Google Reveals How YouTube Recommends Videos

Last month YouTube’s VP of Engineering, Christos Goodrow, revealed a plethora of details about the myriad elements that comprise Google's recommendation algorithm for videos on Youtube. The number one factor being personalization for each individual logged-in user.

According to the post,

Recommendations drive a significant amount of the overall viewership on YouTube, even more than channel subscriptions or search. ... But all too often, recommendations are seen as a mysterious black box. We want these systems to be publicly understood

What is a recommendation system?

Google's recommendation system is an attempt to help people find videos they want to watch. They are displayed in two main places: the homepage — what you see when you first open YouTube— and the Up Next panel.

The homepage is a mixture of personalized recommendations, subscriptions, and the latest news and information. The Up Next panel appears when a user is watching a video. It suggests additional content based on what the user is currently watching, displayed alongside other videos that Google thinks you may be interested in.

Recommendations are highly individualized and heavily based on personalization. Google compares a user's viewing habits with those that are similar, using that information to suggest more content. This means that, if a user likes tennis videos and their system notices that others who like the same tennis videos also enjoy jazz videos, then jazz videos may be recommended even though that user has never before watched a jazz video on YouTube.

How Google personalizes recommendations

The algorithm that produces recommendations has evolved over the years. In 2008 it was based solely on popularity. Over the years a number of refinements have been layered into the formula being used today as seen in the graphic below...