Size recommendation API
Size recommendations with the odds, not a single guess.
WooChim scores a shopper's measurements against your size chart and returns a probability per fit — tight, true, or loose — plus a confidence score and the assumptions behind it. Honest sizing that reduces wrong-size returns.
Calibrated, not binary
A distribution over tight/true/loose instead of a confident 'M fits' that's wrong on a third of orders. Render the odds, set expectations honestly.
Models uncertainty
Unknown fabric, sparse measurements, or a chart with tolerance all widen the estimate and lower confidence — surfaced, never hidden.
Returns-focused
Wrong-size is the #1 apparel return reason. A trustworthy size call at the point of decision is the cleanest way to bring it down.
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.
See it on your body in a minute
Free to start, no card required. Set up your fit model and check any garment.