Today, I mainly share, the underlying logic of short videos and the principle of algorithm recommendation.

First of all, since we have chosen short videos to sell, we need to understand the recommendation system of TikTok short videos, which mainly includes three parts: user portrait, content portrait, and between users and content. Matching, video recommendation sorting. User portrait: After the account is registered, the system will collect data sets based on the basic attributes of the user, such as gender, age, hobbies, etc., and preliminarily define relevant tags for the account. Many people do not understand what is the difference between the age options when registering an account.

User portrait:

After the account is registered, the system will collect data sets based on the basic attributes of the user, such as gender, age, and hobbies, and preliminarily define relevant tags for the account. Many people do not understand the difference between the age options when registering an account. If you have registered an account aged 18+ and an account under the age of 18, you will find that accounts under the age of 18 have many functional limitations. As an account selling products, we need short video tags, not user viewing tags. Only short video tags will allow our videos to be accurately distributed to those marked with such user tags.

Content image:

The system will analyze the characteristics of the videos we send based on the hierarchical classification, keywords, entity words, etc., and label various types of content. According to TikTok’s official reviewers, the manual marking of their monthly assessment should be at 90%. The correct rate, that is, if you happen to encounter the 10%, it will become metaphysics. When reviewing the content, it will look at the quality of the video content, which is divided into useless, weakly useful, and strong and useful. For example, if you send a food tutorial, maybe you have text on it, but you don’t say it, he will think you are useless and have text. But if it is not detailed enough, it is weak and useful. Each step is very detailed and how many grams of seasoning are added, so that this content is considered to be strong and useful.

User matches content:

With user tags and content tags thick, the system will match the user’s favorite content in the content pool according to the user portrait and content portrait and then display it.

Sort:

The system has to face hundreds of millions of users and content, and at the same time, it also needs to consider that users’ likes will continue to change. In order to make the selected content more close to what the user wants and more in line with the user’s likes, the system will sort the content. The system will sort according to your completion rate, likes, comments, reposts, and followings, and videos of the same type will be recommended according to their rankings. This is very important!

Short video weight 100% = user portrait 5% + content portrait 25% + user and content matching 10% + sorting 60%

For short videos, Tiktok has a double review mechanism, which is divided into machine review and manual review.

Machine Audit:

Through artificial intelligence technology detection, video images and keywords are identified. Check whether the title is illegal, and prompt manual attention if it is suspected that there is machine interception. Extract the pictures in the video, match with the existing works in the large database to eliminate duplicates, and make low traffic recommendations or weight reduction recommendations for repetitive content.

Manual review:

The suspected illegal works are screened out by the machine review and reviewed again. Manually conduct detailed review one by one, and if it is determined that the violation is violated, it will be punished according to the offending account, such as video deletion, notice of rights reduction, and account ban.

Mainly focus on reviewing video titles, cover screenshots, and video keyframes. Why do video works never pass the review? All video works published on the site have to go through a double review process. If the video work fails to be reviewed by the machine, it will be intercepted by the machine and entered into manual review, and then due to human problems and the continuous change of the amount of submissions, there will be a delay in review and a backlog of review. When a video work is under review, if you do not receive a violation notice, it does not mean that the video has been restricted or there is a problem with the video content. You can just wait for the review. After the video work passes the review normally, it will not affect the subsequent data traffic, and the key is the content quality.

The platform recommendation mechanism is that after the video has been reviewed, it will be recommended by the system. The recommendation is decentralized, traffic pool planning, and artificial intelligence distribution system. The platform will provide a traffic pool for each work, no matter if you are large or small. As long as the video work has passed the review, it will be recommended. Can the video become popular in the later stage?

It all depends on the data performance of this traffic pool (views, likes, comments, and forwarding).

The recommendation is divided into three stages:

Basic recommendation:

For newly released video works, the traffic distribution is mainly based on the vicinity and attention, and then intelligently distributed to a small number of users with user tags and content tags, depending on the user’s feedback, if the user likes a larger proportion, it may enter the second place. Wheel stacking is recommended.

Overlay recommendation:

The newly released videos will go through the previous round of intelligent distribution, and distribute about 100-500 views according to the account weight. If the forwarding volume reaches 10 (for example), the algorithm will judge the content as popular, and automatically weight the content and superimpose it. Recommend to you 1000, the forwarding volume reaches 100 (for example), the algorithm continues to superimpose the recommendation

When it reaches 10,000, the forwarding volume reaches 1,000 (for example), and it is recommended to stack up to 10w, and push it in sequence. When the forwarding volume reaches a certain level, a mechanism that combines big data algorithms and manual operations is used.

time effect:

The popularity weight will be selected according to the time, and the popularity of a popular video will last for at most 1 week, unless a large number of users imitate and follow. Use your music on beat or in tune, etc., so you also need a stable content update mechanism, and the ability to continuously output hits. It is emphasized here: as a short video creator, we are not going to study the core algorithm mechanism of tiktok in depth, but to understand the logic inside. Targeted content optimization, guide users to watch our videos, and then like, comment, and forward behaviors, so as to get higher recommendation.

How does the TikTok platform judge the quality of a video? When a new video work is released, the first round of basic recommendations will assign you 300-500 online users according to your account level, and judge your video quality based on their data feedback. What data feeds back the quality of the video? Four core indicators: like rate, comment rate, completion rate, and forwarding rate.

The TikTok recommendation mechanism is:

An account that logs in more than 10-20 times a month and publishes 10 videos is a user that the platform must find a way to retain; an account that logs in 5-10 times a month and publishes 0-10 videos is the core conversion user, the long-term development of the platform requires such users to maintain, the main group of people receiving advertisements; accounts that log in 0-20 times a month and publish videos 0-5 times are content providers with unstable traffic; log in 0-10 times a month, for those who don’t post short videos, the platform will mainly convey the value of the product and strive to cultivate user habits. This is why you rarely get ads when you sign up for a new account.

Based on this, in order to improve the account level, each account must log in to the platform at least 20 times a month and publish 10 pieces of content. Based on this, it can be inferred that when the account is stable, if a video is played 500 times, the more videos are posted, the wider the traffic and the higher the conversion rate.