Retail Returns Statistics September 2026

Product returns cost U.S. retailers hundreds of billions of dollars every year. As e-commerce keeps growing, so does the volume of goods travelling back up the supply chain — and the cost of handling them. This report compiles the most current, source-verified statistics on U.S. retail returns for journalists, analysts and business teams.

Every figure below names the release it came from and the date that release was published. Where two credible sources give different answers — as they do on the size of the market — both are here, with an explanation of why.

Key findings at a glance

• U.S. retail returns are estimated at $849.9 billion in 2025, a 15.8% return rate — down from 16.9% in 2024 (NRF / Happy Returns, 2025 Retail Returns Landscape, Oct 2025).
• E-commerce return rates remain higher, at an estimated 19.3% of online sales in 2025 (NRF / Happy Returns, Oct 2025).
• 9% of all returns in 2025 are classified as fraudulent, and 85% of retailers are deploying AI to detect and prevent return fraud — though only 45% consider those tools effective (NRF / Happy Returns, Oct 2025).
• Gen Z consumers (ages 18–30) averaged 7.7 online returns in the past 12 months, more than any other generation (NRF / Happy Returns, Oct 2025).
• Counting return transactions rather than surveying retailers, Appriss Retail puts 2025 returns at $706 billion — of which $68 billion, 9.6% of return volume, came back with no receipt (Appriss Retail, 2026 Total Retail Loss Benchmark Report).
• Handling returns costs the industry $211–212 billion, roughly 30% of the value of the item coming back (Appriss Retail, 2026).
• 72% of retailers now charge for at least one return option, up from 66% a year earlier (NRF, 2025).
• No full-year 2026 figure exists yet. The NRF publishes its annual returns report in October, so any 2026 return rate in circulation today is a forecast (NRF).

U.S. retail returns by year, 2020–2025

The NRF has published an annual U.S. returns estimate every year since 2020. Here is the whole series — and, just as importantly, who produced each figure.

U.S. retail returns by year, 2020–2025, as the NRF reported each one

Value of merchandise returned to U.S. retailers, per calendar year

1. 2020$428B
2. 2021$761B
3. 2022$816B
4. 2023$743B
5. 2024$890B
6. 2025$849.9B

Year	Returns	Return rate	Online return rate	Reported by
2020	$428B	10.6%	—	NRF & Appriss Retail (Jan 2021)
2021	$761B	16.6%	20.8%	NRF & Appriss Retail (Jan 2022)
2022	$816B	16.5%	16.5%	NRF & Appriss Retail (Dec 2022)
2023	$743B	14.5%	17.6%	NRF & Appriss Retail (Dec 2023)
2024	$890B	16.9%	—	NRF & Happy Returns (Dec 2024)
2025	$849.9B	15.8%	19.3%	NRF & Happy Returns (Oct 2025)
2026	Not yet published	—	—	Due Oct 2026

Three things to take from that table before quoting any row of it.

There is a documented break at 2023. The NRF states plainly that, because the methodology changed, its 2023 numbers cannot be compared with the 2022 report. So the apparent fall from 16.5% to 14.5% is partly a change in measurement rather than purely a change in shopper behaviour. Almost nothing that quotes this series mentions it.

The research partner changed after 2023. The 2020–2023 reports were produced with Appriss Retail. From 2024 the NRF has published with Happy Returns, a UPS company — and Appriss began publishing its own separate estimate. That is the origin of the two competing figures for the same years, and the next section takes it apart.

2020 is the outlier, not 2021. The 10.6% rate was recorded in a year when many stores were shut for months and a great deal of merchandise was never bought in person to begin with. The NRF attributed the 2021 increase primarily to overall sales growth rather than to a change in return behaviour.

Set 2020 aside and the rate has sat in a 14.5%–16.9% band for five years, with the dollar value between $743 billion and $890 billion. Both the rate and the total fell in 2025 — worth noting, because the prevailing narrative is that returns rise every year. They did not in 2025.

Why two credible sources disagree

The NRF figure and the Appriss figure are built in opposite directions.

US retail returns for 2024 and 2025, as each source counts them

Value of merchandise returned to U.S. retailers, per calendar year

1. 2024 · NRF$890B
2. 2024 · Appriss$685B
3. 2025 · NRF$849.9B
4. 2025 · Appriss$706B

The NRF and Happy Returns number is top-down. They survey retailers — in the 2025 edition, 358 ecommerce professionals at US merchants above $500 million in revenue — ask each what share of annual sales comes back, and apply the resulting rate to total US retail sales. It answers "what do retailers say their return rate is", which is the right question if you want to know what the industry believes about itself, and it inherits whatever total retail sales figure is used as the multiplier.

The Appriss Retail number is bottom-up. Their benchmark is built on point-of-sale return transactions across roughly 250 million unique customer identifiers, extrapolated to the market. It answers "how many return transactions actually happened", which is the right question if you want to count events rather than opinions, and it inherits whatever coverage bias their client base has — it is strong in categories where Appriss is deployed and weaker where it is not.

So the spread is a methodology gap, not an error. Two practical consequences:

• Do not mix them. A return rate from one source against a dollar total from the other produces a sales base that never existed. This happens constantly in secondary coverage.
• Match the source to the claim. Arguing about retailer sentiment, planning or policy? Use the NRF. Arguing about transaction volume, fraud mechanics or anything per-return? Use Appriss.

Appriss's own year-on-year series makes the point again: they had 2024 returns at $685 billion and a 13.21% rate, against the NRF's $890 billion and 16.9% for the same year. The gap is stable, which is what you would expect from two consistent methods rather than from one of them being unreliable.

Return rates by channel

Channel is the sharper split. Online returns ran at 19.3% against 15.8% across all retail in 2025 — roughly a fifth of everything bought online comes back. The structural reason is fit: you cannot try a garment through a screen, so shoppers buy two sizes and return one. Appriss's channel mix shows where those returns physically land:

Return journey	Value, 2025	Share
Bought in store, returned in store	$367B	52%
Bought online, returned in store	$208B	29%
Bought online, returned online	$131B	19%

Nearly a third of all returned value is bought online and carried into a shop. That single row explains why "ecommerce returns" and "store returns" cannot be managed as separate problems, and why the counter staff absorbing them are rarely the team whose budget the return is charged to.

It is also why the barcode is still on the receipt. When an online order is returned to a counter, the printed slip is what lets the till recall the original sale — which is the job that barcode does. In our measurements of 632 receipts, 37.3% carried one and not a single receipt carried a QR code.

Holiday returns run above the annual rate: retailers expected 17% of holiday sales to be returned, concentrated in January. The industry nickname for it is Returnuary.

What returns actually cost a retailer

The dollar value of returned merchandise is not the cost of returns. The cost is what it takes to get an item back, inspect it, decide what it is now worth, and route it somewhere — and that is the figure most reports leave out.

Appriss puts industry-wide returns processing at $211–212 billion, which they characterise as roughly 30% of the item's value. On a $60 return, about $18 goes on handling the reversal, before any markdown on resale.

That number is why the return-fee wave happened. In the NRF's retailer survey, 40% of retailers charging for returns cited the rising cost of processing them and another 40% cited carrier shipping costs. It is also why returnless refunds exist at all: below some threshold, telling the customer to keep a $12 item is simply cheaper than paying to get it back, inspect it and write it down.

Set against shrink, returns are the larger line. Appriss's 2026 benchmark totals $796 billion of retail loss for 2025 — $706 billion of returns and about $90 billion of shrink. Loss prevention attention has historically gone to the smaller half.

Returns without a receipt

This is the figure we came into the data looking for, and it is bigger than we expected.

Return type	Value, 2025	Share of volume
Receipted	$638B	90.4%
Non-receipted	$68B	9.6%

$68 billion of US merchandise came back in 2025 without proof of purchase. About one return in ten.

Non-receipted returns are the pressure point in the whole system. Without the transaction record a retailer cannot verify the price paid, the date, the store, or whether the item was bought at all — so the controls that exist are proxies. In practice that means ID capture and a per-person cap over a rolling window, refund at the lowest recent selling price rather than the price paid, store credit instead of cash, and manager approval above a threshold.

A material share of that $68 billion is not fraud at all — it is thermal paper that faded, or a receipt that went in the bin before the item turned out to be faulty. If that is the position you are in, the order to work through is asking the merchant for a duplicate first; most chains can reprint from the card used, and a reprint is worth more at the returns desk than any reconstruction.

It also explains the shape of return fraud. The overwhelming majority of fraudulent returns are attempts to make a non-receipted return look receipted, or to make a receipted return cover something it does not. That is a document problem before it is a merchandise problem, which is why the structure of the document matters at the returns desk. How receipts are designed sets out what a returns clerk is actually reading when they check one.

Abuse and fraud are not the same line item

The single most misquoted returns statistic is the fraud rate, because two credible numbers are in circulation and they are counting different behaviour.

9% of returns are fraudulent — retailers' own estimate, from the NRF's 2025 survey. This is what retailers perceive.

2% of returns are fraud and 12% are abuse — Appriss, from transaction data, splitting the category:

Category	Value, 2025	Share of returns	What it means
Abuse	$86B	12%	Excessive but not illegal — wardrobing, serial bracketing, exploiting a policy as written
Fraud	$14B	2%	Malicious — stolen goods, stolen tender, empty boxes, counterfeit swaps
Both	$100B	14.2%	The preventable total

Abuse is six times the size of fraud. That inverts the usual framing, and it changes what you would do about it. Fraud is a detection problem: catch it, refuse it, refer it. Abuse is a policy design problem — the customer is doing exactly what the published terms permit, at a volume the terms did not anticipate. You cannot prosecute your way out of bracketing.

The tactics retailers report seeing, from the NRF's 2025 survey: overstated return quantities (71%), empty-box or "box of rocks" returns (65%), and decoy or counterfeit items substituted for the real one (64%). Appriss's earlier 2024 research put wardrobing at 60% of retailers, returns on fraudulent or stolen tender at 55%, and returns of stolen merchandise at 48%.

85% of retailers now run AI or machine learning against returns fraud — but only 45% of them consider those tools effective. That gap is the honest state of the art.

Policies got stricter, and shoppers noticed

Returns policy tightened across 2025 and consumers registered it. In a Savings.com survey of 1,000 US adults, 79% had met at least one stricter return rule in the preceding year:

Stricter rule encountered	Share of shoppers
Returnless refund — "keep the item"	38%
Store credit only, no cash refund	25%
Return window under 14 days	19%
Return fee or restocking charge	12%
ID required to return	9%
"All sales final"	9%

On the retailer side, 72% now charge for at least one return option, up from 66% a year earlier.

The returnless row is the interesting one, because it is a cost-control measure that shoppers experience as generosity. Roughly a third of retailers already offer returnless refunds and about another 28% plan to. It reads as a gift and functions as a write-off.

Heading into the 2025 holiday season, retailers surveyed by ReturnPro — 500 senior executives at companies above $500 million — were tightening further: 75% said return fraud worsens over the holidays, and 65% were putting measures in place to monitor or prohibit excessive return patterns. The NRF found the holiday plan to be third-party logistics support (49%), seasonal returns staff (43%) and, going the other way, extended return windows (37%).

How consumers behave

Returns policy is a purchase-stage decision, not a post-purchase one. That is the finding with the most commercial weight in the whole dataset.

• 82% say free returns matter when deciding where to shop, up from 76% in 2024
• 81% read the return policy before buying
• 76% prefer an instant refund or exchange
• 71% would not buy from a retailer again after a bad return experience, up from 67%
• 80% would tell friends and family about a bad return experience
• 90% say a generous policy increases their loyalty to a brand

Volume, by generation and by income:

Group	Online returns per year
Gen Z (18–30)	7.7
All shoppers, average	13
Household income $125,000+	22
Household income under $35,000	10

The Gen Z and the all-shopper figures come from different surveys with different definitions, so read them side by side rather than against each other — the NRF's 7.7 counts online returns by Gen Z specifically, while Savings.com's 13 is a self-reported annual average across all adults. The income gradient inside that second survey is the cleaner comparison, and it is steep: the highest income bracket returns more than twice as often as the lowest.

Why things come back, from the same survey: wrong size or fit 52%, damaged or defective 44%, misleading description 23%, changed mind 18%, wrong item sent 16%, found a better price 6%. Fit alone is over half.

And on the behaviour retailers worry about: roughly two-thirds of consumers admit to at least one costly return behaviour, and 45% think bending the truth on a return is acceptable.

One number cuts the other way. 60% of shoppers have kept something they did not want because returning it was too much hassle, at an average abandoned value of $50. Friction suppresses returns — and suppresses repurchase at the same time, which is what the 71% figure above is measuring.

Refund method is the last piece, and it is moving. As more of a retailer's inbound payments arrive over instant rails rather than cards, the reversal path changes with them — a card refund has a defined chargeback route and a bank-to-bank transfer does not. We cover what has happened to those rails in the peer-to-peer payment report.

What we could not verify

• Any full-year 2026 return rate. The NRF's next annual report lands in October 2026. The 2026 rates circulating now are projections, and the ones we traced did not name a method.
• Appriss's extrapolation base. The benchmark states 250 million unique customer identifiers but not the number of retailers or their category mix, so we cannot assess coverage bias in the $706 billion figure.
• Exact fee amounts by retailer. Widely reported as $4–12 per return. These change without announcement and we could not confirm a current schedule for any named retailer from that retailer's own terms on one date, so we have not published a table of them.
• The returnless-refund adoption share. The "about a third of retailers, another 28% planning to" figure traces to a vendor survey we could not read in full. Treat it as directional.
• Reconciling the two fraud rates precisely. The NRF's 9% is retailer perception and Appriss's 14.2% is measured preventable loss. They are not the same quantity and we have not tried to bridge them arithmetically.

<a href="https://www.repceipt.com/reports/retail-returns-statistics">Retail Returns Statistics September 2026</a> — Repceipt