What people gets wrong about the YouTube algorithm...

One of the biggest misconceptions about the algorithm is that it pushes your content out to viewers. The reality is: YouTube doesn't push your content to viewers, it pulls viewers to your content. So how does this all work?

The system uses your performance data (such as Average View Duration and Click-Through Rate) as input to try to understand and predict for each viewers. It's similar to the way you would ask a friend what movie you should watch. If you tell your friend if you like or dislike the movie, they'll have an idea of what to recommend you next. Todd Beaupré—YouTube's Senior Director of Growth & Discovery—describes the algorithm as automating word of mouth.

Recommendations can come in waves; meaning that your video will reach a certain audience, then six months later, there might be a resurgence because of a related video and it makes yours relevant again. This could also happen because of the news or nostalgia.

Other inputs YouTube takes in to account.

YouTube takes in factors like when you watch videos and what device you're watching it on. If you watch news in the morning and comedy at night, you're more likely to see these types of recommendations depending on the time of the day. The same applies to what devices you're watching, like on your phone or TV.

People have preferences on what kind of videos they watch on different devices. More often people would want to watch podcasts and series on TV while watching Shorts on their phone. Keep that in mind when you're creating content.

When you browse your home page, you may see a satisfaction survey from YouTube asking about the latest video you've watched. YouTube then collects these millions of responses and feed that directly into the recommendation system. The same goes if you choose to like or dislike a video, and whether you choose the "Not interested" or "Don't recommend channel" option.

Ultimately, YouTube wants to keep you on their platform. They want to build a relationship with their audience, the same way as creators. So the better they understand you, the better they can recommend.

Some bad (but also good) news for those who compare their channel with others.

The metric of data is dependent on who saw the video and how broadly it was distributed. Absolute metrics—like CTR, AVD, etc—can't be compared across channels. However, YouTube provides you with these analytics because most creators want to be able to benchmark and compare.

If you must compare yourself to other channels, take a look at the subscription tab to see what your core audience are reacting to without the influence of the algorithm. To see this, you need to be on YouTube Studio desktop. Head into Advanced Analytics > Traffic Sources > Browse Features > Subscriptions. Compare multiple videos from your channel to see whether your consistent audience is clicking and watching your videos or not.

Don't freak out about the ebb and flow of views.

Just because views are going down doesn't mean your channel's going to die. Typically, viewers will binge one channel for a while before moving on to another or life gets in the way and they need to ta. On the KDCC Discord, this is a frequent concern from creators. Just know that this is part of a natural flow. Don't expect to always be at your highest level of views. Channels that go down in views tend to go back up.

Instead of worrying about the view count, focus on responding to the feedback you're getting from your audience. There will be times when you need to move on from the subject or format that has worked for you in the past.

Sometimes, it comes down to supply and demand. When a trend picks up, there might only be a few channels and videos about it, thus creating a low supply with high demand. Eventually, more creators will join in on this trend, and the result can be a high supply with low demand. Keeping up with trends or predicting one is a good way to garner more viewers. You can do this through the Trends tab on YouTube Studio or on Google Trends.

When you look at your analytics, don't just check the last 28 days. Instead, look at the year to see if there are spikes. Then check for any seasonality. For example, during the holiday season, people are taking time off work or school and would likely be watching YouTube more. After the holiday, they'll return to their regular schedule and your views may decline.

YouTube is bringing new capabilities into their recommendation system.

With the introduction of multiple audio tracks, YouTube had to make their system aware that the video has multiple languages. They added a feedback loop to understand how each audio track were performing instead of looking at the video as a whole. This helps both YouTube and you know if the video's working in a particular language or not.

The benefit of having your videos in more than one language is being able to reach a broader audience. And we're not just talking about only dubbing one video, but at least 80% of your catalog. The more a viewer can choose from, the more they'll likely find another video they'll enjoy from you. YouTube says creators who dub 80% of their videos benefit more in watch time. However, it's better to focus on 80% of one language than 20% with multiple languages.

When you add in different language tracks, remember to translate your title and descriptions as well. When YouTube recommend your videos, the viewers of whatever language should be able to read and understand what sort of video they're going to watch.

The second piece that YouTube wants to bring in is a larger language model to the recommendation system. What this means is that the larger model can understand more nuance. In a smaller language model, it knows that the viewer enjoys cooking videos and memorizes that. With the larger model, it might be able to understand the ingredients, the emotions from the creator, how the video is presented, and more.

An example YouTube chose is the difference between a cook (smaller model) and an expert chef (larger model). A cook can read the recipe, which is a fixed set of instructions. An expert chef can not only follow the recipe, but understand the fundamentals behind cooking and can dynamically respond to different conditions. If someone says they can't eat meat, the expert chef is able to adjust while still making a delicious dish.

​

This will be an interesting upgrade to YouTube's recommendation system, but we'll have to see how this will work out. If you haven't checked out the video from Creator Insider, we highly recommend you do!

​