Size recommendations with the odds, not a single guess.

Frequently asked

What does the size recommendation API return?

A recommended size, plus tight/true/loose probabilities that sum to 1, a prediction-confidence score (0–100), and a list of assumptions (e.g. fabric unknown) — per garment, against the shopper's measurements.

How does it decide the size?

It compares the shopper's body measurements to the garment's size-chart numbers per dimension, models the uncertainty in each, and integrates over fit bands to produce probabilities — then recommends the size with the strongest fit.

Why probabilities instead of one size?

Because clothing fit is genuinely uncertain when fabric and cut aren't fully known. Returning honest odds builds trust and avoids the false precision that erodes it after a bad order.

What inputs do you need?

Body chest/waist/hips (and optionally more), plus the garment's flat size-chart measurements and category. Naming the fabric tightens the estimate.