Key takeaways
- Roughly 30% of online fashion comes back, and the number has barely moved in a decade, fit and size are the leading cause.
- Sizing fails because labels are not measurements: a 38 at one brand is a 40 at another, so shoppers hedge by bracketing.
- AI sizing replaces the guess with a match: the shopper's real body profile against the specific garment's real measurements.
- The results are measurable: Showroomprivé cut returns 11%, ETAM cut size-related returns 16% in an A/B test, and deployments typically see returns fall 20 to 50%.
- Confidence also sells: assisted sessions tend to convert 15 to 25% higher.
The 30% problem nobody priced in
Roughly 30% of online fashion comes back. That is the working number across most European apparel retailers, and it has barely moved in a decade. A third of everything you ship is a round trip: picked, packed, delivered, collected, inspected, and, if you are lucky, restocked in time to sell again at full price.
Fit and size sit at the top of the reason-for-return list, ahead of quality, colour, and changed minds. Shoppers do not return because they hate the dress. They return because the dress does not fit. The garment was fine. The guess was wrong. And the economics of that wrong guess land entirely on the retailer, which is why the industry-wide bill runs into the tens of billions.
Why size is the hardest variable in e-commerce
A size 38 at one brand is a 40 at another and a 36 somewhere else. Vanity sizing shifts the labels. Different pattern blocks shift the cut. Different fabrics stretch or hold. Even within one brand, the denim line and the knitwear line are often built on different assumptions about the body wearing them.
The shopper has no way to know any of this, so they do the rational thing: they bracket. Order two sizes, keep one, send one back. Every bracketed order is a return baked in from the start, the customer did nothing wrong. The store simply never told them which size would actually work. We unpack that behaviour fully in The Bracketing Habit That Is Killing Your Margins.
Why the size chart cannot save you
The classic answer to fit uncertainty is a size chart, and it fails for a structural reason: it maps a label to a range of body measurements, then asks the shopper to measure themselves correctly, interpret a grid, and resolve a borderline result on their own. Most will not, and those who try often measure wrong.
Worse, a chart knows nothing about the specific garment. It cannot say that this blazer runs narrow in the shoulder or that this jersey dress forgives a size either way. Fit lives at the intersection of one body and one garment, and a static chart sees neither.
How AI sizing closes the gap
Kleep Smart Sizing predicts the right size for each shopper against each garment's real measurements. A 30-second questionnaire or a two-photo body scan builds a body profile, then matches it to the actual spec of the item in front of them, the flat measurements, the ease, how the fabric behaves.
The recommendation is garment-specific, not brand-wide. It accounts for the cut of that exact product, which is the part static size charts can never do. And because the answer is one confident size rather than a range, the shopper orders one unit instead of two. That single behavioural change is where a large share of the return reduction comes from.
The numbers in practice
This is not a theoretical case. Showroomprivé cut returns by 11% with Kleep AI sizing at marketplace scale. ETAM ran a rigorous A/B test and saw revenue rise 5% while size-related returns fell 16%. Across deployments, brands typically see returns fall 20 to 50% on assisted purchases, and assisted sessions tend to convert 15 to 25% higher, because confident shoppers buy.
The volume behind those numbers has scaled fast: from 17,000 sizing flows in 2023 to 500,000 in 2024, 4.2 million in 2025, and in 2026 a Kleep sizing flow runs every 5 seconds across 300+ live brands, from Givenchy and Lacoste to A.P.C. and Represent. The pattern holds from luxury to mass market: tell people their size, and the returns that fit anxiety creates stop happening.
The garment was fine. The guess was wrong. The 30% return rate is not a law of physics, it is a fit problem in disguise.
Why is the return rate so high in online fashion?
Because shoppers cannot try garments on, and size labels are inconsistent across brands and even across categories within a brand. Fit and size are the leading cause of returns, ahead of quality or changed minds, which pushes the typical rate to around 30%.
What is the main reason clothes bought online get returned?
Wrong size or poor fit. Shoppers guess from unreliable size charts, or deliberately order multiple sizes intending to return the extras, a behaviour called bracketing that builds returns into the order before it ships.
How can fashion retailers reduce returns caused by sizing?
By replacing the size chart with a per-garment recommendation built from the shopper's actual body profile. AI sizing tools like Kleep Smart Sizing do this with a 30-second questionnaire or a two-photo scan; results in production include an 11% return reduction at Showroomprivé and a 16% drop in size-related returns at ETAM.
Do size recommendation tools actually increase conversion?
Yes, removing fit uncertainty removes a major reason to hesitate. Assisted sessions typically convert 15 to 25% higher, and ETAM's A/B test measured a 5% revenue lift alongside the returns reduction.
Conclusion
The 30% return rate is not a law of physics. It is a fit problem in disguise, sustained by labels that mean nothing and charts that cannot see the garment. Tell shoppers the right size for their body against the real product, and a large share of those returns never happen in the first place, which is exactly what the Showroomprivé and ETAM numbers show.
