Virtual try-on pilots show 40% fewer size returns: 2026 early data

Virtual try-on technology has delivered return reductions averaging 25 to 40 per cent in early commercial pilots, according to data from retailers and platform providers released between late 2025 and mid-2026. The range varies by category, implementation and measurement method, but the direction is consistent: when shoppers see garments on a body that resembles their own before checkout, fewer items come back for fit reasons.

Zalando reported up to a 40% reduction in return rates with its virtual fitting room technology, though the retailer noted the figure came from pilots and was limited by the testing phase. ASOS publicly aligned with a 35% return reduction figure when it launched hybrid virtual try-on with AIUTA in February 2026. Zara reportedly achieved approximately a 10% return rate reduction where its virtual fitting room feature was active after launching progressively from mid-December 2025.

The early data arrives as fashion executives face a stubborn economics problem. Around 70% of shoppers who returned apparel bought online in the last 12 months did so due to size or fit, according to a Coresight Research survey conducted in February and March 2026. Returns cost retailers between 10 and 20 US dollars per item processed, and a 23.4% return rate would translate into approximately 47.1 billion US dollars in returned merchandise for the US apparel market. Virtual try-on addresses the dominant cause, but adoption remains low and results uneven.

What return reduction are retailers actually seeing?

Published pilot results cluster in a 25 to 40 per cent range for return rate reductions, with important caveats around scope and methodology. According to Zalando's June 2026 reporting, its sizing solutions prevented 8% of size-related returns across the business in 2025, while in virtual fitting room pilots, the company reported return-rate reductions of up to 40%. The gap between those two figures illustrates the difference between company-wide averages and controlled tests on specific products.

Zalando reported a 40% drop in returns during an early virtual try on pilot with Levi's, run across multiple European markets. Previous tests with jeans led to a reduction in returns by up to 40%, the retailer stated in October 2024. The pilot allowed customers to virtually try on a curated selection of Levi's products for men and women, launching in 14 European markets.

ASOS unveiled a hybrid virtual try-on experience in February 2026, allowing shoppers to either upload their own image or select an AI-generated model, developed in partnership with AIUTA. Around 10,000 different products were available to try on, initially for selected UK and US customers. The retailer did not publish pilot results but industry sources aligned ASOS with the 35% reduction figure reported by technology providers.

Inditex highlighted its AI-based virtual fitting experience, named Zara Try-on, which since December has let customers create an avatar from their own photos; the technology is deployed in 43 global markets with more than seven million sessions so far as of March 2026. Users submit a selfie and a full-body photograph; AI processes these inputs and generates a personalised avatar in approximately two minutes; Zara reportedly achieved approximately a 10% return rate reduction where the feature is active.

Industry-wide claims suggest a wider band. Industry research by Onix Systems and Forrester associates virtual try-on with a 25 to 40% reduction in fashion returns. McKinsey's study of digital fitting technology documented 25 to 40% return-rate decreases across 14 apparel brands. The consistency of the range across independent sources lends credibility, but the lack of published methodology and the mix of vendor-sponsored research with independent studies means decision-makers should treat the figures as directional rather than guaranteed.

Why do size and fit cause most fashion returns?

Sizing and fit issues are the top reason for online apparel returns; around 70% of shoppers who returned apparel bought online in the last 12 months did so due to size or fit, per Coresight Research. Size and fit was the top reason clothing gets returned, named by 53% of apparel sellers in a Coresight Research survey from 2023, far ahead of colour at 16 per cent and damage at 10 per cent.

The inconsistency stems from three structural problems. First, there is no standardised sizing system across brands and geographies. The industry does not use a standardized sizing system and apparel brands and retailers are unable to capture what the customer really looks like in different styles and sizes of apparel, Coresight noted. Second, online shoppers cannot touch fabric or see drape. Third, many shoppers order multiple sizes with the intention of returning at least one, a practice known as bracketing.

Around one in three online fashion purchases gets returned, and poor fit drives 71% of those returns. Since fit and sizing drive roughly 70% of apparel returns, giving shoppers a visual preview directly addresses the primary cause. Virtual try-on does not fix returns caused by quality issues, colour misrepresentation or buyer's remorse, but it can reduce the largest single category.

What does it cost to process a return?

Processing a single fashion return costs between 17 and 21 pounds or 20 to 30 US dollars, including shipping, warehouse labour, restocking, customer service, and lost sale opportunity. The average cost to process an ecommerce return ranges from 10 to 65 US dollars, depending on the product category and complexity; this includes reverse logistics at 5 to 15 dollars, processing labor at 8 to 15 dollars, restocking at 2 to 10 dollars, and potential write-offs.

Optoro data cited by Coresight puts the typical cost of processing a return at 66% of the product's price, though that figure includes markdown costs and the opportunity cost of lost margin. Optoro's State of Retail Returns puts the industry-standard cost at roughly 30 US dollars per 100 dollars of merchandise returned; that includes return shipping at 6 to 12 dollars, 3PL handling and inspection at 3 to 8 dollars, markdown loss on resale at 8 to 20 dollars, the non-refundable payment fee, and an allocated abuse drag of 3 to 5 dollars.

The fully loaded cost compounds quickly at scale. Apparel and accessories routinely see 25 to 40% return rates; a 25% return rate in apparel costs 375,000 to 625,000 US dollars in processing for a mid-size brand doing meaningful volume. McKinsey's February 2026 analysis found default processing recovers about 50% of product value; for a brand processing 200 returns per month on 50-dollar average items, that 25-point gap is 30,000 US dollars a year.

Which retailers have launched virtual try-on pilots?

A small but growing cohort of large fashion retailers moved virtual try-on from experiment to live deployment between late 2025 and mid-2026. Zalando integrates its body measurement technology with its virtual fitting room, launching in 14 European markets; the pilot allows customers to virtually try on a curated selection of Levi's products for men and women. Since the initial pilot in 2022, over 80,000 customers have used Zalando's virtual fitting room.

ASOS made an initial rollout of its hybrid approach to virtual try-on in partnership with AI fashion platform AIUTA with access to around 10,000 products on the ASOS iOS app in February 2026. Each experience loads in just 4 to 7 seconds, well ahead of typical industry solutions, ASOS stated. The feature was available to select UK and US customers before being rolled out more broadly.

Zalando's size and fit work in Berlin prevented 8 percent of size-related returns in 2025, Inditex extended its AI fitting room from Zara to all of its brands in March 2026, and ASOS launched hybrid try-on with London-based Aiuta on about 10,000 products in February 2026. Inditex highlighted its AI-based virtual fitting experience, named ZARA Try-On, which since December has let customers create an avatar from their own photos and generate images of the avatar wearing real products.

Amazon Fashion launched its AI-powered virtual try-on for shoes in 2024, expanding to clothing in 2025; early results showed a 25% reduction in shoe returns and a 35% increase in customer satisfaction scores. Google Shopping integrated virtual try-on across its platform, allowing shoppers to see how clothes look on a diverse range of body types before purchasing; the feature processes over 10 million try-on sessions per month.

How does virtual try-on technology work?

Three distinct technical approaches underpin commercial virtual try-on systems: photo-based AI rendering, augmented reality overlay on live camera feeds, and avatar-based 3D simulation. Each has different accuracy profiles, cost structures and use cases.

Photo-based AI rendering, the most common approach in 2026, uses generative AI to composite a garment onto an uploaded photo of the shopper or onto a library of diverse model images. Developed in partnership with AI fashion platform AIUTA, the tool enables users to see how selected pieces could look on them in just four to seven seconds. The speed comes from pre-trained diffusion models that learned garment draping and body alignment from millions of training images.

Avatar-based systems take longer to set up but offer 360-degree visualisation. Users submit a selfie and a full-body photograph; AI processes these inputs and generates a personalised avatar in approximately two minutes, delivering both a high-resolution photo and a 360-degree video of that avatar wearing selected items. Zara's implementation uses this approach. The trade-off is a two-minute wait versus instant rendering.

Augmented reality try-on overlays 3D garment models onto a live camera feed, most common for accessories and footwear where the item is rigid and anchored to a fixed point. A convincing try-on places and scales the shoe on the detected foot automatically, at real-world dimensions, with no alignment step or calibration card asked of the shopper; if a shopper has to fiddle with placement, the answer to 'does this look right on me' is already compromised. AR works well for eyewear, watches and shoes but struggles with fabric drape and movement.

What are the accuracy challenges?

Virtual try-on shows what a garment looks like on a body shape, but struggles to predict whether it will fit. The distinction matters because most returns trace to fit, not aesthetics. Sizing is not standardized across fashion; this is where virtual try-on moves beyond visualization into the kind of product intelligence that brands need on top of pure try-on, noted Daria Shapovalova, co-founder and CEO of DRESSX.

Despite rapid progress, virtual try-on limitations still affect accuracy and reliability; understanding these gaps helps brands make informed decisions when adopting advanced fitting technologies. Material realism remains a major limitation due to ongoing fabric rendering challenges in AI; thick materials like knitwear challenge AI accuracy due to depth and layering complexities.

How a garment drapes, stretches, or gathers is driven by its material properties; the main limitation is fabric simulation; drape, stretch, and weight are difficult to replicate digitally with complete accuracy. Silk clings differently than cotton, denim stretches differently than leather, and current AI models approximate rather than simulate the physics. The visual result may look plausible but not predict actual fit.

A 2024 survey from Brunel University on deep-learning virtual try-on models points to persistent weaknesses around preserving clothing characteristics and textures, accurately applying clothing to the person, preserving facial identities and dealing with dataset bias. Lighting, pose, image quality and body diversity in training data all affect output accuracy. Lighting, pose, and image quality drastically affect the accuracy of the generated image.

DressX says its virtual try-on can give customers up to 89% confidence in fit, based on post-interaction surveys and behavioral data, benchmarked against users browsing the same product pages without try-on. That figure measures confidence, not accuracy. The gap between feeling confident and receiving the right size remains the central challenge.

What does virtual try-on cost to implement?

Virtual try-on software cost in 2026 falls into four bands: self-serve SaaS apps from about 20 to 300 US dollars or more per month with included render volumes; usage-based pricing from roughly 0.05 to 0.50 US dollars per render beyond plan limits; mid-market annual licenses commonly quoted between 15,000 and 75,000 US dollars per year; and custom API builds.

Accuracy is moderate, but implementation cost is low, typically 500 to 2,000 US dollars per month for SaaS solutions for basic photo overlay systems. Self-serve options start low: Virtusize's Shopify plan begins at 39 US dollars per month after a 1-dollar first month. Usage-based pricing ranges from 0.16 US dollars per try-on at the entry tier, 0.12 US dollars at mid-tier, to 0.09 US dollars per try-on at the highest volume.

Custom development for large retailers runs substantially higher. Custom development ranges from 50,000 to 500,000 US dollars or more for initial development, with ongoing maintenance at 15 to 20 per cent annually. Integration costs add 5,000 to 50,000 US dollars depending on platform complexity and existing technology stack.

Content creation adds another variable. 3D modeling per garment costs 50 to 500 US dollars; catalog conversion can be significant, though many SaaS providers work from existing 2D product photography. The hidden costs include testing, quality assurance, training and ongoing catalog maintenance as new products launch.

When does the ROI justify investment?

Return on investment depends on three levers: baseline return rate, average order value, and the volume of returns that trace specifically to size and fit. A retailer with a 25 per cent return rate spending 25 US dollars per return on 10,000 monthly orders spends 62,500 US dollars per month processing returns. A 30 per cent reduction in that rate saves 18,750 US dollars monthly, or 225,000 US dollars annually.

A mid-size fashion brand with 10,000 monthly orders at 85 US dollars average order value and a 28% return rate spending 18 US dollars per return pays 50,400 US dollars monthly in return costs; after implementing AI try-on with a conservative 36% relative reduction, the return rate drops to 18%, cutting monthly return cost to 32,400 US dollars, a savings of 18,000 US dollars monthly.

Against implementation costs of 500 to 2,000 US dollars per month for SaaS or 15,000 to 75,000 US dollars annually for a mid-market license, the payback period ranges from weeks to several months for mid-size retailers with meaningful return volumes. By implementing a virtual try-on system that costs 2,000 US dollars a month in API tokens and infrastructure, you only need to reduce that return rate by 5% to break even; if you reduce it by 15%, which is the industry standard for high-fidelity try-on implementations, you are adding tens of thousands of dollars directly to your bottom line.

Conversion lift adds a second revenue stream. In one multi-category marketplace pilot, customers who completed a try-on converted at twice the rate of standard shoppers, while in a fashion marketplace test users who performed a try-on added items to cart 52% more often and converted 35% more frequently. DRESSX studied over a million luxury fashion shoppers for its 2026 report and found that try-on users converted 50% higher, were about 3 times more likely to add to cart, and viewed 7 times as many product listings.

The ROI calculus breaks down for small-volume retailers. A brand processing 500 orders per month with a 20 per cent return rate and 20 US dollars per return spends 2,000 US dollars monthly on returns. A 30 per cent reduction saves 600 US dollars, below the cost of many SaaS plans. Finance rarely rejects try-on because the monthly fee is outrageous; the objection is whether the volume justifies the overhead.

What are the risks and limits?

Virtual try-on addresses one category of returns but does nothing for quality issues, colour mismatch, damage in transit or buyer's remorse. Size and fit is the single largest reason for online apparel returns, cited by 53% of respondents in Coresight Research's survey of US apparel retailers, well ahead of colour at 16% and damage at 10%. That leaves roughly half of returns untouched.

Adoption remains low despite proven ROI. The adoption gap is the headline: 71% of consumers say they'd shop more with AR try-on, yet only about 1% of e-commerce stores offer it. Adoption is still far from universal; surveys suggest that roughly half of U.S. online shoppers express interest in virtual try-on experiences, which means a large portion of consumers remain unconvinced.

Privacy concerns sit below the surface. Virtual try-on tools capture facial geometry, skin tone, and sometimes body measurements, which constitutes biometric data under laws like Illinois BIPA; users should review privacy policies before uploading photos. Retailers must handle, store and delete customer images according to local data protection regulations, adding compliance overhead.

The technology favours certain categories. Formal wear and dresses show a high baseline return rate of 40% or more, delivering large dollar savings per return avoided due to high price and high return rate. Accessories, eyewear and shoes see strong results because AR try-on handles rigid items well. Basic tops and casualwear show smaller gains because shoppers are more tolerant of fit variance at lower price points.

Zalando itself has been upfront that a single pilot needs a larger sample before that number becomes a stable industry benchmark, and that caution is worth keeping in mind rather than treating the figure as settled fact; broader industry claims about return rate reductions commonly land somewhere between 20% and 40%, but that range shifts significantly by retailer and product category. Vendor-reported figures should be discounted accordingly.

Frequently asked questions

What should fashion companies do now?

1. Audit your return data by reason code. Separate size and fit returns from quality, colour and other causes. Virtual try-on only addresses the first category, so confirm it represents at least 50 per cent of your return volume before evaluating tools.
2. Calculate your fully loaded return cost. Add reverse logistics, inspection, restocking, markdown loss and payment fees. Use 20 to 30 US dollars per return as a mid-band estimate if you lack internal data. Multiply by monthly return volume to establish your baseline cost.
3. Run a controlled pilot on high-return SKUs. Select 50 to 200 products with return rates above your category average. Implement virtual try-on on half, leave the other half as a control. Measure return rate, conversion and engagement over 60 to 90 days.
4. Start with SaaS before building custom. Self-serve plans from 20 to 300 US dollars per month let you test demand and measure lift without capital commitment. Reserve custom development for retailers processing at least 10,000 orders monthly where the ROI is unambiguous.
5. Combine visual try-on with size recommendations. Pilot data shows the best results come from pairing AI rendering with fit prediction algorithms that recommend which size to order. Visual try-on alone does not solve the sizing problem.
6. Disclose data handling clearly. Publish a privacy notice explaining how customer photos are processed, stored and deleted. Compliance with GDPR, CCPA and biometric data laws is not optional, and customer trust depends on transparency.
7. Measure engagement, not just installs. Track the percentage of product page visitors who activate try-on, the conversion rate of try-on users versus non-users, and the return rate by cohort. Low engagement means the feature is not visible or trusted; high engagement with flat conversion means the implementation is not accurate enough.