Netflix is employing 300 people to maintain and improve its content recommendations, and spending a total of $150 million dollar on recommending movies and TV shows to its members every year, according to the company’s Chief Product Officer Neil Hunt. But the money is paying off for Netflix, as even improving recommendations a little bit could lead to dramatically increased revenue due to smaller churn, Hunt said during a talk he gave earlier this week at the 8th ACM Conference on Recommender Systems.
Hunt explained during his talk that [company]Netflix[/company] has a very limited window to convince a customer to watch something. The typical user only looks at the Netflix app one or two minutes, he said, and may browse 20 to 50 titles before either choosing something to watch or giving up entirely and doing something else.
That’s why Netflix is trying to find the best possible content for everyone, which can include diving deep into niche and even fringe content. “There are no bad shows, just shows with small audiences,” Hunt said. Embracing the long tail with its recommendations also helps to maximize the opportunity for long time content producers, he said, arguing that this could lead to a democratization of content distribution.
Hunt contrasted this algorithmic approach with regulatory efforts to promote certain kinds of content, like the ones in place in France. French laws demand that 40 percent of content aired by local broadcasters is of local origin, a rule that Netflix circumvented by launching its French subsidiary out of the Netherlands. “It’s a bad idea for culture in general, and for France in particular,” Hunt said. He added that some had even demanded that Netflix should give French regulators access to its recommendation algorithms to make sure that they wouldn’t favor U.S. movies.
That’s despite the fact that algorithms may actually help to promote French content worldwide. “If we do the right job with our recommender systems, we can truly enable a global culture,” he said. Hunt explained that Netflix currently treats algorithms by market, taking into account that U.S. audiences may like different movies and TV shows than viewers in Europe. However, in the next couple of months, the company will test globalizing its recommendations with a subset of its subscribers. In the end, it may be that global pools of data are even better in recommending niche content to niche audiences that regional pools of data, he mused.