Ever since you “discovered” TikTok, you have been fascinated by this platform and have always wondered about the “secrets” behind how it works. How does it choose the content to receive extra views for TikTok and more engagement? How does it understand what interests them? You have been asking yourself these questions for some time, and today, we want to give you the answers you are looking for.
Those who use TikTok frequently know the feed (the stream) of recommended videos seems perfectly designed to keep you glued to your smartphone screen for as long as possible. The algorithm can read our mind and state of mind: it knows which are the right strings to pull, the topics we are most passionate about, and those that motivate us to share a video with one of our contacts.
In the upcoming sections of this article, we’ll tell how the TikTok algorithm operates. Our insights are based solely on the official details shared online, steering clear of the rampant speculations found on the internet.
Given that TikTok’s algorithm and underlying code aren’t open to the public, we lean on the platform’s official communications for clarity. With that out of the way, let’s discuss it. We’ll first unpack the mechanics of the TikTok algorithm. We’ll then offer strategies to improve this knowledge, suitable for casual users and content creators. Happy reading!
How the TikTok Algorithm Works
Since TikTok (like all social platforms) has a certain interest in keeping the attention of its users to keep them active on the platform for as long as possible (which significantly increases the platform’s advertising appeal to advertisers), its algorithm tries to ‘understand’ what each user is interested in. When it succeeds in ‘hitting the mark’ by proposing content that interests it, the user does not abandon the platform, and the time they spend looking at content increases considerably.
In December 2022, the developers of the famous Chinese social network decided to reveal some information that would help us to understand better how TikTok’s algorithm works and, to go into a little more detail, that allows us to understand based on which factors the platform proposes content to users in the ‘For You’ section, which I mentioned in the previous paragraph, and which is for each individual ‘TikTok’.
The contents to be proposed are classified by the algorithm based on a combination of factors, starting from the interests that the user expresses at the moment in which he subscribes to the platform (not by chance; this is one of the first things that is asked of those who subscribe to TikTok) and then going on to ‘correct the shot’ as the user uses the platform since the factors that I list below are analyzed.
- User interactions – all the actions carried out by the user with their account (likes, comments, content shared on the platform and content created, accounts followed, etc.) heavily influence the content that the TikTok algorithm will propose to see later.
- Data present in the content – captions, sound bites and hashtags are some of the information present in the content one interacts with, and these also influence the algorithm, which takes them into account when feeding its recommendation system.
- Popularity in the region – when a piece of content becomes very popular in the geographic area where the user is located, it will likely be proposed for viewing by the algorithm.
- Freshness of content – the algorithm is unlikely to show old content, as it tends to show newer content.
- Device and account settings in use – language preference, country setting, and type of device in use- allow the algorithm to optimize its performance.
Once the algorithm has analyzed and combined all these factors, it processes the information it can obtain and ‘decides’ whether a piece of content interests a specific user (obviously, these ‘decisions’ are made by the algorithm very quickly). Videos are thus ‘ranked’ internally by the platform to determine the likelihood of a user’s interest in a piece of content to be shown in its ‘For You’ section.
If you want to see why certain content was proposed to you by the algorithm and what ‘evaluations’ it made specifically, know that TikTok offers such a possibility through a dedicated function, which you can call up by pressing the Why this video? Button attached to the content of your interest. The message on the screen will tell you why the video was proposed to you (e.g. Why do you follow [username], Why is this video popular, etc.). Interesting.
Guillaume Chaslot, founder of Algo Transparency (which studied YouTube’s recommendation system), reviewed the document at Smith’s request. “This system means that viewing time is critical. The algorithm tries to make people addicted rather than give them what they want,’ he added a reflection on children: ‘With every video a child watches, TikTok gets a piece of information about them. In just a few hours, the algorithm can detect his musical tastes, his physical attractiveness, whether he is depressed, whether he might be on drugs, and many other sensitive pieces of information. There is a high risk that some of this information could be used against him. It could be used to micromirror him or make him more dependent on the platform’.
However, according to reports in the New York Times, viewing time is not the only factor considered: every video on the platform has a score, which is achieved through an equation based on machine learning and actual user behavior, mixing data on likes, comments, and playback time. So the recommendation system scores all videos based on this equation and returns those with the highest scores to users, the document says.
Among the company’s goals is also ‘monetization of the creator’: this might suggest that the videos most frequently shown by the platform are not only the most entertaining but also the most profitable. The company let the C Corriere know through a spokesperson that ‘TikTok wants to bring joy to people, and we know that too much exposure to any one theme doesn’t fit the diverse discovery experience we want to create. That’s why we continue to invest in creating new ways to customize content preferences, automatically skip videos that aren’t relevant or age-appropriate, and remove content that violates our community guidelines.”
Algorithmic magic? An unsuspected secret formula? According to University of California at San Diego computer science professor Julian McAuley, reached by the New York Times, absolutely not. The professor called the description of TikTok’s recommendation engine ‘absolutely reasonable, but traditional’, arguing that the company’s advantage does not come from a surprising equation but from the combination of machine learning with huge volumes of user data and the strong engagement of the people who use the app. Once again, the linchpin is data.
Nothing transcendental, then. Another thing the document reveals, however, is TikTok’s close ties with parent company ByteDance and the Chinese version of social, Douyin. Have we already forgotten the war that former US President Donald Trump waged on the Chinese app, intending to force ByteDance to sell the video social to an American company, Oracle? Fears that the Chinese parent company was so anchored to its successful son TikTok were well-founded, and this document is proof of that. Fears that disappeared with the change of tenant in the White House?
Not really, because concern about Chinese-originated technology has no political color in the United States. In August 2020, President Trump issued an executive order aiming to prohibit TikTok. The order stated that TikTok’s data-gathering practices could potentially grant the Chinese Communist Party access to Americans’ personal and confidential details. This, in turn, could enable them to compile personal profiles for potential blackmail and engage in industrial spying. The ban stalled in court and disappeared after Biden arrived, who rescinded the executive order but at the same time announced an investigation into the security threats posed by TikTok.
On this point, company spokeswoman McQuaide said: ‘TikTok has never provided user data to the Chinese government, nor would we do so if requested’. Moreover, the company tried to allay American concerns by storing user data in the US, with a backup in Singapore. As if that were not enough, a report earlier this year by Citizen Lab, the Toronto-based cybersecurity watchdog organization, says it found no indication that TikTok was censoring sensitive topics or transmitting data to China.
While currently, there’s no solid ground to question TikTok’s data safety or any Chinese government involvement, one glaring concern stands out: the app’s compelling algorithm. This isn’t by chance; it’s why it’s the top video app globally in such a brief span. The best approach?
As with everything, balance is key. This stems from being more conscious of our online actions. Does an account’s fame sway the algorithm? Surprisingly, no. Even if a video from a popular account might get more eyeballs (given the vast audience), the follower count and past video hits don’t directly dictate TikTok’s content suggestions.
Although there is a chance that a video will receive more views if it is shared by an account with many followers (which is to be expected if you have a very large user base), the actual number of followers and the presence of any videos published in the past that have performed well in terms of visibility, are not factors that influence the recommendation system of the TikTok algorithm, at least not directly.
How to Use the TikTok Algorithm to Your Advantage
Understanding the TikTok algorithm’s intricacies, you might ponder how to harness its power for your benefit. The strategy for casual TikTok users aiming to curate a feed for their tastes is straightforward: Engage originally. Avoid mindlessly liking or following accounts in hopes of reciprocation. Such actions might muddle the algorithm. Instead, let your interactions guide the platform, refining its content suggestions for you.
So, the recommendations in the last section can be helpful for content producers. Specifying terms or hashtags related to content that users do not want to see in their video stream is one of the features the service is working on. Because TikTok’s algorithm is based on user preferences, assessing whether content resonates with the target audience is imperative. Consider questions like “What captures my current and potential followers?” or “What keywords do my videos encapsulate for hashtag use?”
One successful tactic is niche focus. Preferably, rather than producing videos with a broad theme, target a unique segment for a distinct group of users. Creating content that speaks directly to them, the content should use a combination of clarity and creativity. This improves the visibility of the video and creates a dedicated viewer base. Just remember that in the large sea of TikTok, the unique and authentic voices often resonate the loudest. So whether you are a casual scroller or an aspiring TikTok star, understanding and adapting to the algorithm can significantly increase the number of views.
