Stitch Fix uses a suite of machine learning algorithms to rank and recommend clothing items to each client, which human stylists then review to select a curated box of five items. The system takes in client style profiles, purchase history, fit feedback, and Style Shuffle ratings to predict which items a given client is most likely to keep.

Stitch Fix's main algorithm predicts 'probability of sale' by scoring each item in inventory based on the likelihood that an individual client will purchase it. The system ingests structured data (sizes, style preferences) and unstructured data (written feedback notes, Pinterest boards) processed via natural language processing using pre-trained language models like BERT and GPT-3. Stylists receive a pre-ranked, filtered shortlist from the algorithm and retain final selection authority; they can override algorithmic suggestions.