Size & fit prediction:
The right size first time
How a collaborative-filtering model uses purchase history and return patterns to reduce the fit-related return rate for jeans from 38% to 31% - with a checkout widget.
This module is part of a six-part model calculation about a fictional mid-sized fashion company (1,800 SKUs, 18 markets). All figures, datasets, model metrics and euro amounts are entirely fictional - they illustrate the methodology and the kind of effect achievable. They are not results of real clients and not a performance promise. What values are achievable on your real data is established only in a pilot project.
01 The problem - "does not fit" is the #1 return reason
Why 34% of all jeans returns are due to incorrect size selection
Jeans are the most return-intensive garment: the return rate stands at 38%, with 34% attributed to "does not fit". The problem is not quality - it is uncertainty at the point of purchase. A "32/32" from Brand A fits completely differently than from Brand B. And within the same label, slim fit, straight and wide leg can vary by up to 2 dress sizes.
02 Model - collaborative filtering + body-shape matching
The model calculation assumes: for 68% of fit-related returns, a correct size recommendation would have been sufficient. This is not AI magic - it is the systematic evaluation of data that already exists: which customer kept which size in which cut, and which they returned. The saving of €1.61M (61,824 avoided returns × €26) requires no hardware investment - only an algorithm in the checkout. Published vendor cases reach up to -20% on the overall return rate (Fit Analytics/Mammut) - our -18% relative reduction sits below that; the 38% baseline is an assumption (published: German market overall ~50%). For new customers without purchase history, the model falls back on category averages - full recommendation quality is reached from the second purchase onwards.
03 Next steps for your brand
Export your returns data with size information. Calibration of the size run per SKU.
Size recommender as a widget on the product page: "Based on your previous purchases, we recommend size 31/32".
Automated detection of miscalibrated size tables. Report to product team: "SKU X runs 1.2 sizes too large."