YouTube released a handy new rundown of how its recommendation system works to help developers build an audience on the platform and reach more viewers with their clips.
The overview is presented by Rachel Alves, a product manager for Discovery on YouTube. Alves has provided various insights into how the platform’s recommendation systems work, and she notes that this presentation is particularly used at conferences and the like to help people understand the systems that affect the diffusion of videos.
As noted by Alves:
“You don’t have to be an expert on analytics algorithms to be successful on YouTube.”
Alves begins with an overview of the core objectives of YouTube’s recommendation systems.
As mentioned here, the main goal is to keep users from coming back by making sure they have a good experience on the platform.
“This is how we maximize long-term satisfaction so that viewers keep coming back to YouTube.”
That, of course, is what you would expect. Alves notes that YouTube has changed what it optimizes over time.
“When we go back to 2011, we’ve optimized for clicks and views […] But that’s not a great metric because it indirectly incentivizes clickbait-y or sensational titles or thumbnails that get people to watch a video but don’t make them very satisfied or happy. “
Alves says a lot of the feedback YouTube received in the early days of its algorithmically defined feed was that people’s home feeds were being filled with “sensational or disgusting videos” so the focus on showing time as a key metric Instead, the switch was made in 2012.
“How much time someone spends watching a video or channel is a lot more indicative of the quality of the content. If you spend more time watching something, the more likely you are interested in it.”
However, Alves says that time is not a perfect metric either, because while you may spend more time watching something, it doesn’t necessarily mean that spending the time afterwards will make you feel good in the end.
For this reason, YouTube has since tried to better define the viewing time “quality or value” and to optimize user satisfaction.
YouTube goes through this:
- User surveys help optimize what people like and enjoy (Alves says they send out millions of user surveys every month).
- Prioritization of relevant content from recognized, established sales outlets, especially for news content
- Reduction of the spread of “borderline violative content”
Hence, the focus is on both user satisfaction and ensuring that users are comfortable about their experience on the platform and taking responsibility for what is reinforced by the recommendations.
YouTube, of course, faced various challenges in this regard. The platform has been regularly reviewed to add to controversial content such as misinformation, conspiracy theories and politically divisive material. There’s no perfect solution to this, but YouTube tries to take these considerations into account in order to refine its approach to the videos it recommends to users.
As mentioned earlier, regarding polls, YouTube sends millions of polls to viewers every month and gathers feedback on a wide variety of video uploads.
Alves says they don’t share this information with the creators at the moment because often they don’t have enough feedback on each individual clip to provide useful feedback, but they can use the information to better inform their algorithms and systems.
“We want to add more satisfaction data and pass it on to the developers. So that’s what we’re working on.”
And yes – as you can probably expect, in addition to direct feedback via survey prompts, YouTube also uses signals such as:
- When people tap / click Not Interested in the menu for individual videos
- Likes and dislikes about clips
- Shares of clips
As with other social platforms, these actions are important in defining the reach of videos through YouTube’s recommendation interfaces (e.g. the home page and recommended entries). As much as YouTube wants to rely on deeper user feedback to determine whether a video is providing a good experience for viewers, these lower, more immediate response metrics also play a role in determining your performance.
Alves also notes that different algorithms are actually used on the homepage and in the proposed lists. Therefore, the idea that there is a central YouTube algorithm is incorrect.
“The home page has a wide variety of videos when you visit youtube.com and uses signals similar to ‘Suggested’, but they are designed to do slightly different tasks.”
With this in mind, Alves notes that developers often want to know how to optimize for each element – which Alves says you can’t.
“You can’t optimize for a traffic source, only for people or viewers.”
With that in mind, Alves says developers who want to maximize the home page views should try to look at their content from the perspective of someone who has been recommended based on their interest in similar clips, but whose content may not yet be more specific Channel.
A reference to a joke with your audience might work in “Suggested” as these are more tailored to each video clip. However, in the home feed, you want a more general appeal and reason for newer viewers to click on a related interest.
As you can see here, Alves also notes that posting consistently can help keep your videos showing up on users’ relevant home feeds and bringing them up to date if they have already watched some of your other clips.
In the “Suggested” feed, Alves says these highlighted clips should show viewers what to watch next after the video they are watching.
This means that the suggested feed will be more closely aligned with the current clip.
In a sense, you can think of the two surfaces as the “top of the funnel” or as a viewer of more general but related interest with the Home Recommendations and then as the “middle of the funnel” with “Suggested” as those viewers have already shown specific interest on your content by tapping through, and now the recommendations are more closely aligned.
Alves says the most effective tactic that creators use to maximize their appearances on people’s “Suggested” lists is to develop a video series or create currently related videos that differ from one another.
Alves also recommends using a consistent title and thumbnail style.
“You can imagine that when a viewer is looking at everything they might be looking at next, they have many options. When you have really strong, identifiable branding that is consistent, it’s really easy to figure out which of them are Videos it comes from your channel, and it only makes that decision all the faster for viewers. “
Alves also notes that CTA buttons to show more, as well as playlists and end screens, are also powerful tools to encourage viewers to keep watching your content.
Alves’ main lesson from this review is that the YouTube algorithm is designed to “find videos for viewers, not viewers for videos”.
“Sometimes the developers feel that the recommendation system is moving or promoting videos to viewers. In reality, the system is designed to work the other way around when a viewer visits youtibe.com and then gives a recommendation system and then ranks.” the best candidate for this viewer depending on the page they’re on. “
So when you get to the home page, YouTube tries to show you the content you want to watch based on personalized recommendations (e.g. history, location, trend, etc.) as you click through to specific content as the video becomes the “Suggested” content is largely defined by that particular clip. YouTube’s system is not designed to reinforce specific clips or creators as such, but rather its overall goal is to align with the interests of the individual.
This seems like a logical process, but it is also an important point of clarity in this context.
Essentially, you want to create content that speaks to your target audience and then consistently build on themes and themes to keep your viewers coming back, while maintaining branding elements to strengthen those connections. Part of this is achieved through research and understanding of what works in your niche. However, establishing a strategic approach and sticking to that process is also key to building a YouTube audience over time.
Here are some valuable pointers that could be helpful in your platform planning. Given that YouTube serves more than 2 billion monthly active users and there has been significant growth in television on televisions, this should at least be included in any marketing plan.