Skip to content

A Global Approach to Recommendations

Carlos Gomez-Uribe

Vice President, Product Innovation

A Global Approach to Recommendations

When Netflix launched in 130 additional countries in January, we wanted our service to work well everywhere in the world, and that meant solving many complex technical problems with a global approach.

For example, the personalized recommendations we offer members are central to the Netflix experience, helping people find something great to watch quickly and easily. In general, our algorithms analyze vasts amounts of data to produce these recommendations, and making this process work globally was challenging on several fronts. In this blog post, my colleagues Yves Raimond and Justin Basilico dive more deeply into the technology challenges we faced. After an entire year, efforts from dozens of teams across the company, and intensive research, we developed and deployed a global recommendation system that will benefit Netflix members across the world.

Through this journey, we saw that great stories transcend borders, and that viewers around the world have more in common than they may realize. For example, one way that Netflix generates personalized recommendations for individual members involves identifying communities of other members with similar movie and TV show preferences, and then making recommendations based on what is popular within that community. Rather than looking at audiences through the lens of a single country and catalog, Netflix’s global recommendation system finds the most relevant global communities based on a member’s personal tastes and preferences, and uses those insights to serve up better titles for each member, regardless of where he or she may live.

Simply put, tapping into global insights makes our personalized recommendations even better because now our members benefit from like-minded viewers no matter where they are in the world. While this is especially helpful if a member is in a new or smaller market, we’re also able to better serve members in larger, more established markets who have highly specific or niche tastes.

Take Anime for example. Our data helped us identify a community of members that really enjoys the specific type of Anime exemplified by the recommendations below:

Although it is not surprising the country with the largest representation in this community is Japan, it’s important to note fewer than 10% of people in this community are actually in Japan -- the rest come from all over the world! In this case, pooling data across all countries for this community really helps us improve our recommendations for all Netflix members in this group, no matter where they actually live.

Another example of what our global recommendation system means for members around the world comes from the global community of health-conscious foodies, who are very interested in learning about food and the industry around it. The recommendations below exemplify the types of videos that this global community enjoys:

The percentage of members from each country in this community is actually relatively small. So if we were relying just on the data from a single country (especially a new one with a smaller number of members), our personalized recommendations would suffer as a result. By leveraging data from across the world and countries of all sizes, our global algorithms are able to tap those insights to make recommendations for this food conscious community that are more accurate and robust.

The global recommendation system also complements the work we are doing to improve localization around the world, with the common goal of delivering a more personalized service to members everywhere. While our localization efforts center on adding local content, languages, payments, etc., the global recommendation system is separately allowing members to benefit from being part of the global community we are building.

-Carlos Gomez-Uribe