Explore-Exploit algorithms used at IKEA recommendations

Ikea has a digital product called the Inspiration Feed which is available in the web and in the IKEA App.

The Inspiration Feed contains “shoppable” images in which IKEA products are displayed in purposefully atmospheric settings for a wide range of room types. In the Inspiration Feed, we apply contextual bandits to deliver personalized recommendations for customers and optimally balance data exploration and exploitation.

I will explain why we chose contextual bandits and why it is optimal for exploration and exploitation in our case and how we scale it for 300 million yearly users.

Tóth Balázs
Senior Data Scientist, IKEA

Balázs Tóth works as a Senior Data Scientist at IKEA where he builds modern recommender systems. He has 9+ years experience in the internet industry where he built multiple machine learning solutions for recommendation and fraud detection.