Returns and ROAS live in two different systems, and nobody joins them up. Your ad platform counts an order the moment it is placed. Your refund arrives three weeks later. It is processed by a team that does not report to marketing. The campaign that bought the sale never hears about it. So the media buyer optimises against a number that treats a returned jacket as revenue. Nobody is lying, and the reporting is still wrong. This page gives you the arithmetic and the one return figure with a disclosed sample. It also says plainly how returns and ROAS look on my own accounts.
This is for ecommerce operators whose ad reporting and finance reporting have never agreed.
Why returns and ROAS never meet in the same report
Because they are produced by different systems, on different clocks, for different people.
Your conversion event fires at checkout and sends a revenue value to Meta or Google. Refunds happen in the store back end or a helpdesk tool, days or weeks later. Nothing routes that event backwards.
Timing does the rest of the damage. By the time a refund is processed, the campaign has already had its budget raised. That decision rested on the order now going back.
So the buyer scales a winner that was partly imaginary. Nobody in that chain has done anything wrong. That is exactly why returns and ROAS stay unreconciled for years in well run accounts.
Fixing it starts with naming it. Every return figure in your ad platform is gross. Put the word gross next to it in any report you send.
The reconciliation gap, and who owns it
Nobody owns it, and that is the whole finding.
Media buyers optimise to platform revenue because it is the only revenue they can see. Finance reports net revenue after refunds because that is what the bank shows. Both numbers are correct inside their own system.
Ask which one the ad targets were set from, and the room goes quiet. Targets are almost always built on the gross figure, then judged against the net one two quarters later.
One shared monthly number closes most of the gap. Net revenue after refunds, for exactly the same period as the ad spend, handed to whoever sets the bids.
That is a fifteen minute job for a finance team. It changes what the media buyer does with the budget. Very few businesses have somebody whose job description includes it.
Returns and ROAS on a 14% return rate, worked
Round numbers, no client attached, so you can run it against your own.
Take 100 orders at $100 each, bought with $2,500 of advertising. Reported revenue is $10,000, so reported return on ad spend is 4.0x. At a 40% gross margin those orders contribute $4,000, leaving $1,500 of profit after media.
Now apply a 14% return rate. Fourteen orders come back, $1,400 of revenue disappears, and net revenue is $8,600. Net ROAS is 3.44x.
Profit moves further than the multiple does. Eighty-six orders contribute $3,440 against the same $2,500 of media, which leaves $940. Your profit fell by 37% while your headline number fell by 14%.
One detail makes it worse. Advertising for the fourteen returned orders was never refunded. You paid full price to acquire revenue that walked back out.
The apparel problem has no primary source
Everybody knows apparel returns are the worst, and nobody can show you the study.
Search for a category return rate and the same numbers repeat across dozens of pages. None carries a sample size, a date range, a geography or a stated method. Trace them upstream and they stop at another page doing the same thing.
Salesforce is the exception worth citing. Salesforce put returns at 14% of purchases, worth $181B, in holiday data published on 8 January 2026. That is up 10% year on year, across 1.5 billion shoppers in 89 countries.
Read the caveat with it. That sample skews towards enterprise retail, and 14% is an overall figure rather than a category one.
Sizing uncertainty makes a higher apparel rate entirely plausible. Plausible is not measured, so I will not put a number on it, and neither should the pages that do.
Returns and ROAS inside my own accounts
Time for the uncomfortable part, because this article would be dishonest without it.
Returns on every ecommerce account I run are handled by the client’s own in-house team. None of them sends refund data back to me. So no return on ad spend I have ever optimised towards had refunds deducted from it.
That applies to my published figures as well. My €14.18 across 1,715 orders came from an EU fashion client, full year 2022, across Meta and Google combined. That is a gross cost per order. My 1.99 return on a US flag and banner client across June to August 2026 is gross too.
Both are real, both are dated, and both sit above the true net figure by roughly whatever those businesses refund.
Saying so costs me nothing and tells you something useful. Any consultant quoting a return figure without that caveat has solved a problem I have not. More likely they have not thought about it.
What a return costs beyond the refund
Refunded revenue is the visible part, and the rest of the bill is bigger than people expect.
Outbound shipping has already been paid and is rarely recovered. Return postage is often free to the customer and never free to you. Then somebody has to receive the item, inspect it, repack it and put it back into stock.
Payment processing adds another slice, because fees on the original transaction are not always returned when the sale is.
Resale value is the last one. Some goods go back on the shelf at full price. Others come back opened, worn or seasonal, and get sold at a markdown or written off entirely.
None of that appears in an advertising report. All of it comes out of the same contribution you were using to justify the bid.
Whether the platform ever sees a refund
A refund only reaches your ad platform if somebody deliberately sends it there.
Both major platforms can accept adjustments to a recorded conversion, and doing it properly means wiring your order system to report cancellations and refunds against the original event. On every account I have inherited, nobody had done it.
Even where it works, the correction arrives late. A refund processed in week four cannot un-scale a budget decision taken in week two, so the wiring improves your history rather than your bidding.
Broken counting makes this academic anyway. One US automotive equipment ecommerce account in my portfolio reports a return on ad spend of 44.37, on perfectly ordinary delivery metrics of $8.28 CPM and $0.28 CPC. That is a counting fault, not a result.
An account that cannot count a sale correctly has no chance of counting the refund.
Returns and ROAS targets: gross the number up
Adjusting the target is one line of arithmetic, and it is the single most useful thing in this article.
Divide your break-even multiple by one minus your return rate. On a 40% gross margin, break-even is 2.5x, and a 14% return rate lifts it to about 2.91x.
Margin decides how much that hurts. Across my ecommerce clients gross margin runs from 20% to 50%. At the thin end, break-even of 5.0x becomes roughly 5.81x. At the generous end, 2.0x becomes 2.33x.
Use your own return rate rather than a published average, because the only figure that governs your business is the one your customers produce. Pull it from the store, by month, for the same period as the spend.
Then set the bid target on the adjusted number and write the return rate assumption next to it, so the next person knows what it was built from.
Cutting returns is a site job before it is a media job
Advertising can influence the return rate, and the website decides most of it.
Sizing guidance, real measurements, and photography on more than one body type reduce the guesswork that produces a refund. Delivery timing stated honestly at checkout stops the cancellation that follows a late parcel.
Product copy matters just as much. An item that arrives exactly as described comes back far less often than one sold on a flattering photograph.
Media has a smaller lever, and it is worth pulling. Creative that shows the product in realistic use sets a more accurate expectation than creative built purely to stop the scroll.
Measure the two together. A campaign with a lower reported return and a lower refund rate can beat a campaign that looks stronger in the platform, and you will never see that unless somebody reports both.
What to change this week
Three things, and none of them needs a new tool. Each one closes part of the gap between returns and ROAS.
Ask finance for net revenue after refunds, for the last complete quarter, matched to the same dates as your ad spend. Divide it by that spend and compare it with what the platform reported.
Then divide your current break-even target by one minus your own return rate, and update the target you are bidding to.
Finally, pull the refund rate by product category from your store and look at your top spending campaigns beside it. The worst offender is usually getting the most budget, because gross reporting rewarded it.
Working out what an order is genuinely worth is the valuation step this arithmetic sits on. Two dated figures for what a purchase costs sets out my own volumes and periods, gross of refunds like everybody else’s. The acquisition number hiding inside your blended figure covers the denominator problem, and what a promotion does to your break-even covers the other half of the margin story. Sizing, photography and honest delivery promises are the cheapest way to cut refunds, which is what my web development work handles. To have your gross and net numbers reconciled properly, book a teardown.