Core recommendation algorithm and underlying logic for TikTok

Many students who work in TikTok do not understand the underlying logic of TikTok, which is actually necessary to understand, because the underlying logic and recommended algorithms are helpful for TikTok to maintain the number, which can let you understand which data should be paid more attention to to do a good job in TikTok! 1.
The underlying logic of TikTok TikTok and Tiktok are the flagship products of ByteDance.
Both are very similar in terms of client interface and back-end logic.
The core mechanisms of TikTok and Tiktok are: (1) attract users to the platform first through high-quality content;
(2) then match the two to appropriate users through continuous content updates and high-quality recommendation algorithms, so that users can become "addicted".
The core keys of staying on the platform for a long time are [continuously updated high-quality content] and [three-dimensional recommendation algorithm].
2.
For the content platform or e-commerce platform, the most important thing for TikTok's three core algorithms is to achieve a balance and consideration, which is to [make high-quality content more exposed].
However, traffic cannot be monopolized by some works, and new works should also be more exposed.
So there is both the logic of horse racing and the decentralized mechanism of "thousands of people and thousands of faces".
(1) Decentralization algorithm Decentralization algorithm is actually to avoid traffic monopoly.
For big V, no matter how many fans you have, your works will not be seen by the full number of fans.
Unless fans enter your home page to watch, according to normal logic, about 10% of fan users will see your works in the recommended browsing.
This part of the magnitude can guarantee the lower limit data of video works, But whether we can make a further breakthrough depends on the data performance.
So we often see that some videos of big V have very low playback volume, some are very high, and the very low ones basically have a basic value, which is actually the role of this algorithm! (2) Popular algorithms are used to make recommendations based on user tags or attributes.
For example, new users who have just registered will definitely recommend more highly liked content to you.
The purpose of this stage is to retain new users, so you need to use the content that most people like to make you "addicted", which naturally belongs to [highly liked content].
For another example, if the user's tag is a fishing enthusiast, it will naturally recommend outdoor related content to you, as well as other fishing enthusiasts' common interests.
So here is flexible matching based on user tags, because if the recommended content is too monotonous, even if it is users' favorite, it is likely to end up tired, so it is necessary to constantly dig out the content that users are more interested in, different dimensions, and continue to push it to you.
(3) The core of traffic pool racing algorithm is to perform better and give more traffic, which is also the logic of TikTok's traffic pool.
For TikTok, it is distributed according to the content, and each distribution will set a flow pool for the work, and then enlarge your flow pool level by level according to the data feedback, that is, the more the work is displayed