NRF and Happy Returns projected that 19.3% of online sales would be returned in 2025. Their release dated October 15, 2025 also estimated returns at 15.8% of annual retail sales overall, worth $849.9 billion. These are attributed, dated estimates, not observations newly collected by TrueMargin. [1]
The useful benchmark is the one whose definition matches your question. A fashion writer, merchant and warehouse operator may all ask about returns while needing different measures. Returned sales value, orders with a return and units sent back are not interchangeable. Neither is an online-sales estimate interchangeable with one covering retail as a whole.
Ecommerce and retail return estimates, with their scope
Keep the source year attached when citing these figures. The forecast status is part of the finding: a release issued before the year ends does not establish a final audited outcome for that year.
| Source period | Population or measure | Reported estimate | Interpretation |
|---|---|---|---|
| 2025 | Online sales | 19.3% | Projected share returned; not an order-count rate. [1] |
| 2025 | Annual retail sales overall | 15.8%; $849.9 billion in merchandise | Blended retail estimate, not in-store-only returns. [1] |
| 2024 | Annual retail sales overall | 16.9%; $890 billion in merchandise | Projection in the release dated December 5, 2024. [2] |
Do not turn the online percentage into a count of returned items or orders. The source expresses the estimate as a share of sales, and its retail total is stated in merchandise value. An expensive returned item and a cheap returned item affect a value-based rate differently, even though each is a unit. A partial refund can also affect sales value without representing a complete returned order.
Who was surveyed, and when?
For its 2025 research, NRF describes a summer survey of 358 ecommerce professionals working for US merchants with revenue above $500 million. A separate survey covered 2,006 consumers who had returned an online purchase during the preceding twelve months. Retailer estimates and consumer answers are different evidence populations. [1]
Those samples are not a census of every ecommerce store or every order. They do not establish the expected rate for a particular small business, product or country. The public release does not provide enough detail to reconstruct a transaction-level sample, confidence interval or category-specific estimate here. Those fields remain unknown rather than being inferred.
The earlier 2024 release describes a fall survey involving 249 ecommerce and finance professionals from large US retailers and a separate group of 2,007 consumers who had made an online return during the previous year. Differences in respondents and timing matter when comparing editions. [2]
Why category and channel averages can disagree
A benchmark compiled from apparel orders answers a different question from a survey covering several retail verticals. A figure about people who returned a purchase is different again: the share of shoppers who returned something does not tell you the share of revenue refunded. Check whether a category refers to the retailer's main business, the actual returned product or a shopper's reported behavior before combining sources.
The NRF releases used here do not establish comparable category-specific ranges for fashion, footwear, electronics or beauty. Use a category study only when its period, geography, sales channel and counting method match the comparison you want to make.
Likewise, subtracting a blended retail percentage from an online percentage does not produce the in-store rate. You would need compatible sales totals and return values for each channel. Compare like periods, geography and accounting definitions before interpreting a difference as a trend.
Calculate a rate that matches your decision
- Sales value: state whether the numerator is returned merchandise value or cash refunds, and match it to the original sales cohort.
- Orders: define whether any returned line makes an order a returned order. Do not count the same order repeatedly because several items came back.
- Items: count units returned against units sold. Preserve the product mix when comparing categories.
- Timing: attach returns to their original purchase cohort and allow time for the return window. Dividing this month's refunds by this month's unrelated sales can distort the result.
- Outcomes: separate cash refunds, exchanges, store credit, cancellations and unsuccessful return requests. The operational event and financial result can differ.
Keep the definition next to the chart or table. If the metric changes, record the change before comparing periods. A warehouse can need an item-count measure for capacity planning while finance needs refunded value and the costs left behind.
Refunded revenue is different from the cost of a return
A refund reduces sales. Reverse shipping, inspection and payment fees are expenses. Recoverable merchandise affects inventory and cost of goods sold; it does not create another sale merely because it comes back. Combining these effects into a single "net revenue" line obscures what actually changed.
Hypothetical example: separate sales, expenses and recovered inventory
Assume 100 single-item orders at $100 each, giving $10,000 gross sales. Suppose 20 orders receive full refunds, so refunded value is $2,000 and retained sales are $8,000. These chosen inputs are not an industry return-rate benchmark.
Also assume product cost of $30 per item, outbound shipping and nonrefundable payment costs of $8 per original order, extra return handling of $10 per returned order, and $2,000 in advertising. Every returned item is assumed recoverable at its original cost, with no markdown or write-off. All figures use the same settled cohort; overhead, salaries, financing and income taxes are excluded.
| Line | Calculation | Amount |
|---|---|---|
| Gross sales | 100 × $100 | $10,000 |
| Refunds | 20 × $100 | −$2,000 |
| Retained sales | $10,000 − $2,000 | $8,000 |
| Cost of products kept | 80 × $30 | −$2,400 |
| Outbound and payment costs | 100 × $8 | −$800 |
| Extra return handling | 20 × $10 | −$200 |
| Advertising | Chosen cohort spend | −$2,000 |
| Contribution before excluded costs | $8,000 − $2,400 − $800 − $200 − $2,000 | $2,600 |
The 20 returned items restore $600 of inventory at the assumed $30 cost each. That recovery is why product cost charged to the kept sales is $2,400 rather than the $3,000 initially shipped. It is not extra revenue or a cash receipt. Cash remains tied up until the stock is sold, and actual damage or markdowns would change the calculation.
Build the same reconciliation using your own contract, invoices and order records. Our profit margin calculator and Shopify profit calculator can help check the inputs; keep return-related costs explicit in your supporting ledger.
Use return reasons to choose the next operational test
Start with the reason attached to each returned product. For fit complaints, review measurements and fit guidance. For description mismatches, inspect the actual listing and images. For transit damage, compare packaging and carrier records. These are testable actions, not a ranked promise that one tactic will deliver a particular reduction.
Evaluate a policy using retained contribution and customer outcomes as well as return volume. A return fee can change who buys, who requests a return and how the cost is divided. An exchange still requires handling, and store credit creates an obligation to supply something later. Neither should be treated as costless retained cash.
Methodology and limits
This analysis compares the named NRF releases and separates their reported period, respondent population, forecast status and sales denominator. It is a source reconciliation, not a new survey, transaction dataset or claim that TrueMargin discovered the underlying statistics. The date above records when this analysis was completed; it does not change the source years.
The cost example is constructed independently to explain the accounting distinction. Its prices, quantities, recovery assumption and expenses are hypothetical. It cannot establish typical processing costs or prove that one retailer's returns are acceptable. For a real comparison, substitute actual settled returns, inventory recovery, fees and operating costs.
Frequently asked questions
What is the average ecommerce return rate?
NRF and Happy Returns projected that 19.3% of online sales would be returned in 2025. This is a dated US retail estimate, not a newly measured 2026 rate or a universal benchmark for every store. Source
Is the overall retail return rate the same as the online rate?
No. The 2025 NRF release gives 15.8% for annual retail sales overall and 19.3% for online sales. Those measures cover different sales populations; the overall figure is not an in-store-only rate. Source
What is the return rate for fashion ecommerce?
The cited NRF releases do not establish one verified fashion, footwear or beauty return-rate range. Use a category-specific study only when its period, geography, sales channel and counting method match the comparison.
How should I calculate my store's return rate?
Choose a denominator first. A sales-value measure divides the value returned or refunded by the matching original sales value. An order-based measure counts orders with a qualifying return, while an item-based measure counts returned units. Label partial refunds and exchanges separately.
Are refund processing costs deducted from net sales?
Refunds reduce sales; return shipping, inspection and other processing costs are expenses. Keep those lines separate. The value of stock recovered for resale also matters when calculating contribution, and recovered inventory is not a cash receipt.
Sources
- NRF and Happy Returns: Consumers Expected to Return Nearly $850 Billion in Merchandise in 2025 — release dated October 15, 2025; estimates and survey population described in the release.
- NRF and Happy Returns Report: 2024 Retail Returns to Total $890 Billion — release dated December 5, 2024; prior-period projection and respondent scope.