How does AI improve fit and sizing? From size advice to better patterns

KEY TAKEAWAYS Summary by the editors

1. AI size and fit tools combine brand size charts, purchase and return history, customer feedback and sometimes body measurements to recommend a size or flag that an item runs small or large.
2. Zalando reports that its size and fit solutions prevented 8% of size-related returns in 2025 and cover around 70% of its assortment, and that return rates in categories such as jeans can reach 65%.
3. Zalando also reports that recent virtual fitting room pilots reduced return rates by up to 40%, a pilot result rather than a company-wide figure.
4. The larger long-term value lies upstream: aggregated fit feedback and body data can show brands which patterns, grades and size charts cause problems.
5. Body measurement tools raise privacy questions, and Zalando states that photos for its measurement feature stay on the customer's device and are deleted.

AI improves fit and sizing in two ways. At the point of sale, it recommends the right size to each shopper using size charts, past purchases, returns and sometimes body measurements, which reduces size-related returns. Behind the scenes, the same data can show brands which styles, patterns and grading rules cause fit problems, so they can correct them in product development. Reported gains are real but measured in single digits overall, with larger effects in specific categories and pilots.

Why does fit matter so much in fashion?

Sizing is not standardised across brands, and often not even across one brand's categories. A shopper who is a medium in one label's knitwear may be a large in its outerwear. Online, the result is bracketing (ordering two sizes and returning one) and returns that cost money in logistics, handling and markdowns. The online platform Zalando says return rates in categories such as jeans can reach 65%. Fit is therefore both a customer experience problem and a margin problem.

How do AI size recommendations work?

Most systems start from a simple principle: learn from what similar customers bought and kept. If many shoppers who keep a medium in brand A return a medium in a specific style from brand B and keep the large, the model learns that the style runs small. More advanced tools layer additional data on top:

• Size flags on product pages that say an item runs small or large.
• Size profiles where customers record brands and sizes that fit them.
• Body measurement tools that estimate measurements from phone photos or video.
• Virtual fitting rooms that show a garment on a 3D avatar in different sizes.

Zalando describes this layered approach in a June 2026 company article, combining brand sizing information, purchase and return patterns, customer fit feedback, insights from specialised fitting models and what it calls the largest anonymised dataset of body measurements from European fashion customers. It says an in-house team of 80 specialists works on these tools.

Read also

Virtual try-on in fashion: how does it work and does it reduce returns?

What results have retailers reported?

Zalando states that its size and fit solutions prevented 8% of size-related returns overall in 2025 and that its foundational solutions cover around 70% of the assortment. It reports that more than 1.5 million customers have tried its body measurement experience, and that recent virtual fitting room pilots reduced return rates by up to 40%. Earlier, in 2023, the company reported a 10% reduction in size-related returns for items with size advice compared with items without it.

Two cautions apply when reading such figures. They are company-reported, and pilot results in selected categories are not the same as an average across a full assortment. Brands should expect their own results to depend heavily on data volume and category mix.

Can fit data improve patterns and size charts?

This is the less visible but potentially larger opportunity. Every return reason such as too tight at the hips or sleeves too long, combined with the customer's usual sizes and, where consented, body measurements, is a data point about the garment itself. Aggregated by style and size, these signals can show that a particular block is graded too narrowly in larger sizes, or that a size chart does not reflect the brand's actual customers. When Zalando launched body measurement-based size advice in 2023, it said the feature could also give brand partners further insight into their customers' size and fit requirements.

Some brands bring this into development through digital tools. HUGO BOSS says it uses virtual try-ons with avatars and 3D simulation in product development to simplify processes without compromising on performance and fit. Research is also exploring how AI could generate sewing patterns directly from design inputs, as in the Design2GarmentCode work presented at CVPR 2025, though that remains at research stage.

What data does AI fit and sizing need?

The table also explains a structural challenge: the data that reveals fit problems often sits with retailers, while the ability to fix patterns sits with brands. Sharing structured, anonymised fit feedback between them is what turns return reduction into better products.

Practical sharing does not require exotic technology. A seasonal fit report per style, listing the share of returns marked too small or too large by size, together with the most frequent free-text comments, already gives a technical designer something to act on. Over several seasons, the same data shows whether corrections to a block actually reduced fit complaints, which is the evidence most product teams lack today.

What are the limits and privacy risks?

Recommendations need volume: new brands, new styles and low-traffic categories have little history to learn from. Fit preferences are personal, so a technically correct size can still be returned because the customer wanted a looser look. Body measurement tools handle sensitive personal data, and customers need clear consent and control. Zalando states that the photos used by its measurement feature never leave the customer's device and are automatically deleted, an approach that reduces exposure.

Read also

AI, 3D and virtual sampling: fewer samples, faster decisions

Where should a brand start?

1. Capture structured return reasons, including fit direction (too small, too large, too long).
2. Publish accurate garment measurements, not only body size charts.
3. Analyse fit issues by style and size to find blocks or grades that need correcting.
4. Share fit insights with product development each season.
5. Introduce recommendation tools once there is enough history to learn from.

AI size advice reduces the symptom at the point of sale. Using the same data to correct patterns, grading and size charts reduces the cause, and that is where brands with control over their product development have the most to gain.

Frequently asked questions

How do AI size recommendation tools work?

They learn from size charts, purchases, returns and fit feedback, and sometimes from body measurements, to predict which size a shopper is most likely to keep. Some tools also flag when a specific item runs small or large.

Do AI size tools really reduce returns?

Retailers report measurable but moderate effects. Zalando says its size and fit solutions prevented 8% of size-related returns in 2025, with larger reductions of up to 40% in virtual fitting room pilots. Results vary by category and data volume.

Can AI help brands improve their patterns?

Yes, indirectly. Aggregated fit feedback and return reasons by style and size can reveal patterns, grades or size charts that cause problems, which product teams can then correct. Generating patterns directly with AI is still mainly a research topic.

Is body scanning for size advice safe for privacy?

It depends on the implementation. Customers should give explicit consent and know how images and measurements are stored. Zalando, for example, says photos for its measurement feature stay on the device and are deleted automatically.

GuideThe complete guide to AI in fashion design and product developmentRead the complete guide