Key takeaways
- A size chart is a statistical generalisation pretending to be personal advice, it assumes proportional bodies that mostly do not exist.
- Millions of fit sessions show the same patterns: shoppers sit between sizes, need different sizes per category, and mis-measure themselves.
- The chart has one input (a label) and zero knowledge of the individual or the specific garment, so it hedges with ranges that force a coin flip.
- Replacing the chart with a body-to-garment match works: ETAM's A/B test cut size-related returns 16% and lifted revenue 5%.
- The fix is not a better grid, it is removing the grid from the shopper's job entirely.
A size chart is a generalisation
A size chart maps a label to a range of measurements. It assumes bodies are proportional: that someone with a given bust has a predictable waist and hip, that height and inseam move in lockstep, that one grading curve fits a population. Real bodies do not read the memo.
The chart is not lying maliciously. It is lying structurally, it presents an average as if it were advice for an individual. For the minority of shoppers who happen to sit near the centre of every measurement distribution at once, it works. For everyone else, it is a suggestion wearing the costume of an answer.
What millions of fit sessions revealed
Kleep's sizing volume has grown from 17,000 flows in 2023 to 4.2 million in 2025, which means we can now see, at scale, where the chart and reality diverge. The gap is not occasional. It is systematic.
- Shoppers routinely sit between two sizes, and the chart forces a coin flip.
- The same person needs different sizes across categories: one in knitwear, another in denim.
- Self-reported measurements are frequently off, because most people measure themselves wrong.
- Brand-to-brand drift means a shopper's usual size is a poor guide on a new site.
Why the chart cannot fix itself
A static chart has one input, the label, and no knowledge of the individual. It cannot account for body shape, fabric behaviour, or the specific cut of one garment versus another. A stretch jersey midi and a structured cotton shirt in 'size M' are two entirely different fit propositions, and the chart renders them identically.
So it hedges. It gives a range wide enough to be technically correct and practically useless, '38 to 40' is the chart admitting it does not know. The shopper is left to guess, and guessing is what drives bracketing and the 30% return rate. The distrust this breeds is well earned, and it compounds: as we explore in The Psychology of Size, shoppers who have been burned by charts stop consulting them at all.
Replacing the guess with a match
Smart Sizing flips the model. Instead of asking the shopper to interpret a chart, it builds a body profile from a short questionnaire or a two-photo scan, then matches that profile to the garment's real measurements, the actual spec of this product, in this fabric, with this cut.
The output is one confident recommendation for that exact item, not a range. The shopper stops guessing, orders a single size, and the store stops paying for the wrong guess. Where the chart says 'you are probably somewhere around here', the match says 'for this dress, take the 38'.
The proof from the A/B tests
When ETAM replaced chart-guessing with garment-level recommendations and A/B tested it properly, revenue rose 5% and size-related returns fell 16%. When ba&sh tested Kleep head-to-head against True Fit, add-to-cart rose 88% and usage of the tool rose 20%, shoppers engage more with an answer than with a grid.
These are not outliers; they are what happens when the guess is removed. Across deployments, returns typically fall 20 to 50% on assisted purchases. The chart never had numbers like these, because the chart never actually answered the question.
A size chart presents an average as if it were advice for an individual. It is a suggestion wearing the costume of an answer.
Why are size charts so inaccurate?
Because they map one label to a range of measurements and assume proportional bodies, while real bodies vary independently in bust, waist, hip and height. They also know nothing about the specific garment's cut or fabric, so the same chart serves products that fit completely differently.
Why do I wear different sizes at different brands?
Brands cut on different pattern blocks, grade differently, and apply vanity sizing to different degrees. There is no enforced standard behind the numbers, so a 38 at one brand can genuinely be a 40 or 36 elsewhere.
What is the alternative to a size chart for online stores?
Per-garment size recommendation: building a body profile from a quick questionnaire or two-photo scan and matching it against each item's real measurements. In ETAM's A/B test this approach cut size-related returns 16% and lifted revenue 5%.
Do shoppers actually use size recommendation tools?
Yes, when the tool gives a specific answer rather than another grid. In ba&sh's head-to-head A/B test, Kleep saw 20% higher usage than True Fit and an 88% higher add-to-cart rate.
Conclusion
Size charts are not malicious, but they are misleading by design: one label standing in for millions of different bodies and thousands of different garments. The data from millions of fit sessions is unambiguous. Match each shopper to each garment, and the lie the chart tells stops costing you returns.
