The Fit Gap: Using AI to Solve Fashion’s 30% Return Problem

We’ve all done the “bracketing” dance: ordering a Size 6, 8, and 10, knowing most of them will end up back in the mail. For shoppers, it’s a safety net. For retailers, it’s a margin killer.

Fashion e-commerce return rates hover around 30%. It’s easy to blame logistics, but the deeper issue is data. Shoppers are guessing because static size charts cannot convey fit, proportion, or drape. A Medium in a boxy linen shirt fits differently than a Medium in a bias-cut dress, and no spreadsheet or chart can capture that nuance.

This mismatch between customer expectation and product reality is where merchandising intelligence becomes essential. Front-end tools help shoppers make better decisions, but without understanding how products truly perform for different body types, returns will persist, and margins will suffer.

The Limits of the “Blind Buy”

Online shopping asks consumers to trust a 2D image and a generic size guide. For decades, that has been the standard, but it comes with obvious limitations. A size chart might tell a customer their usual size, but it does not capture how the garment drapes, how the fabric stretches, or whether a hemline hits in the right place for their frame.

This limitation doesn’t just affect shoppers; it affects merchandising teams. When returns pile up, teams often focus on volume metrics: how many items were returned, or which SKUs had high return rates. But those numbers tell only part of the story. They show what was returned, not why.

Brands have started to experiment with technology that digitizes the fitting room. Virtual try-on tools allow shoppers to visualize garments on models with different body types. Precision sizing apps can take measurements from phone cameras, enabling a more personalized experience for custom-fit categories like intimates.

These tools help customers understand how a garment fits and educate them to evaluate silhouette and proportion, not just a number on a tag. But they cannot fix the underlying assortment if the products themselves do not align with customer needs.

Stocking Shapes That Fit

Front-end tools improve confidence, but they cannot compensate for a misaligned product mix. If midi skirts are returned repeatedly because the hips are too narrow, a virtual try-on experience will show the customer the mismatch, but it will not solve the underlying issue: the assortment itself is misaligned.

Traditional merchandising systems often fall short here. Spreadsheets can report high return rates for a particular SKU, but they cannot explain which visual attributes are causing returns. Teams are left guessing, making incremental adjustments based on incomplete data rather than systemic insight.

This is where a visual intelligence platform like Prévoir becomes critical. By capturing and analyzing the attributes that truly drive fit like waist placement, shoulder structure, sleeve volume, hemline, and cut, teams can understand why returns happen and which shapes are most successful for their specific audience.

Turning Fit Into Strategy

Prévoir approaches fit not as a size problem, but as a product attribute problem. Instead of relying solely on text descriptions, it examines product imagery to track attributes and silhouettes at scale.

• Visual attribute tagging: Product imagery is analyzed to tag attributes such as “Empire Waist,” “Drop Shoulder,” or “Bias Cut.” This removes subjective inconsistencies from manual tagging and creates a standardized database of product characteristics.

• Matching fit to sales: By comparing these visual attributes to sales and returns, merchandising teams can detect patterns. Perhaps “Oversized Blazers” are trending broadly, but your audience returns the ones without structured shoulders. Knowing this allows you to adjust your assortment intelligently.

• Smarter sizing curves: Historical sales data, combined with attribute-level insights, helps teams adjust size distribution. Instead of stocking deeply in shapes that consistently underperform, teams can allocate more to silhouettes that drive full-price sell-through.

  
  

With this approach, fit ceases to be anecdotal and becomes measurable. Merchandising decisions move from reactive to proactive, informed by a deeper understanding of both customer behavior and product characteristics.

Why Returns Are Not Just a Logistics Issue

Reducing returns is not simply about improving fulfillment efficiency. Returns are a signal. They tell a story about how well a product matches customer expectations.

When teams misinterpret this signal, they make decisions based on incomplete information. They may overproduce SKUs that seem popular in the abstract, or underproduce shapes that sell consistently but are miscategorized. By connecting visual product attributes to sales and returns, teams can close this feedback loop.

Front-end fit tools help the customer make better choices in the moment. Back-end insights help the business make better buying decisions across the season. Together, they reduce returns not by limiting experimentation, but by aligning products with actual demand.

The Takeaway

Fit is both a customer and a merchandising problem. Customers need the confidence to select items that suit them, and teams need the visibility to understand why some products fail while others succeed.

Front-end tools like virtual try-ons and precision sizing improve choice. Visual intelligence platforms like Prévoir ensure the assortment works. When product decisions are informed by both what sold and why it fit (or didn’t) teams can stop chasing returns and start delivering products that customers keep.

In the end, better fit isn’t just about fewer returns. It’s about clearer insight, smarter assortments, and stronger margins.

Start Your Free 14-day Trial of Prévoir