Stitch Fix is an online retailer that sends customers a “fix” - a small box of clothing selected for them - rather than letting them browse a catalogue. The business is built on a deliberate combination of machine learning and human judgement, which the company documents publicly in its “Algorithms Tour.”
According to that tour, Stitch Fix uses a range of techniques: collaborative filtering and mixed-effects models to track how a client’s preferences evolve, neural networks for visual similarity, and natural language processing to interpret written requests. These models score and rank inventory for each client, but the company is explicit that machines and humans do different jobs - “while machines are great for tasks involving rote calculations, there are other tasks that require improvising, knowledge of social norms and the ability to relate to clients.” Algorithmically ranked options are handed to a human stylist who makes the final selections and adds personal notes. The same data science also drives warehouse assignment, stylist-client matching, logistics, demand forecasting, and even genetic-algorithm-inspired design of new garments.
For a general reader, Stitch Fix is a well-documented example of “human in the loop” AI in commerce: the algorithms narrow and rank the choices, but a person makes the call - a division of labour many businesses are now trying to copy.