Fit recommendation via collaborative inference

Abstract

Some embodiments of the invention provide techniques for recommending a size of a subject item to fit a subject consumer. In some embodiments, clusters of consumers with fit characteristics similar to the subject consumer are identified, using one or more data clustering algorithms, based on any of numerous consumer attributes (e.g., self-reported and/or inferred height, weight, body shape, body characteristics, and/or purchase histories (e.g., consumers with high overlap in terms of sets of products purchased)). Information on other consumers in the cluster may be analyzed to draw conclusions on how different sizes of the subject item may fit the subject consumer. For example, the purchase history of other members of the cluster may be analyzed to determine whether other members purchased a particular size of the item, and if so, the size purchased by the other members may serve as a basis to recommend a size that may best fit the consumer. For example, if other members of the cluster purchased a particular size, then that size may be recommended to the subject consumer, or if other members of the cluster purchased and then returned a particular size (e.g., for being too small), then another (e.g., larger) size may be recommended to the subject consumer.