This paper explores, in the movie industry setting, how the rise of big data presents firms with the opportunity to acquire knowledge that can influence product performance and shape firm strategy.

It asks the question: what is the value of insights gained from big data for firms? Specifically, the paper focuses on Amazon Instant Video’s customers who rented this [focal movie] also rented this…” lists to (1) evaluate similarity between movies based on users’ rental patterns, (2) show that movies that are implicitly similar to others have better box office performance than movies that are far from others, and (3) provide reasons for these performance differences.

This paper also explores the formation of implicit clusters of movies, made of movies with many co-rentals in common, and shows that they differ from usual classification schemes (genres). Observable characteristics such as genre, actor/director reputation, studio type and MPAA rating have modest power to explain why these clusters form. This points to the possibility that implicit similarity dimensions may exist within clusters. An initial exploration shows that one such dimension is theme, which transcends coarse classifications.

Exploring other possible latent common features could provide further insights to studios to help them make movies better targeted to different audience segments. Methodologically, this paper employs a combination of rigorous econometric methods and machine learning tools.