Consumer Tech

Your Return Rate Just Became a Software Problem

Online shoppers return nearly one in five orders. Zalando and Zara are now running AI virtual fitting rooms in production to fix it. Smaller retailers can too.

Consumer TechBy Servola Tech Desk2026-08-234 min read

AI-assisted, edited by humans. | Editor-in-Chief: Leon Soliman | Editorial standards

Your Return Rate Just Became a Software Problem

What are Zalando and Zara actually doing differently?

Both companies now run AI-driven virtual fitting tools in live production, but they took different technical paths to get there. Zalando's system leans on precision: it combines a shopper's own body measurements with detailed 3D scans of each garment to predict whether a specific size will actually fit, an approach reported by Bloomberg as more accurate but, by its own account, sometimes uncomfortable for shoppers to interact with, since research shows a body avatar can affect a person's confidence and buying behavior.

Zara took the opposite route. Its tool uses generative AI to render an outfit onto an avatar resembling the shopper, prioritizing an engaging, inspirational image over exact measurement. Zara rolled the feature out in December alongside new return fees for online orders, and each generated outfit currently takes about two minutes to load with a daily cap of five looks per shopper, a real usability tradeoff the company is accepting in exchange for the fit signal it produces.

Why does a fit problem cost this much money?

Because returns are no longer a rounding error in retail; they are one of the largest line items in the industry's cost structure. Retailers took back 15.8 percent of everything sold in 2025, a total of 849.9 billion dollars, and the online-only figure was worse at 19.3 percent, according to industry return-rate data reported this year. Gen Z shoppers, the demographic every fashion retailer is chasing hardest, returned nearly eight online orders each on average last year, and 82 percent of consumers now say free returns are essential to whether they buy at all, which means retailers cannot simply charge their way out of the problem without losing the sale in the first place.

Sizing mismatch, not damaged goods or changed minds, is the single largest driver of that return volume, which is exactly why a software fix aimed at predicting fit before checkout, rather than a logistics fix aimed at handling the box after it comes back, has become the direction big apparel retailers are actually investing in.

How much could this actually save, in real numbers?

Early results from retailers willing to disclose them suggest a genuine, measurable effect rather than a marketing claim. ASOS reported a 160-basis-point reduction in its own return rate after testing a comparable AI fit-prediction tool, a small-sounding number that translates into real reverse-logistics volume at ASOS's scale, and McKinsey's estimate of the global upside puts a hard ceiling on how large this opportunity actually is.

Metric	Figure
Overall retail return rate, 2025	15.8 percent ($849.9 billion)
Online-only return rate, 2025	19.3 percent
ASOS return-rate reduction after AI fit tool	160 basis points
McKinsey estimate, global returns eliminable via virtual try-on	$100 billion to $150 billion a year

Is this only for retailers Zalando's size?

No, and this is the part that matters for a smaller European online retailer reading about Zalando and Zara's rollout. Virtual try-on has moved from a bespoke engineering project to a vendor category: platforms including Shopify's Genlook, AIUTA, and Catches now sell fit-prediction and try-on tools that plug directly into an existing storefront rather than requiring an in-house 3D-scanning pipeline. Catches, one such vendor, projects a 10 percent increase in conversions alongside a 20-to-30 times return on the cost of implementation for merchants that adopt it.

That gap between what Zalando built in-house and what a mid-sized retailer can now buy off the shelf is closing quickly, which changes the calculation from should we build this to which vendor should we test first.

What should an EU online retailer actually do with this?

Start by finding the real number before buying a tool to fix it: pull your own return rate broken down by size category and product line, since the categories driving your highest reverse-logistics cost, typically fitted clothing rather than accessories or footwear with more forgiving sizing, are where a try-on pilot will show a measurable effect fastest. Run the pilot on that narrow slice of the catalog rather than the whole storefront, so the conversion and return-rate delta is attributable to the tool rather than lost in overall seasonal noise.

Measure the win on logistics cost per order, not just on-site engagement or time spent on the product page; a shopper who spends two extra minutes with a virtual fitting tool but still returns the item has cost you the same reverse-shipping fee as before, with an added hosting bill on top.

Frequently asked questions

What is virtual try-on technology?

It is software that shows a shopper how a specific garment and size will look or fit on their own body, either through 3D body-and-garment scanning, as Zalando uses, or through a generative-AI image of the outfit on an avatar resembling the shopper, as Zara uses, before the shopper commits to an order.

Does virtual try-on actually reduce returns, or is it a marketing claim?

ASOS reported a measured 160-basis-point reduction in its own return rate after testing a comparable AI fit-prediction tool, and McKinsey separately estimates that broad adoption of virtual try-on could eliminate 100 to 150 billion dollars of returns globally each year, both figures from disclosed results rather than vendor marketing.

Is this only relevant to large fashion retailers like Zalando and Zara?

No; vendors including Shopify's Genlook, AIUTA, and Catches now offer virtual try-on and fit-prediction as a plug-in service for existing online storefronts, which means a smaller retailer can pilot the same underlying technology on its highest-return product categories without building any of the underlying 3D or generative-AI infrastructure itself.

Zalando and Zara are not experimenting anymore; they are running this in production because the return math already worked. The retailers still treating fit as a customer-service problem instead of a software one are the ones paying for both.

ZalandoZaraVirtual Try-OnE-commerceRetail AI

Fairphone 6 Plus Debuts Into a Memory Cost Squeeze

Fairphone launched the Fairphone 6 Plus on August 18 at 650 to 690 euro depending on VAT, a mid-gen refresh landing as memory chip costs rise industry-wide.

3 min read

Consumer Tech

ChatGPT Can Now Send Texts From Your Work Mac

OpenAI's new ChatGPT plugin reads, drafts and sends iMessage, SMS and RCS texts on Apple Silicon Macs. Here is what IT and security teams should decide before employees turn it on.

5 min read

Decisions

The Third Pivot Was the Warning Magic Leap Missed

Magic Leap filed a WARN notice for 193 layoffs effective 1 October 2026, its fourth strategic pivot after $4 billion raised. The lesson is the pattern, not the pivot.

4 min read

← All articles