What's special about TikTok's technology?

Here's how this algorithm works, and why it's causing more controversy than the technology used by competitors like Meta's Instagram, Google's YouTube, and Snapchat.

Algorithm

Algorithms are considered core to ByteDance's entire operations, and according to Reuters sources, ByteDance has said it would rather shut down the app than sell it to another company.

Additionally, the Chinese government amended its export laws in 2020, giving the government approval power over any export of algorithms and source code, further complicating any attempt to sell the short video app TikTok.

It's not just the algorithms themselves, but also the way they work with the short-form video format, that has made TikTok such a global success, researchers and former employees of the company say.

Before TikTok's emergence, many believed that technology that connects users' social relationships was the secret to the success of a social networking app, based on the popularity of Facebook and Meta's Instagram.

But TikTok has proven that an algorithm that uses user preferences to personalize their experience can be more powerful. Instead of building their algorithm based on a “social graph” like Meta, TikTok executives, including CEO Shou Zi Chew, have claimed that their algorithm, based on “interest signals,” is a powerful tool for personalizing the user experience.

Ms. Catalina Goanta - Associate Professor at Utrecht University (Netherlands) said that although competitors also have similar interest-based algorithms, TikTok can boost the effectiveness of the algorithm thanks to the short video format.

“Their recommendation system is very popular. But what really sets TikTok apart as an app is the design and the content,” Catalina Goanta added.

The short video format allows TikTok's algorithm to be much more dynamic and even capable of tracking users' changing interests and preferences over time, down to how granular a user's interests might change over a given period of time during the day.

Fast data collection

The short video format allows TikTok to learn about users' interests at a much faster rate, said Jason Fung, TikTok's former head of gaming.

“Because videos are small and short, you can collect data about user preferences much faster than YouTube. Imagine collecting data about the average user every 10 minutes versus every few seconds,” added Jason Fung.

And positioning TikTok as an app built for mobile devices from the start also gives it an advantage over competing platforms that have to adapt their interfaces from a desktop screen.

TikTok’s early entry into the short-form video market also gives the company a major first-mover advantage. Instagram didn’t launch Reels until 2020, while YouTube launched Shorts in 2021, both of which have years of experience in data collection and product development behind TikTok.

Explore new content

Additionally, TikTok frequently recommends content outside of users' interests, something the company's leadership has repeatedly asserted is essential to TikTok's user experience.

A study published last month by researchers from the US and Germany found that TikTok's algorithm "mined user preferences in about 30% to 50% of video recommendations" after examining data from 347 TikTok users and five automated chatbots.

This finding suggests that TikTok's algorithm is opting to recommend a large number of discovery videos in an attempt to better infer user interests or maximize user retention time by recommending more videos outside of a user's interests.

Attract users to public groups

TikTok uses a variety of strategies to attract users to public groups, build community, and drive engagement on its platform. Some of the key strategies include suggesting groups based on interests, highlighting trending group challenges and hashtags, searching for groups, and suggesting groups based on users’ interests and activities, among others.

Mr. Ari Lightman, a professor at Carnegie Mellon University (USA), said that by encouraging users to form public groups, TikTok can learn more effectively about users' behavior, interests, similarities and ideologies.

By implementing these strategies, TikTok effectively engages users in public groups, fostering a sense of togetherness, participation, and shared experience, thereby enhancing the overall user experience on the platform.

Hire a data tagging expert

TikTok's recommendation algorithm is largely taken from Douyin, TikTok's Chinese version, which launched in 2016. While ByteDance often emphasizes that TikTok and Douyin are separate apps, a source with direct knowledge of the matter said the algorithms for both apps are the same.

“Around 2018 and 2019, Douyin was trying to tag every user. So they would manually tag every video clip. Then they would tag their users based on the videos they watched. Then they applied this tactic to TikTok as well,” said Yikai Li, a former executive at ByteDance.

While hiring data taggers is now a common and important practice for AI companies, ByteDance is one of the pioneers in adopting this strategy.

“Sorting these cards requires a lot of effort. It takes a lot of time. So Chinese companies have an advantage here. You can hire more people, and it’s cheaper than North American companies,” said Yikai Li.