Shop your Right Size: A System for Recommending Sizes for Fashion products

Size selection is a critical step while purchasing fashion products. Unlike offline, in online fashion shopping, customers don’t have the luxury of trying a product and have to rely on the product images and 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. Hence providing size recommendation for customers enhances their buying experience and also reduces operational costs incurred during exchanges and returns. In this paper, we present a robust personalized size recommendation system which predicts the most appropriate size for users based on their order history and product data. We embed both users and products in a size and fit space using skip-gram based Word2Vec model and employ GBM classifier to predict the fit likelihood. We describe the architecture of the system and challenges we encountered while developing it. Further we also analyze the performance of our system through extensive offline and online testing, compare our technique with another state-of-art technique and share our findings.