How does TikTok’s algorithm actually work?

After working in the space for a few months, I noticed that I have a very different take on algorithms than your average person, due to my day job. For context, my SFW day job heavily involves being able to make video content that succeeds on social media platforms. In my day job, I actually mentor video creators on how to create viral content. So, the topic of algorithms comes up often, especially when working with newer creators.

Successful creators have a different view of algorithms

Not surprisingly, people’s opinions of algorithms change once they become incredibly successful: all of the ones that I work with personally would unanimously agree: algorithms are not what most people think they are.

I find it's helpful to take a step back and just look at what the purpose of an algorithm is. An algorithm isn’t really created to celebrate some creators and not others. It’s not a mean machine that wants to suppress good content. In reality, algorithms are the opposite.

What algorithms really are

Every social media platform (TikTok, YouTube, Instagram, etc) has an incentive to keep users on their platform for as long a possible, so that they can generate as much money as possible through advertising.

Since these companies have a huge incentive to keep users on their platform, they create and iterate on an algorithm with one purpose: keep users engaged so they continue to consume content. The more content that users consume, the more money these companies make. These algorithms simply exist to feed content to users that they (the algorithm) have identified is interesting and engaging.

The algorithm is not really designed to “punish” people, it is designed to identify great content so that users stay on their platform for as long as possible.

TikTok videos that get “stuck” at 300 views

Now, have you ever noticed that a lot of people complain on the TikTok subs that their videos get stuck somewhere between 200-300 views? There's a reason for that! That's the TikTok algorithm at work: when a new video is posted, TikTok sends a little bit of traffic to that video - about 200-300 people to be exact.

When TikTok sends this traffic over, the algorithm is evaluating how well this content performs. It’s measuring things like watch time, engagement, and other markers for success. If that video gets shown to 200 people and the video has high likes/follows/watch time, the algorithm knows that this piece of content is beneficial for TikTok. What happens next? The algorithm will show it to more and more people, while continuing to measure those markers for success.

People who claim their videos get "stuck" at 200-300 views created content that the algorithm has deemed "not engaging" for one reason or another: maybe low watch time, low engagement, a combination of that and more. Why should TikTok's algorithm continue promoting a video that has proven to not be engaging? TikTok is not guessing what kind of content is interesting, it would not be beneficial for the company if they were just guessing. Instead, they are using data to make that decision.


What if your content used to do well, and now it doesn’t?

Whenever this happens, newer creators tend to believe that they are shadowbanned or that the algorithm simply hates them. This is most likely not the case - it’s much more likely that while you used to create content that users found engaging (hence the past success), your more recent content is simply less engaging. This could be due to so many reasons. Some examples are:

  • What worked in the past doesn’t work anymore due to niche saturation - now, the bar for what is “good” content has been raised

  • Perhaps your own subscribers engaged poorly with the content, telling the algorithm that your own fans didn’t like it, so the content should not be promoted further

  • New viewers (viewers who are not subscribed to you) are simply scrolling through your content quite quickly, indicating to the algorithm that this video is not engaging

Now, how does all this actually apply to a creator's work?

I think once a creator understands that the algorithm's sole purpose is to keep users entertained, they begin to craft content in a way that honors what the algorithm wants. If your content isn't doing well, you most likely aren't being shadow banned / the algorithm doesn't hate you / etc etc - you just aren't making content that the algorithm believes is "good content". In fact, you're probably simply making content that the algorithm has proven as "not engaging" in some way.

Then the question becomes, how do you make good content?

  1. Identify what content IS succeeding on the platform of your choice. Look for recent viral content across multiple creators in multiple niches. Literally put this in an excel sheet and write down things you find in common amongst the successes.

  2. Recreate the content that you KNOW is successful, but do it in your own way. This way you aren't randomly making content that you hope people will like, but instead you're rooting your content strategy in recent positive data.

To conclude: how to make the algorithm happy

Algorithms are quite complex and always evolving + improving so that the company can better identify what kind of content will keep users engaged. With that said, my guess is that there are two big components worth talking about.

Component #1: What happens when TikTok tests your content out with strangers? Do strangers like it? Do they watch the entire thing? Do they further engage with the platform? Or, do they scroll past it after a millisecond?

Component #2: How do your actual subscribers react to the content you post? This is a huge concept that receives a lot of emphasis when learning on YouTube in particular: if your own subscribers don't have a high thumbnail click-through rate and retention/watch time - that's often a bad sign as your own audience should be the most engaged. Since TikTok doesn’t have total transparency on what their algorithm looks for, we can only speculate. However, it makes sense conceptually that these two components have an impact on how the algorithm determines how engaging your content is.

Previous
Previous

How to turn unsolicited weiner photos into an opportunity for more revenue

Next
Next

What to do when a fan is chatty, but doesn’t buy anything.