which convention your revenue line uses before deducting anything from it.
Counting only the refund
Shipping out, shipping back, labor and any loss of value are all real and none of them appear in the refund amount.
Measuring returns per store instead of per product
A store-level rate is an average of things that should be managed separately. The blended 8% in the example above was two products at 22% and 5%.
Judging a month's return rate on that month's sales
Returns lag. Dividing this month's returns by this month's sales understates the rate while you are growing and overstates it while you are shrinking. Cohort the returns back to the period of the original sale when you need precision.
How ORVX handles returns
Revenue comes from delivered order items only, so a returned order is never part of the top line and is never deducted from it either. Unit return rate and refund-to-revenue ratio are reported as separate metrics, deliberately, because collapsing them into one "return rate" loses the distinction this guide is mostly about.
Return behavior is visible per product, which is where it is actionable — see how to calculate ecommerce profit for where all of this sits in the P&L.
Frequently asked questions
How do I calculate ecommerce return rate?
Decide which one you mean. Unit return rate is returned units divided by units sold. Monetary refund ratio is refund value divided by realized revenue. Track both — a large gap between them tells you returns are concentrated at one end of your price range.
What is the average return rate for ecommerce?
We do not publish one. The figures in circulation come from surveys with undisclosed methodology and blend categories that behave nothing like each other, so comparing your store to them tells you little. Compare your products against each other and against their own history instead.
Should returned orders be subtracted from revenue?
Only if your revenue figure included them in the first place. ORVX recognizes revenue from delivered items only, so returned orders were never in the total and subtracting them again would remove the sale twice. Whichever convention you use, apply exactly one.
Do returns affect ROAS?
They affect profitability but usually not the reported figure, because platforms record attributed revenue at purchase and returns arrive later. That is one of the reasons a campaign can report a strong ROAS and still reduce profit.
What does a return actually cost?
The margin you did not earn, plus outbound shipping, return shipping, inspection and restocking labor — and the full landed cost of the unit if it cannot be resold. That last condition is the one that matters most: it changes the cost of a return several-fold.