Return fraud statistics used to be a curiosity — a footnote in retail reports, cited once a year at NRF Big Show. In 2026 they are an operating budget line. The share of online returns classified as fraudulent has roughly doubled in four years on both sides of the Atlantic, driven by three compounding shifts: more returns overall (a function of pandemic-era buying habits that never reverted), professionalised fraud networks running at industrial scale, and the arrival of generative AI as a cheap way to fabricate damage evidence.
This article is the benchmark reference. Numbers are rounded for readability but the source for every figure is named; where 2026 itself has not yet produced an industry report, we lean on the 2025 data and the growth trajectory between 2022 and 2024. Update this article in Q3 when NRF, Appriss, and bevh release their 2025-full-year reports.
Global Headline Figures
| Region | Total returns processed | Fraud share of returns | Estimated fraud loss | Source |
|---|---|---|---|---|
| United States | $890B | 13.7 % | $122B | NRF Consumer Returns 2024 |
| European Union | ~€180B (est.) | ~14 % (est.) | ~€25B | bevh DACH + Statista EU aggregation |
| United Kingdom | £80B | ~13 % | ~£10B | BRC + Statista UK |
| DACH (DE/AT/CH) | ~€55B | ~15 % | ~€8–10B | bevh 2025 + aggregation |
| Rest of world | ~$200B (est.) | ~11 % | ~$22B | Claimscan aggregation |
The US is the largest and best-measured market. NRF's 2024 report, conducted with Appriss Retail and a sample of 141 retailers, is the most-cited source in the industry and the reason US numbers are a step more confident than the rest. The European picture is more fragmented — bevh (the German e-commerce association) reports headline return volumes, Statista aggregates by category, and the Forschungsstelle Retourenmanagement at the University of Bamberg publishes annual fraud-rate samples. Our EU estimate is a weighted blend.
The US Trajectory: 2019 to 2024
The US number is the one to watch, because the five-year growth rate previews the next two years for Europe.
| Year | Returns processed | Fraud % | Fraud loss | Online return rate |
|---|---|---|---|---|
| 2019 | $309B | 5.9 % | $18.4B | 18.1 % |
| 2020 | $428B | 6.0 % | $25.3B | 20.8 % |
| 2021 | $761B | 10.6 % | $80.6B | 20.8 % |
| 2022 | $816B | 10.4 % | $84.9B | 16.5 % |
| 2023 | $743B | 13.7 % | $101B | 17.6 % |
| 2024 | $890B | 13.7 % | $122B | 17.0 % |
Two things jump out. First, the fraud rate effectively doubled from 2019 to 2023 (5.9 % → 13.7 %). Second, while the total return rate compressed slightly in 2024 (store policy tightening worked), the fraud share did not. Fraud is now a structurally larger chunk of the return volume.
Per NRF's 2024 segmentation, the dollar-weighted fraud loss splits as:
- Wardrobing / BORIS / BOPIS abuse: 38 % of fraud loss
- Chargeback fraud (first-party fraud): 22 %
- Empty-box and item-swap: 14 %
- Return-of-stolen-merchandise: 11 %
- Collusive-employee fraud: 7 %
- Photo-based fraud (photoshopped damage, AI-generated): 8 % and fastest-growing
Photo-based fraud is the category that didn't exist as a standalone line item in the 2019 NRF report. In 2023 it was measured for the first time at 3 % of fraud loss; in 2024 at 8 %. It is the fastest-growing category in the benchmarks.
The European Picture
Europe does not have a single NRF-equivalent report. What we do have: bevh's annual return-volume data for Germany, Austria, and Switzerland; Statista's category breakdowns across the EU-27; the Forschungsstelle Retourenmanagement at Bamberg University publishing fraud-rate samples; and a handful of consulting reports (Accenture, Deloitte) that aggregate across the continent.
| Country | Online return rate | Fashion return rate | Est. fraud % | Source |
|---|---|---|---|---|
| Germany | 53 % | 70 % | 14 % | bevh 2024 + FH Bamberg |
| Austria | 48 % | 65 % | 13 % | Handelsverband AT + aggregation |
| Switzerland | 41 % | 58 % | 12 % | KOF + aggregation |
| United Kingdom | 29 % | 45 % | 13 % | BRC + Statista UK |
| France | 25 % | 38 % | 11 % | Fevad + Statista FR |
| Netherlands | 36 % | 48 % | 13 % | Thuiswinkel.org + aggregation |
| Italy | 17 % | 24 % | 9 % | Netcomm + Statista IT |
| Spain | 19 % | 28 % | 10 % | Adigital + Statista ES |
The German number — 53 % overall return rate online, 70 % in fashion — is famously the highest in the OECD. Three structural reasons: payment-on-account (Kauf auf Rechnung) dominant, statutory 14-day withdrawal right strictly enforced, and cultural normalisation of "order three sizes, keep one". The fraud percentage is close to the EU average but the absolute loss is larger because the return denominator is larger.
The Italian and Spanish numbers are the mirror image: lower return rates driven by cash-on-delivery and card-up-front payment mix, with fraud at the lower end of the European range.
Fraud by Product Category
Category risk is the planning number — which SKUs to prioritise for photo-forensic review, hem tags, or seal validation.
| Category | Return rate | Fraud rate (of returns) | Dominant pattern | Avg. fraudulent claim |
|---|---|---|---|---|
| Fashion — womenswear | 62 % | 10 % | Wardrobing + worn-returned | €95 |
| Fashion — menswear | 48 % | 8 % | Wardrobing + size arbitrage | €78 |
| Electronics | 12 % | 15 % | Item-swap + box-swap + empty-box | €420 |
| Beauty | 8 % | 6 % | Used-sent-back + decanted product | €55 |
| Home & garden | 18 % | 8 % | Photoshopped damage + wardrobing | €140 |
| Sports & outdoors | 19 % | 9 % | Event wardrobing + worn-shoes | €110 |
| Jewellery | 22 % | 18 % | Item-swap + worn-event | €380 |
| Books & media | 6 % | 4 % | Chargeback + digital resale | €25 |
Three observations worth acting on. First, the highest-fraud-rate categories (jewellery, electronics) are not the highest return-rate categories (fashion). High-value, low-volume categories attract professional fraud; high-volume fashion attracts amateur abuse. Second, the average fraudulent claim ranges from €25 to €420 — the value-weighted risk sits firmly in electronics and jewellery. Third, each category has a dominant pattern, which means defence is tractable: you do not need to defend against everything simultaneously, you need to defend fashion against wardrobing and electronics against item-swap.
The Fastest-Growing Fraud Patterns
Not all patterns are growing at the same rate. Tracking growth is how you avoid over-investing in yesterday's fraud.
| Pattern | 2023 share of fraud | 2024 share | Growth rate | Why |
|---|---|---|---|---|
| AI-generated evidence | 1.2 % | 4.8 % | +300 % | Generative image tools cheap and ubiquitous |
| Photoshopped damage | 2.1 % | 3.9 % | +86 % | Established practice, rising with AI-tool normalisation |
| Chargeback abuse (first-party) | 17 % | 22 % | +29 % | Card-issuer dispute process frictionless by design |
| Social-media return tutorials | 0.4 % | 1.5 % | +275 % | TikTok how-to content for bricking, swapping, wardrobing |
| Organised return rings | 4 % | 5 % | +25 % | Professional operators scaling across retailers |
| Wardrobing (classic) | 34 % | 31 % | -9 % | Absolute loss flat; share declining as others grow faster |
| Empty-box | 11 % | 14 % | +27 % | E-commerce share of returns still rising |
Two of the three fastest-growing patterns are evidence-fabrication patterns. That is the shift this article is most interested in. Five years ago the typical fraudulent claim was behaviour-based (wardrobing, keeping-and-disputing). Today a growing share is documentation-based — photographs that lie, AI-generated "damage", reverse-image-searched stock photos submitted as personal evidence. The defensive stack for a 2026 CS team therefore looks different from a 2020 one: less focus on policy-based wardrobing defence, more on photo forensics.
The Repeat-Offender Concentration
A less-reported but operationally important statistic: fraud is heavily concentrated. Both NRF and Appriss consistently find that roughly 20 % of fraudulent returners account for 60–70 % of the loss.
| Decile of fraudulent returners | Share of fraudulent claims | Share of fraud loss | Avg. claim value |
|---|---|---|---|
| Top 10 % | 18 % | 38 % | $340 |
| Top 20 % | 29 % | 61 % | $280 |
| Top 50 % | 58 % | 86 % | $195 |
| Bottom 50 % | 42 % | 14 % | $45 |
The operational implication is stark: detecting and blocking the top 10 % of repeat fraudulent customers closes more than a third of total loss. This is why cross-tenant pHash matching, repeat-claim flagging, and retailer-consortium data sharing (even imperfectly) have disproportionate impact.
What Merchants Are Actually Doing
NRF's 2024 retailer-survey section — a separate set of questions from the fraud-loss estimation — asks retailers what they spend on return-fraud defence.
| Defence mechanism | % of retailers using | Median annual spend | Self-reported effectiveness (1–5) |
|---|---|---|---|
| Tighter return policies (shorter windows, restocking fees) | 78 % | $0 (policy change) | 3.2 |
| CRM-based repeat-claimant tagging | 66 % | $12k | 3.7 |
| Manual photo review (CS human) | 61 % | $50k labour | 2.9 |
| Third-party chargeback-defence vendor | 42 % | $85k | 3.4 |
| ID verification for high-value returns | 29 % | $28k | 3.8 |
| Photo-forensic software | 11 % | $18k | 4.1 |
| Organised-fraud consortium data sharing | 14 % | $22k | 4.0 |
Two patterns: the cheapest defences (policy tightening) are the most-used but have the lowest reported effectiveness. The most-effective defences (photo-forensic software, consortium data sharing) are the least adopted, largely because they are relatively new categories. Photo-forensic software adoption doubled from 5 % in 2022 to 11 % in 2024 and is on a clear adoption curve.
The 2026 Outlook
Three trends we expect to compound through 2026:
- AI-generated evidence will be the single largest category of new fraud growth. Generative image tools that can produce photorealistic damage photos have collapsed in cost from $20+ per image in 2023 to under $0.05 per image in 2026. The economic floor for fraud has fallen to near-zero.
- Return rates will continue to compress slowly (policy tightening works) but fraud rates will continue to rise as a share of returns. The net effect on absolute fraud loss is approximately flat in the US and modestly rising in Europe.
- Detection technology will become the differentiator. The retailers that invest in photo forensics, cross-tenant hashing, and chargeback-defence automation will pull ahead; those relying on policy tightening alone will see the loss ratio climb because the fraud patterns that policy addresses (wardrobing) are growing slower than the ones it does not (evidence fabrication).