What retailers have measured about fit, try-on and returns
Zalando, Levi's, Shopify, Google and others have published their own numbers. We put them in one place, with what each one measured and where it comes from. None of them are AI Frame clients. Below them, our first case study from a store that uses AI Frame.
Returns are where fashion margins go
In the US, retailers expected about 19.3% of online sales to come back in 2025, worth $849.9 billion across all retail. In fashion, a large share of those returns have nothing wrong with them. The size was wrong, or the item looked different on the shopper than they expected.
Size help: what Zalando measured
Zalando, which serves 62 million active customers across 29 markets, has teams in Berlin and Zurich working only on size and fit. It compared items where it shows size advice with items where it doesn't.
About 380 size recommendations matched to purchases in one month
A US unisex streetwear brand on Shopify, with a focused catalog of around 70 to 80 products. These numbers come from AI Frame analytics on their core hoodie, over about four weeks.
Their hoodies, bombers and hats are unisex and cut roomy. Shoppers couldn't tell their size from a men's or women's chart, so the brand wanted every shopper to get a clear answer on the product page.
They published their Size Chart with a height range per size (for example 5'0" to 5'3" is XS) and turned on Sizing assist on the product page. Shoppers get a size from their height alone, with no tape measure.
The hoodie is so roomy (a garment chest of about 117 cm, even at XS) that matching on chest alone would put almost everyone in XS. Height ranges match how the brand actually sizes it, so we added support for that chart type.
Put size help where shoppers already are. Sizing assist sits on every product page and takes a few seconds, so shoppers use it week after week. About 90 sized purchases a week on one hoodie shows how much shoppers want a clear answer.
What buyers ordered, compared with our recommendation
Half ordered exactly the size we recommended, and the most common choice after that was one size up, a common pick for a hoodie worn loose.
What the brand learned about its shoppers
| Buyers who | Profiles | Avg height | Avg chest |
|---|---|---|---|
| Sized up | 35 | 162 cm | 99 cm |
| Kept our size | 38 | 174 cm | 101 cm |
| Sized down | 9 | 173 cm | 91 cm |
Saved Fit Profiles showed which shoppers like to size up: shorter shoppers with a broader build. That's useful for the brand, for example to add a fit note to the product page, and it helps us make recommendations sharper for shoppers like them.
About this data: one store, one hoodie range, about four weeks (September to early October 2026), from AI Frame analytics. The brand stays anonymous. Next, we're adding return data to see which sizes shoppers kept.
Virtual Try-On: what has been measured
"How will it look on me?" is the question a Size Chart can't answer. These are the published results from tools that show shoppers the product before they buy.
What shoppers say
Why both on one page
Shoppers ask two questions before they buy. "Will it fit me?" is answered by a Size Chart. "How will it look on me?" is answered by Virtual Try-On. Zalando's numbers come from size help. The try-on numbers come from showing the product before purchase. AI Frame puts both on the same product page, so the shopper doesn't have to leave it to answer either one.
How to read this data
Most of these numbers come from companies that build or sell this technology, including us. Comparisons of shoppers who used a tool with those who didn't also flatter the tool, because interested shoppers are more likely to use it. Results vary by store, category and where the tools sit on the page. Read these numbers as a strong direction, not a promise for your store.
Measure it on your own store
The only numbers that matter for your store are your own. Here is a simple way to check:
- 1Note your product page conversion rate and return rate for the last 30 days.
- 2Add the Size Chart and Virtual Try-On to your product pages.
- 3After 30 days, compare. In AI Frame you can also see how many visitors used the tools.
Sources
- Zalando, How Zalando leverages technology to help customers find the right size
- Zalando, Size recommendations based on body measurements (press release, July 2023)
- Zalando, Returns at Zalando
- Zalando, Rewriting the rules of fit in Europe: 3 key takeaways from CPHFW AW26 (Feb 2026)
- Forbes, "Google, DressX And The New Fashion AI Virtual Try-On Stack", Moin Roberts-Islam (14 Apr 2026)
- DressX, 25 AI Virtual Try-On statistics (2026 report)
- Shopify Plus case study, Rebecca Minkoff
- NRF and Happy Returns, 2025 Retail Returns Landscape (Oct 2025)
- Google, Virtually try on clothes with a new AI shopping feature (2023)
- Bitkom, Trend im Online-Shopping: die digitale Umkleide (Jan 2023)
- AI Frame analytics, anonymous US streetwear merchant, September to October 2026 (case study above)
