Size Recommendation System for Fashion E-commerce

Understanding user size preference in addition to style preference is a critical aspect of fashion e-commerce domain. Unlike offline, in online fashion shopping, customers don’t have the luxury of trying a product and have to rely on the product image and the size charts to select a product that fits well. As a result of this gap, online shopping yields a large percentage of returns due to size and fit. Also, explicit elicitation of a users body shape or measurements does not scale well. In this paper, we propose a size recommendation system to automatically pre-select consumer’s size based on past purchase and content data without explicitly asking for users measurements. We use skip gram based word2vec model on our purchase data to learn the latent representation of all our products and users in a common size and fit space, thereby enabling a similarity notion among different products and user-products. Gradient boosting classification model is further employed on both the learnt latent features and observable features (like users estimated chest size, products fit etc.) to predict the preferred product size for a user. The effectiveness of the proposed algorithm is validated through extensive experiments on real world data. Further we derive distinct users’ body shapes and glean insights from their return behavior on our platform.