Organizing for Successful Data Science at Stitch Fix with Eric Colson - TWiML Talk #257

Algorithm Diversity

Eric Colson, Chief Algorithms Officer at Stitch Fix, highlights the diverse range of algorithms developed at the company, extending beyond recommendation systems. While only seven developers focus on recommendations, over 93 work on other crucial algorithms like inventory management and demand forecasting 1. These algorithms emerged naturally, driven by the right team and motivations, rather than top-down directives. Colson emphasizes the importance of quantifying their value, often through A/B testing, to determine their worth to the company 2.

   

Emergent Algorithms

Many of Stitch Fix's algorithms emerged organically through data scientists' curiosity and exploration. Colson notes that only one algorithm was explicitly requested, while others developed from unexpected insights during data analysis 3. This emergent behavior is supported by the company's organizational structure, which encourages innovation and exploration. Colson explains that this approach allows data scientists to discover new capabilities without being constrained by predefined goals 4.

   

Algorithmic Fashion

Stitch Fix leverages algorithms not only for recommendations but also in fashion design and merchandise management. Colson describes how a genetic algorithm was used to create new clothing designs by recombining old styles, demonstrating the innovative use of data science in fashion 5. Additionally, algorithms optimize merchandise buying by predicting the right quantities and sizes to purchase, enhancing efficiency and reducing excess inventory 6. This strategic use of algorithms significantly impacts the company's operations and client satisfaction.