Roughly one in five online orders comes back. The National Retail Federation, in its 2025 returns report produced with Happy Returns, estimated that 19.3% of online sales would be returned in 2025, against 15.8% across retail as a whole. At 30% gross margin, a 20% return rate consumes most of that margin before anyone counts what it costs to process the returned units.
The headline numbers
The National Retail Federation’s 2025 returns report, published October 15, 2025 with Happy Returns, projected total retail returns of $849.9 billion for the year. That represented 15.8% of annual sales, which the NRF described as broadly consistent with the prior year at 16.9% and $890 billion.
Inside that total, the online figure is the one that matters to a marketplace seller. The NRF put it at 19.3% of online sales. Holiday returns ran higher still, with retailers expecting 17% of holiday sales to come back.
One more number from the same report explains why nobody has fixed this: 82% of consumers said free returns are a major consideration when making a purchase, up from 76% the year before. The cost is not going away because customers are voting for it.
A second count that does not agree
Appriss Retail’s 2026 Total Retail Loss Benchmark Report, drawing on 2025 data, puts total consumer returns at $706 billion. That is $144 billion below the NRF number for what is nominally the same phenomenon.
The gap is methodology, not error. Appriss builds from transaction data across roughly 250 million unique customer identifiers among its own retail clients, plus a consumer survey run in December 2025. The NRF figure comes from a survey of 2,006 consumers and 358 ecommerce professionals at US merchants above $500 million in revenue, projected across the industry. Two defensible approaches, two different answers.
Use one or the other and say which. Anyone presenting a single authoritative returns total is not reading the sources.
What a return actually costs to process
This is the figure most sellers have never priced. Appriss Retail’s 2026 report states that returns cost retailers an average of 30% of an item’s value to process, and puts the industry total at roughly $211 billion to $212 billion. The report itself gives both figures in different sections, which is a good reason to cite the percentage rather than the total.
That 30% sits on top of the revenue you refunded, and it is the figure most sellers leave out of margin models entirely.
Fraud and abuse, separated
The NRF report found that 9% of all returns are fraudulent. It publishes no dollar figure attached to that rate, and multiplying it against $849.9 billion produces a number the NRF never stated.
Appriss breaks the problem in two and does attach dollars. Fraud, meaning fake receipts and stolen merchandise, accounts for 2% of all returns and $14 billion. Return abuse, which Appriss defines as excessive but legitimate returns, accounts for 12% and $86 billion. Together with related categories, Appriss classifies 14.2% of returns, or $100 billion, as preventable loss.
The distinction matters because abuse is the half a seller controls. It responds to return windows, restocking terms and limits on repeat returners, all of which you can change this quarter without anyone’s permission.
Where returns physically land
Appriss reports that in-store returns now account for 81% of all returns, split across bought in store and returned in store at 52% and $367 billion, bought online and returned in store at 29% and $208 billion, and bought online and returned online at 19% and $131 billion.
For a pure marketplace seller with no retail footprint, that last category is your world, and it is the smallest slice of a very large problem. It is also the slice with the highest per-unit reverse logistics cost, because nobody is consolidating your returns into a truck already going back to a distribution center.
The size of the base
Context for all of the above: the US Census Bureau’s Quarterly Retail E-Commerce Sales release for the first quarter of 2026, published May 18, 2026, put seasonally adjusted ecommerce sales at $326.7 billion against total retail of $1,929.0 billion. Ecommerce accounted for 16.9% of total sales in the quarter.
One caveat the Census Bureau states directly and most citations skip: nonemployer businesses were removed from this series in the benchmark revision released in April 2025. Ecommerce percentages you may have quoted from 2023 are not comparable to this one.
Working the math on a single returned order
A $60 eBay order that comes back works out as follows.
eBay’s published fee structure for most categories, as listed on its selling fees help page in August 2026, is 13.6% of the total amount of the sale up to $7,500 per item, plus a per order fee of $0.40 for orders over $10.00 and $0.30 for orders at or under $10.00. eBay states explicitly that the total amount of the sale “includes the item price, any handling charges, any shipping costs collected from the buyer, sales tax, and any other applicable fees.”
So on a $60 item with $8 shipping collected, the fee base is $68, not $60. The final value fee is $9.25, plus $0.40, so $9.65 leaves before you have paid for the product.
Now the customer returns it. You refund $68. Apply the Appriss finding of 30% of item value in processing cost and you are carrying roughly $18 of handling, inspection, restocking and inbound freight against an item you now own again. If it comes back unsellable, the entire cost of goods goes with it.
Run that against a 19.3% return rate on a hundred orders and the arithmetic stops being abstract. Nineteen refunds, roughly $342 of processing cost by the Appriss ratio, and an unknown number of units that cannot be resold at full price.
What to actually measure
Three numbers, tracked monthly, at the SKU level rather than the account level.
First, return rate by SKU, not blended. Blended return rates hide the two or three products doing all the damage. Second, resale rate on returned units, meaning what percentage comes back sellable at full price. Third, net margin after returns, which is the only margin figure that means anything in a category running near 20%.
Most sellers cannot produce those three numbers because returns, refunds and restocked inventory move through three different systems that never reconcile. Tools built for this problem, ConnectBooks among them, post returns and refunds back through the accounting file and adjust inventory in the same pass, which is what makes SKU-level margin after returns computable rather than theoretical.
No software will decide to stop selling the SKU with a 40% return rate. That call sits with whoever writes the next purchase order, and it is usually the one nobody wants to make, because the SKU is often a strong seller by revenue.