AI Case Study

Google YouTube drives 60% of video clicks through its machine learning recommendation system for users

YouTube has developed a video recommendation system to increase the amount of user time spent on the site. In order to do so, the relevancy of the content shown to users must be optimised, which has been done through machine learning based on past user behaviour.

Industry

Consumer Goods And Services

Entertainment And Sports

Project Overview

YouTube states that "the goal of the system is to provide personalized recommendations that help users find high quality videos relevant to their interests. In order to keep users entertained and engaged, it is imperative that these recommendations are updated regularly and reflect a user’s recent activity on the site. They are also meant to highlight the broad spectrum of content that is available on the site."

Reported Results

"The recommendations feature has been part of the YouTube homepage for more than a year and has been very successful in context of our stated goals. For example, recommendations account for about 60% of all video clicks from the homepage."

Technology

"To compute personalized recommendations we combine
the related videos association rules with a user’s personal
activity on the site: This can include both videos that were
watched (potentially beyond a certain threshold), as well as
videos that were explicitly favorited, “liked”, rated, or added
to playlists."

Function

Marketing

Product Marketing

Background

"Users come to YouTube for a wide variety of reasons which
span a spectrum from more to less specific: To watch a
single video that they found elsewhere (direct navigation), to
find specific videos around a topic (search and goal-oriented
browse), or to just be entertained by content that they find
interesting. Personalized Video Recommendations are one
way to address this last use case, which we dub unarticulated
want."

Benefits

Data

User site activity - videos watched, uploaded, rated, favourited