In its early days, Netflix faced the challenge of effectively offering movie recommendations through its Cinematch algorithm. This system, fundamental to the platform’s strategy, had to deal with the polarization of opinions on films like ‘Napoleon Dynamite’, which received extreme ratings, with one or five-star ratings predominating. This situation greatly complicated the task of recommending titles to subscribers.
The Curious Journey of Napoleon Dynamite
In 2007, the company launched a competition offering a million dollars to those who could improve Cinematch by 10%. This initiative attracted approximately 30,000 participants willing to solve the riddle of recommendations, facing challenges similar to those posed by the aforementioned cult comedy. The algorithm, which had been refined over several years, was still not up to the task of addressing the diverse preferences of users.
The surprising box office success of ‘Napoleon Dynamite’, destined to be a polarizing film, highlighted the limitations of the algorithm, leading the contestants to focus on this and other hard-to-classify titles. The participants competing for the prize also encountered other complicated films like ‘Fahrenheit 11/9’ and ‘Kill Bill’, which presented similar challenges.
In 2009, the BellKor’s Pragmatic Chaos team won the challenge by achieving a 10.06% improvement in Cinematch. However, over time, Netflix has changed its approach. Instead of focusing on personalized recommendations, the company began to promote its latest releases, recognizing that users were more interested in what their friends enjoyed through platforms like Letterboxd. This shift has sparked criticism about the dehumanization in the content selection process, a change that reflects the evolution in viewer preferences.