How often do you return to an app, or decide to keep a streaming subscription? Retention measures a platform’s success at keeping users returning or active over a period. Its exact measure depends on the platform: for example, a social media service might count how many times a user opened its app in the past week. section 4.4 section 6.1
Retention is one of the metrics platforms may use to judge a recommender system. In an A/B test, a platform compares an existing model with a new one using predetermined measures such as retention and engagement. These measures define what counts as an improvement for the test; they do not establish that a recommendation is good in every sense. section 4.4 section 4.8
Keeping a streaming subscriber
If a subscriber cannot find anything appealing to watch, they may conclude that the service has little to offer and cancel. Recommendations can help users find content they enjoy, while also serving the platform’s goal of long-term retention. The two interests can overlap, but a platform’s goal of keeping users is not identical to the quality of any one recommendation. section 6.1
Retention is typically considered alongside measures such as time spent, engagement, and content consumption. Choosing it as a priority makes continued use a central sign of success, so the choice of metric matters when deciding which system to deploy. section 4.4







