YouTube's recommendation engine is one of Google's most successful innovations. An astonishing number of 70% watch time on YouTube is dictated by YouTube's recommendations.
Despite this, the SEO industry tends to focus on "19459005] search engine and focus on ranking in YouTube search results or getting YouTube ads in the results Google search.
However, the SEO sector rarely refers to this document.
This article explains what it contains and its impact you are approaching SEO for YouTube.
To this day, metadata remains much more important for SEO on YouTube than for search results in Google.
While YouTube can now create automatic subtitles for videos and their ability to extract information from the video has improved significantly over the years, you should not rely on them if you want YouTube to recommend your video
YouTube's recommendation algorithm paper states that metadata the fact that metadata is often incomplete or even completely missing is an obstacle that their recommendation engine is also designed to overcome.
To avoid forcing the recommendation engine to do too much work, make sure you are filled with the right information with every video you download:
Include your target keyword in the title of the video. The title also attracts attention and arouses users' curiosity.
Perhaps the most attention-grabbing titles on YouTube are probably more important than traditional research, as the platform is based more on recommendations than on research results.
Include a full description that uses your keyword or variant, and make sure its length is at least 250 words.
include here, the more data that YouTube has to work with, allowing you to capitalize on the long tail.
Include the main points you will cover in the video and the main issues that you will address.
In addition, the use of descriptions associated with other videos, from the point of view of the user, can help you find the recommendations for these videos.
still matter on YouTube, unlike the keyword meta tag for search engines, which is completely obsolete.
Include your main keyword and all variants, related topics that appear in the video and other YouTubers mentioned in the video.
Include your video in playlists with related content and recommend your playlists at the end of your videos.
If your playlists work properly, your YouTube video users longer, leading to the display of your video in recommendations.
Use an eye catching vignette. Good thumbnails usually include text to indicate the subject and a catchy image that creates an immediate emotional response.
Although automated YouTube subtitles have misinterpretations of your words. If possible, provide a complete transcript in your metadata.
Use your keyword in your file name. It probably does not have as much impact as in the past, but it certainly does not hurt
2. Video Data
The data in the video becomes each more important days.
The YouTube recommendation engine paper explicitly refers to the raw video stream as an important source of information.
] Since YouTube is already analyzing audio and generating automatic transcripts, it is important that you indicate your keyword in the video itself.
Reference the name and YouTube channel of all videos you answer in order to increase the odds you show in their video recommendations.
Finally, it may become more important to rely less on the "talking head" video style. Google has a Cloud Video Intelligence API able to identify objects in the video.
Include videos or images in your videos that reference your keywords
3. User Data
Needless to say, we do not directly control user data, but we do not understand the operation of the recommendation engine or its optimization without understanding the role of the data user.
YouTube recommendation engine paper divides user data into two categories:
Explicit: This includes videos and video channel subscriptions. Implicit:
To optimize user data, it is important to encourage explicit interactions such as approval and subscription, but it is also important to create videos that lead to good implicit relationships. user data.
Audience retention should be closely monitored, especially relative retention of audience.
Videos whose relative audience retention is mediocre hy, and videos with particularly poor retention must be removed to not detract from your overall channel.
4. Understanding Co-Visit
Here's where we start getting into YouTube's recommendation engine.
The YouTube article explains that the ability to map a video is a fundamental part of the recommendation engine. to a set of similar videos.
It is important to note that similar videos are defined here as videos that the user is more likely to watch (and probably watch) after viewing the original video. the content of the videos is quite similar.
This mapping is done with the help of a technique called co-visit.
The number of simultaneous visits is simply the number of times that two videos were viewed. over a given period, for example 24 hours.
To determine the link between two videos, the number of simultaneous visits is then divided by a normalization function, such as the popularity of the candidate video.
In other words, if the number of co-visits is high in two videos, but the video candidate is relatively unpopular, the relationship score for the candidate video is considered high.
In practice, the kinship score should be adjusted by considering how the recommendation engine itself promotes co-viewing, watch time, video metadata, and so on.
In practice, this means that if your video From the recommendations, people who watch another video should also watch your video quickly.
There are several ways to accomplish this:
Create response videos quickly after a video. The initial video is created. Post videos on platforms that have also uploaded traffic to another popular video. Select keywords associated with a specific video (as opposed to a larger subject). to watch your other videos.
5. Factoring In-User Personalization
YouTube's recommendation engine does not simply provide videos with a high kinship score. The recommendations are customized for each user and the procedure to follow is explicitly described in the article.
For starters, a set of videos is selected, including videos that the user has watched, weighted by factors such as
For the simplest recommendation engines, videos with the highest relationship score would then simply be selected and included in the recommendations.
However YouTube discovered that these recommendations were simply too narrow. The recommendations were so similar that the user would probably have found them.
Instead, YouTube has expanded the recommendations to include videos with a high kinship score for these initial recommendations, and so on. number of iterations.
In other words, to appear as a suggested video, you do not necessarily need a high number of co-hits with the video in question. You could do this by having a high number of co-visits with a video that in turn has a high number of co-visits with the video in question.
For this to work eventually, your video will also need to figure prominently in the recommendations, as shown in the next section.
6. Ratings: Video Quality, Specificity, and User Diversification
YouTube's recommendation engine is not content to rank videos by the highest degree of kinship. Being in the top N's best kinship scores is simply passing / failing. Rankings are determined using other factors.
The YouTube article describes such factors as video quality, user specificity and diversification.
] User Rating.Comment.Favoriting.Sharing.Upload time.View count.
The article does not mention it, but session time has since become the determining factor here, in which videos driving the user to spend more time on YouTube (not necessarily on this video or this YouTube channel) rank better.
Specificity of the user
These signals optimize videos based on the user's history. It essentially consists of a kinship score based on the user's history, rather than on the original video in question.
Too similar videos are removed from rankings so as to present users with a more meaningful selection of options.
This is accomplished by limiting the number of recommendations using a particular start video to select candidates or by limiting the number of recommendations from a specific channel.
The YouTube recommendation engine is critical to the way users interact with the platform.
Understanding how YouTube works will dramatically improve your chances of success on the world's most popular video site.
Consider what we discussed here, consider adding a look to the newspaper, and integrate that knowledge into your marketing strategy.
Other referencing resources on YouTube:
Subscribe to SEJ
Receive our weekly Loren Baker Newsletter , founder of SEJ, on the latest industry news!