A marketer's guide to how TikTok's algorithm works
Note: A version of this story ran in Episode 4 of my Marketing & The Machine podcast.
In a call with reporters last month, TikTok executives revealed some of the inner workings of their apps’ algorithm that keeps hundreds of millions of users hooked FOR AN AVERAGE OF 52 MINUTES A DAY!
The algorithm that has boosted TikTok’s valuation to somewhere in the ballpark of $30 billion relies on machine learning which learns to find content you’ll initially engage with…and then serves videos that are liked by similar audiences.
Here’s how it breaks down. When you first open the app, TikTok serves you 8 popular videos featuring different trends, music, or topics. After you’ve sped through that first batch of 8, the app serves up a fresh set of 8 more videos based on how you engaged with the initial set.
From that point, TikTok can then start pinpointing new videos that similar groups of people like that you also have engaged with.
And all this data is processed in real-time too, so when you click a new video, your feed will quickly change based on your latest actions.
For those of you scoring at home, data scientists refer to this type of algorithm as a recommendation engine. At its core, it’s nothing too much different than what Netflix, Amazon, or YouTube offer.
And note, TikTok doesn’t just take into account the videos you like, they’re looking at your location, whether you have an Android or iPhone, language settings, and more.
The more videos you comment, like, or engage with the better their algorithm gets. In short, they’re putting you in proximity with similar users or what the company calls “clusters.” You and I might call it our own little filter bubble. Ah, home sweet home.
TikTok did acknowledge that their algorithm can contribute to filter bubbles, which then could spiral into misinformation, hoaxes, and conspiracy theories, but they are trying to get ahead of the issue by using all their data to detect unusual patterns in real-time before they spread.
The good news is it’s not too late to embrace TikTok. If you’re a brand or influencer just getting started, TikTok says that neither follower count nor whether the account has had previous high-performing videos are direct factors in the recommendation system. So you can make a name for yourself if you hit on the right niche.
The meat and potatoes of what you should focus on (and what the algorithm puts more weight on) are things like (1) user interactions - follows, comments, likes, and shares, (2) video information - sounds, hashtags, or captions within the content, and (3) device information like I mentioned above.
As a tangible takeaway that means it’s a good idea to post often, using suitable hashtags, encouraging people to comment (and do so yourself), giving your videos a clear description, bringing in influencers when you can, and then slowly and subtly working in your branded content.
Want to get really geeked out about TikTok’s inner workings? Check out this explainer covering TikTok’s infrastructure and its recommendation engine’s workflow.