Fashion’s Return Rate Is Large Enough Now to Fund a Physics Engine

The National Retail Federation and Happy Returns, in a joint report published in October 2025, estimated that 15.8% of all retail sales would be returned in 2025, totaling $849.9 billion. For online purchases the figure ran to 19.3%. Gen Z shoppers, currently aged 18 to 30, averaged 7.7 online returns per person over the prior twelve months, more than any other generation. Eighty-two percent of consumers cited free returns as a major consideration when making a purchase, up from 76% the year before. The retailers surveyed in that same report identified reducing return rates as a top priority for 2026, alongside growing online sales.

Those two goals pull against each other: every percentage point of online volume growth adds to the pool of potential returns without a corresponding reduction in the cost of processing them. The industry has understood this for years and responded with better product photography, expanded size charts, and, for most of the past decade, periodic announcements of virtual try-on pilots. Catches, ASOS, and Google have each concluded something different about where that gap sits.

A startup bets on physics over pixels

In mid-March 2026, AI startup Catches launched its RealFit technology at Nvidia’s GTC conference in San Jose, with luxury brand AMIRI as its first live deployment. According to the company’s March 16 press release on Business Wire, Catches spent two years developing a proprietary GPU-accelerated simulation framework modeled on physical fabrics, producing what it describes as a 1:1 representation of a garment’s weight, structure, drape, and movement, built on Nvidia’s CUDA platform and Omniverse environment. Shoppers enter their measurements and a photograph to generate a digital twin, then toggle between sizes to see exactly how each piece fits before purchasing.

The distinction Catches is drawing is between tools that render clothing onto a body and tools that model how a fabric actually behaves on that body in motion. Ed Voyce, Catches’ founder and CEO, said at the GTC launch that “fit uncertainty lowers online conversions and causes over 50% return rates in some categories.” The startup has raised $10 million from investors including Antoine Arnault, Director of Image and Environment at LVMH; Gary Sheinbaum, former CEO of Tommy Hilfiger; and Sarah Willersdorf, former Head of Luxury at BCG. Catches projects a 10% increase in conversions and a 20- to 30-times return on investment for brand partners, and targets what Voyce describes as “massive reductions” in return rates, though the company has not published a figure from the AMIRI deployment.

Higher average order values in luxury mean the margin recovery from even a small improvement in conversion or return rates is larger per transaction than in mass-market apparel, a unit economics reality that shapes Catches’ decision to start there. The company is launching first with AMIRI and Nanushka, neither of which is owned by LVMH, and has indicated additional brand partnerships across luxury and non-luxury are forthcoming. Nvidia’s VP and GM of AI for Retail and CPG, Azita Martin, confirmed Catches as an independent software vendor on the Nvidia platform, stating the companies are working to bring these capabilities to more fashion brands.

RealFit is not a widget that drops into an existing product page. It is GPU compute running physics simulation at scale, which means integration timelines, infrastructure costs, and technical requirements differ in kind from generative AI tools built on top of existing image models. Technology and merchandising teams evaluating this category will find that difference relevant to scoping both the implementation project and the realistic timeline to measurable return rate data.

The return rate moved at ASOS, and virtual try-on was not the cause

The clearest documented return rate improvement in online apparel over the past year came from ASOS, and it did not involve consumer-facing virtual try-on.

ASOS, in its fiscal year 2025 results published in November 2025, reported reducing its underlying returns rate by approximately 150 basis points year-over-year. The company’s annual report and interim results attribute that improvement to upgraded size guides, more customer reviews, and refinements to its fair use policy, including return rate thresholds that apply fees to customers with the highest return behavior. The interim results also note the use of AI internally to analyze and address the reasons for returns, a back-end application distinct from the consumer-facing virtual try-on the company subsequently launched. ASOS reported adjusted EBITDA of £131.6 million for the fiscal year, a 250-basis-point margin improvement despite a 14% revenue decline.

ASOS launched its virtual try-on feature in February 2026, in partnership with deep-tech startup AIUTA. Per ASOS’s own press release, the tool initially covers around 10,000 products on the ASOS iOS app for select UK and US customers, offering the option to upload a personal photo or select from 20 AI-generated digital models representing a range of body types, sizes, and skin tones, with results loading in four to seven seconds. That launch postdates the period covered by the FY2025 returns improvement, so its contribution to the 150-basis-point reduction is not present in current reporting. ASOS is explicit that the tool is designed to give general guidance and that shoppers still need to consult size guides before purchasing.

The following January, ASOS introduced a returns transparency feature in its UK app, displaying each customer’s personal return rate and alerting them when they approach the thresholds at which fees apply. Customers with a return rate below 70% retain free returns; those above 70% face a £3.95 charge if they keep less than £40 worth of items from an order; those above 80% face an additional £3.95 restocking fee. Ben Blake, EVP Customer and Commercial at ASOS, told TheIndustry.fashion the changes affect only “a small group” of UK customers, with free returns remaining available to the majority.

Zara, owned by Inditex, has run a parallel combination of policy and technology. Inditex began charging for online returns in 2022 across markets including the UK, France, and the US, before extending the policy to Spain. According to Inditex’s FY2025 annual report published in March 2026, the Zara Try-on service has been deployed in 43 markets and registered more than 7 million sessions since its mid-December 2025 launch. Inditex reported a gross margin of 58.3% for FY2025, up 42 basis points, but does not disclose return rate or conversion figures tied to the virtual fitting room. Simeon Siegel, Senior Managing Director at Guggenheim, told CNBC that “there are certainly companies that have absolutely seen benefits” from AI tools in retail, while noting that isolating and measuring those benefits remains the harder problem.

Google bets on surface area over simulation depth

A structural shift affecting how brand and merchandising teams should read this market is the migration of virtual try-on into platforms shoppers already use, where adoption follows search traffic rather than a customer’s willingness to find a dedicated feature.

Google has pursued this direction. TechCrunch reported in October 2025 that Google expanded its virtual try-on capability to Australia, Canada, and Japan and added shoes alongside the apparel categories already available to US shoppers. The company then moved to consolidate the capability into its core product surface: Google’s own Labs page for Doppl confirms the standalone app will shut down on April 30, 2026, with users directed to try-on accessible directly on product listings and apparel image search results. Google announced via its Keyword blog in early April 2026 a new shopping experience in AI Mode, including a virtual dressing room within product search, drawing on a Shopping Graph of more than 50 billion product listings refreshed at more than 2 billion updates per hour.

Catches is building precision fit modeling for brands willing to invest in GPU infrastructure, targeting the pre-purchase decision of a shopper who has already arrived at a product page. Google is embedding a lighter-weight tool into a surface that reaches every apparel shopper who runs a product search, including those who would never seek out a dedicated try-on feature. The implication for performance marketing and attribution teams is that shoppers may now arrive at a brand’s product page having already used Google’s tool, with purchase confidence that existing pre-site attribution frameworks do not account for.

The NRF and Happy Returns research found that 51% of Gen Z shoppers bracket their purchases, ordering multiple sizes with the intent to return, compared to 24% of Baby Boomers. Brands with younger core customers carry this cost more acutely than industry averages suggest, which means the potential value of fit tools is higher for those brands than aggregate return rate data implies. Category managers deciding where in the assortment to prioritize virtual try-on will find the bracketing data points to a specific behavioral driver rather than a cost distributed evenly across age groups or product categories.

What the current evidence does not support is a clean causal line from any virtual try-on implementation to a documented, isolated reduction in returns at scale. ASOS produced 150 basis points of improvement through information and policy changes, with virtual try-on launched separately afterward. Catches has not published return rate data from its AMIRI deployment. Inditex reported 7 million sessions for Zara Try-on but no return or conversion figures attached to the tool. Google’s own Doppl shutdown notice makes no mention of return rates, conversion impact, or purchase metrics, directing users instead to screenshot their saved looks before the app closes.

Conversations On Retail

Conversations On Retail is a gathering place and resource center for retail and CPG executives, built to make it easier to stay current, discover the technologies and solutions shaping the industry, and connect with the people driving it forward.

We publish news, views, and reviews from staff editors, contributing experts, and trusted partners. Some articles are developed internally, while others are submitted by industry contributors or adapted from interviews and recorded conversations with industry leaders.