Details of the twitter flow push mechanism: the whole process from content to recommendation

Today I would like to share with you the main streaming mechanism of the dithering platform

The streaming logic of dithering is a centralized traffic distribution mechanism The dithering platform uses multi-level flow pool for hierarchical recommendation, and pays more attention to the top products and hot spots. First, match the user tags with the content that such users like. He may first send to the primary flow pool of 200 to 300 users. The probability of fans seeing new videos is not high, which is about 10% According to the user's feedback effects such as likes, comments, and forwarding, determine whether your content is of high quality.

If the score is high, it will make a second recommendation, a week's recommendation to a larger flow pool at a higher level, and the recommendation will be superimposed to form popular content. The user feedback will determine whether the content will be recommended to the next larger flow pool, and the creators of high-quality works will get a lot of resource tilt, faster access to fans, and the dithering will be filtered through multi-level flow pools, After that, a large number of head works will be presented, covering the hot spots of the current blockbusters, which is a basic logic of dithering.

Dithering algorithm

The dithering deduction can be divided into two parts: push flow and pull flow. Push streaming refers to the communication process in which the content creator uploads the short video created by himself to the dithering server for display by other users or platforms. Pull streaming refers to the communication process in which the dithering server pushes the data to the user's equipment to play when the user watches the video. It includes the following key links and technical elements:

1. Upload traffic management

With the increase of the number of users and the diversification of the network environment, the dithering server must manage and optimize the upstream traffic, and at the same time avoid affecting the user's experience due to upload failures.

2. Quality control and safety protection

In order to ensure that users upload high-quality characteristic videos, Treble has implemented a series of content quality control and audit mechanisms, including manual audit, machine audit and other means. When users upload videos, the platform will automatically detect and review the uploaded videos to ensure that the videos comply with regulations and platform regulations.

3. Content distribution and optimization

The powerful content distribution and optimization capability of dithering is an important part of its streaming mechanism. It distributes different content to interested users and optimizes it by using the exposure of the flow pool.

4. Personalized recommendation algorithm

The personalized recommendation algorithm of dithering is another important part of its streaming mechanism. Its purpose is to intelligently match the content that users are interested in with the recommended short videos. The personalized recommendation algorithm for dithering recommends videos that may be of interest to users by analyzing their interest and behavior data, and dynamically adjusts the recommended content according to user feedback. In this way, valuable content can be pushed to users to improve user satisfaction and convenience.

Streaming in the live broadcast room

The mechanism of dithering and streaming must be based on the competition between live broadcasting rooms. Who is the corresponding competitor of your live broadcasting room?

In order to ensure the fairness of traffic distribution and make it possible for each live broadcast room, the dithering method is Peer competitive traffic pool Principle of. In other words, your competitors are at the same level as you, As long as you can stand out in the peer competition, you can get more traffic recommendations.

Streaming in the live broadcast room is similar to short video streaming. It is recommended to users who may like this kind of live broadcast room according to the tags, characteristics and attributes of the account and the live broadcast room. Users can click the live broadcast room to watch the live broadcast. The process of streaming is shown as follows:

In the process of streaming, the dithering system needs to conduct data collection, data summary and data comparison in the live broadcast room, and then judge whether to continue streaming.

1. System acquisition data
2. Streaming in the live broadcast room is mainly about the data in the live broadcast room. Since the live broadcast, the system has been collecting the data of various clicks, stops, interactions, and transformations in your live broadcast room;
3. System summary data
4. The system will summarize and analyze the data collected from all the live broadcast rooms in the same time period;
5. Live data comparison
6. Compare with accounts in the same industry, at the same level and in the same time period, and conduct real-time ranking;
7. Continue streaming
8. After the system analyzes the data, if the data in your live broadcast room is good and is greater than the threshold of streaming, you will continue to stream.

In general, that is Each account will distribute a certain amount of traffic at the beginning of the broadcast phase to compete with the live broadcast rooms at the same level. The data in the live broadcast room during this period will be evaluated. When the data reaches a certain target, the streaming will continue. If not, the streaming will stop. Until the end of the live broadcast, the whole live broadcast data will affect the traffic of the next live broadcast.

In addition to competing with the live broadcast rooms at the same level, we also have to compete with ourselves. Because the current streaming of dithering does not distribute traffic according to the accumulated data of the whole field, but The flow rate is driven in such time units as 5 minutes, 30 minutes and 3 hours, such as gear.

(Picture source: Feigua Zhitou)

The assessment data of streaming in the live broadcast room can be divided into two categories: popularity data and transaction data.

Among them, live broadcast room Only the popularity data of the live broadcast room will be assessed for Xiaoxuehua's goods As long as the popularity is high, the dithering will continue to push the stream;

But use The live broadcast room of Xiaohuang Che not only needs to assess popularity data, but also the transaction data. If there is no transaction, there will be no streaming.

No traffic, but the data is not good enough.

If you want the live broadcast room to get a continuous stream, you need to carefully do the above data. Only when the data is done well, and the traffic is also available, can the transaction be generated!

Operation article of last year and today

1. 2022:   PDCA │ A management tool that benefits people for a lifetime. ppt (0)
2. 2022:   Are you looking for a job in the right way? [Amway 9 Treasures APP] (0)
3. 2022:   15 necessary toolkits for operators (2022 Collection). xls (0)
4. 2022:   Explain the Productive Thinking Model (0)
5. 2021:   Ultimate Product Zhou Hongyi Reading Notes (0)