Accuracy

Where we can be wrong.

A size suggestion is a starting point built from other people's reports and default conversion tables. Here is exactly where it breaks, so you can tell when to trust it less.

Conversions between label systems

• Bras. US cups E and F mean different things at different brands (some run D, DD, DDD, G; others D, DD, E, F). We flag these as low confidence. EU, French and Australian bands and cups are converted with the Freya international chart, which not every brand follows.
• Shoes. Men's to women's is usually +1.5 sizes but not always; EU sizes do not map cleanly to US half sizes; width codes are brand-specific. Anything not in US women's is medium or low confidence.
• Leggings and jeans. Numeric sizes (0, 2, 4…) and alpha sizes (S, M, L) cover different waist ranges at every brand: a label 27 is a 28.25 inch waist at one brand and 28 at another, and an XL spans 32.5 to 39 inches across brands. We hold brand charts for the main brands; where a chart has not been checked against the brand's current page it is marked unverified, and where we hold none we fall back to a generic table and say so.
• Tights. Sizes depend on each brand's height-and-weight grid. Until a grid is verified, tights suggestions stay low confidence.

The size converter

• The formula column is arithmetic. It says how the systems are defined, not how a garment is cut. Women's clothing, menswear and rings are formula only: we have not checked brand charts for them.
• Brand to brand is two brands' claims. Each chart says which body a size is built for, and brands do not measure that body the same way. When the body-based answer and the plain label sit a size apart the answer drops to medium confidence, and further apart to low, with both ends shown as a range.
• Wearers need five people. What people who own both brands actually wear is shown only once five different people stand behind it. Until then the panel says there is not enough, and nothing is estimated in its place.

Try it at the size converter.

Crowd signals

• Fewer than five distinct reporters: nothing is shown. Five to fifteen effective reports: labelled “early signal”. A single loud group can still skew an early signal.
• Reports carry a self-reported size and coarse shape answers. People misremember sizes and describe their shape differently; similarity is a soft weight, not a filter.
• Products change between production runs. Reports on an older version are down-weighted, but we cannot detect a silent change in a factory.
• The public site never shows generated or sample fit reports. A product with no reports says so and shows only the maker's facts and size chart.

Your anchor

• “Fits well” means different things to different people. Two anchors in the same category beat one; three beat two.
• An anchor from a different category is not used. A shoe tells us nothing about jeans.
• When two of your anchors disagree by more than a size and a half, we say so and lower the confidence rather than average them silently. Older anchors count for less: bodies and products both change.

What to do when it is wrong

Return the item under the retailer's policy, then file a fit report saying what happened. That report is exactly what stops the next person making the same mistake, and it corrects the suggestion for people shaped like you.