Adam Torkildson
Returns cost more than the refund, and they land in the wrong month. Those two facts together explain why an ecommerce P&L can look healthy while the bank balance drifts down. The National Retail Federation and Happy Returns, in the returns report they released in October 2025, estimated that 19.3 percent of online sales would be returned in 2025, against 15.8 percent across retail overall. At roughly one in five online orders, the accounting treatment of returns stops being a rounding question. Here are five specific ways it distorts what you think you earned.
1. The return arrives after the sale is already in the books
A unit sold on February 26 and returned on March 14 puts revenue in one month and the reversal in the next. On a monthly P&L, February looks better than it was and March looks worse than it is.
For a business with steady volume, this mostly washes out. For a seasonal one, it does not. December sales get reversed in January, which is why so many sellers see a strong Q4 followed by a January that reads as a collapse rather than as the return tail of a good quarter. The NRF report found that retailers expect 17 percent of holiday sales to be returned, consistent with previous years.
The fix is a returns reserve: an estimate of the returns you expect against sales already recorded, booked in the same period as the sale. Most sellers under $10 million do not carry one, which is defensible, but only if you know you are reading a lagged number rather than a real one.
2. The refund is the smallest part of the cost
Refunding a $48 item does not cost $48. It costs the $48, plus the fulfillment fee you already paid to ship it, plus whatever portion of the marketplace commission is not credited back, plus the cost of getting it back, plus the labor to inspect and restock it.
Sellerboard, which builds profit analytics for Amazon sellers, breaks its return cost view into exactly these components: the refunded amount, refundable and non-refundable FBA fees, refund commissions, and lost cost of goods on unsellable returns. That breakdown exists because sellers kept discovering that the fee side was larger than the refund side on low-price items.
Marketplace fee schedules change by category and by size band, so check the current schedule for your own before assuming a percentage. The general shape holds: fulfillment fees are the least likely to come back to you.
3. A returned unit is not automatically inventory again
When a return comes back damaged, opened, or seasonally dead, its cost has to leave inventory and hit the P&L. When it comes back sellable, the cost goes back onto the balance sheet.
Sellers who treat every return as restocked carry inventory that does not exist. Sellers who treat every return as a write-off understate their asset base and overstate COGS. Both errors compound quietly across a year, and both surface at the same moment: a physical count that does not match the ledger.
This is where the accounting method starts to bind. IRS Publication 538 states that an inventory is necessary to clearly show income when the purchase or sale of merchandise is an income-producing factor, and that a business required to account for inventory must use an accrual method for purchases and sales. A small business taxpayer exception exists for filers averaging $26 million or less in gross receipts over the three prior tax years who are not tax shelters, though even they must use a method that clearly reflects income. Guessing at the disposition of returned units is not that method.
4. Blended margin hides the SKUs doing the damage
A 19 percent overall return rate is an average across items that come back at 4 percent and items that come back at 45 percent. The 45 percent SKU is frequently one that looks profitable on a blended report and is not profitable at all once returns are attributed to it.
Apparel is the obvious case, and the NRF report identifies the behavior driving it: close to two thirds of consumers admit to at least one costly returns behavior, including wardrobing and bracketing, which is ordering multiple sizes with the intent of sending most back. A bracketed order books three units of revenue and delivers one.
Attributing returns to the SKU requires SKU-level cost and fee data, which is why per-item profit reporting exists as a category. Tools including A2X, Sellerboard, Webgility, and ConnectBooks all approach the problem from different angles, but the shared premise is that account-level margin cannot answer a SKU-level question.
5. Some share of your returns are not real returns
The NRF and Happy Returns report puts return fraud at 9 percent of all returns. Among retailers that track such incidents, the report notes increases in overstated quantity of returns at 71 percent, empty box or “box of rocks” returns at 65 percent, and decoy returns such as counterfeit items at 64 percent. Eighty-five percent of surveyed retailers said they were employing artificial intelligence to detect or prevent return fraud.
On the consumer side, the same report found that 45 percent consider it acceptable to bend the truth when making a return, particularly when unsatisfied with the purchase.
Fraudulent returns do not have their own line in most charts of accounts. They land inside refunds and inside inventory shrinkage, indistinguishable from ordinary buyer’s remorse, which means the loss is real but the cause is invisible.
What the numbers do and do not cover
The NRF figures are worth using and worth qualifying. The report’s methodology surveyed 2,006 consumers who had returned at least one online purchase in the past twelve months, and 358 professionals involved in ecommerce at large US merchants with over $500 million in revenue.
That second group is not you if you are a $3 million brand. Large merchants have different return policies, different carrier rates, different fraud exposure, and different negotiating power. Use 19.3 percent as an orientation point and then measure your own, because the only return rate that belongs in your forecast is the one your own settlements produce.
What to change first
Three things, in order of how much they pay back per hour spent.
Put returns in contra-revenue rather than in expenses. Net sales becomes a real line, and the return rate becomes visible without a calculation.
Split the return cost into refund, non-refundable fees, and lost cost of goods on unsellable units. Three accounts, and the answer to “why is margin down” stops requiring a spreadsheet.
Pull a return rate by SKU for the last twelve months and sort it. The top of that list is usually five or six items, and the decision about what to do with them, reprice, change the listing photography, adjust the size guide, or discontinue, is worth more than any accounting change.
None of this reduces returns. It makes them visible, which is the prerequisite for deciding whether the return rate on a given product is the cost of doing business or the reason the business is not making money.
