How does the YouTube recommendation algorithm work


How does the YouTube recommendation algorithm work

YouTube video recommendations allow users to discover new content and creators related to their likes and videos they have watched, commented on and liked.

YouTube's recommendation system is based on the principle of helping people find the videos they want to watch, but have not yet discovered. In this way, YouTube displays recommended videos on its homepage, but also in the side panel, specifically in the Up Next area.

YouTube's recommendation system was created in 2008, and since then the different algorithms that select content have become very complex. Initially, the algorithm was limited to selecting the videos that performed best within the platform and displaying them in a tab called “Trends”. However, few users have seen these videos, as they rarely match their real interests.

What factors does YouTube consider when recommending a video?

Watch time: One of the main factors YouTube takes as a reference for viewing a video is watch time. The data from the platform shows that when a user watches an entire video, it is because they like the content, therefore, the videos with the longest watch time will be recommended to more users.

Publishing time: While some videos are shown in the recommendations engine from the first day they are posted, the most common thing is that a few days (sometimes even months) pass before the video is shown as a recommendation. This is an explanation, since YouTube has a lot of data about the oldest channels, so its algorithm knows immediately if a video is recommended. However, new channels do not have enough data, so they usually take longer to appear.

Channel size: One of the most common complaints among users and creators is that YouTube does not usually recommend content for smaller channels. Again, this is a data issue, because these channels provide less data for YouTube, so the algorithm doesn't "trust" the content enough to recommend it. However, it all depends on the post that is created, as a small channel that has a lot of interaction with users will see their content appear in the recommendation engine.

Subscriptions: While it is natural to think that subscriptions affect recommendations, the truth is that they do so to a lesser extent. The answer is that we often subscribe to channels from which we never see content. YouTube analyzes this behavior, so it does not recommend content of this type, and does not take into account the "subscriber" ratio of the channel so that its content appears in the recommendations. Again, content will appear in recommendations if users consume and interact with the content.

ليست هناك تعليقات
إرسال تعليق

    وضع القراءة :
    حجم الخط
    تباعد السطور