This article will help you understand Zhihu algorithm!

We all know that self-media people attract user attention and monetize traffic by outputting content on the platform. Therefore, many people believe that the platform is a bridge of communication between self-media people and users.

This view is actually correct, but to be more precise, self-media people are never communicating with users, but rather playing games with platforms.

Why talk about game?

Because any platform has its own fixed mechanism algorithm, what self-media people need to do is to constantly make their content closer to the platform algorithm and constantly trigger the platform's recommendation mechanism.

Let me take Zhihu as an example. Many people should have heard of Zhihu’s “Wilson Formula”, but they don’t quite understand it, nor do they know how this formula is calculated.

Here, I will give you a detailed explanation, hoping it will be helpful to some friends who are engaged in Zhihu operations .

First of all, we need to understand that if we compare operating Zhihu to a monster-killing game, then the algorithm is a detailed strategy manual that can tell us what the monster's weaknesses are and what to do next. This strategy can help us have a clear idea of ​​what to do in the process of killing monsters and upgrading.

Simply put, the concrete manifestation of Zhihu's algorithm is the personal content search ranking.

The larger the value obtained by the operator through the algorithm, the higher the search ranking will be, the higher the exposure will be, and the more corresponding feedback will be obtained. The essence of our operation of Zhihu is to make this value as large as possible.

Secondly, before discussing the specific formula, we need to think about a question first. If you were the operator of the Zhihu platform, which users would you tend to retain?

The answer is simple. Anyone who can play a positive role in the construction of the platform will naturally be rewarded by the platform.

This is one of the basic requirements for a platform to develop in the long run and does not need to be proven by data.

For users, outputting professional knowledge, screening high-quality content, enhancing community activity, increasing Zhihu's influence, maintaining platform order, etc. are all feasible platform construction operations.

Therefore, as we continue to generate these behaviors, we are constantly empowering Zhihu, and we should naturally get higher algorithm values.

After understanding this concept, let’s analyze in detail the official algorithm mechanism of Zhihu, the various values ​​mentioned in the algorithm, as well as its guiding significance and guidelines for closing the pit.

Here is a formula, namely the Wilson formula:

Where u is the weighted number of votes in favor, v is the weighted number of votes against, and za is a parameter.

The figure below can more intuitively show several important characteristics of Wilson's formula.

For the sake of convenience, the axes corresponding to up-vote, down-vote, and score in the left figure are called x, y, and z axes respectively, and the right figure is the contour map of the left figure.

The overall curved shape of the left picture is consistent with the commonly understood correspondence between upvotes, downvotes, and answer quality. This is the algorithm mechanism officially recognized by Zhihu.

Weighted votes refer to the value assigned to the content by other people's likes, but it should be noted that the impact of each person's like is different, which depends on the weight of the liker in the current field.

Many people carefully write an article and then post it on Zhihu Answers, thinking that they will then get thousands or tens of thousands of likes and become Zhihu big Vs. However, after a while, they find that their answers have very few likes, and there is not even a single opposing comment.

At this time, they began to feel disappointed, thinking that they did not have the so-called self-media talent and were even not suitable for Zhihu operations.

This is actually a misunderstanding. Insufficient feedback on your answer does not mean that there is a problem with your content, nor does it mean that this answer will remain unanswered.

In fact, based on long-term operational experience, a valuable article, even if it does not have much feedback at the beginning, will suddenly explode in some period of time in the future. The reason why it does not have a certain degree of influence now is because there are still some problems with your account.

For example, if you wrote an answer and a big V with 100,000 followers and a small V with a few hundred followers liked your answer, the difference in impact will be huge. This is the concept of "weighting".

Similarly, opposition is also weighted. The higher the weight of the opponent, the greater the drop in our score.

This formula looks complicated, and it is normal that you may not understand it at first. I will simplify it and you can use another formula to determine your own weight.

That is: s=Like*Collection*Like*Initial Weight*Comment*Dislike.

Among these influencing factors, except for the field weight, the other factors are realized through the interaction between users. In other words, how high an article ranks is ultimately determined by other users.

In addition, likes, favorites and likes will increase the score of the answer, while dislikes will reduce the corresponding score.

At this point, someone may say, if you had said this earlier, I would have understood it. What is the purpose of using this Wilson formula?

It is really useful, because in addition to expressing the corresponding relationship between various factors, the Wilson formula can also concisely express the development process after the factors change.

The specific manifestations are:

• Fixed number of negative votes, the more positive votes, the higher the score;
• Fixed approval votes, the more disapproval votes, the lower the score;
• The ratio of approval and disapproval is fixed, and the higher the total number of votes, the higher the score;

This should be easy to understand, so I won’t explain it in detail.

When the total number of votes is small, if an answer gets votes, the score will increase quickly. The larger the total number of votes, the slower the increase. What does this mean?

In the process of operating Zhihu, you will often find that some answers with more than a dozen likes are ranked very high, while the answers following them may have thousands or tens of thousands of likes.

That's what this rule means.

For newly created content, Zhihu's system will give the content a weighted basic score based on the weight of the creator's current field so that it can get greater exposure.

This is very reasonable, because compared to earlier answers, if a new answer does not have a certain initial exposure, it will not even be qualified to be judged.

On the premise that the system grants basic exposure, if it can get the recognition of the first batch of users, the system will determine that this is a potential content and will push it to more people, leading to an explosion in a short period of time.

During this period, although its number of likes was not as high as other answers, its like rate was very high, which enabled it to quickly improve its ranking and stabilize in a leading position for a period of time.

For an answer with a high number of approval votes, when it starts to receive disapproval votes, its score will drop rapidly. The greater the total number of disapproval votes, the slower the rate of decline. This is somewhat similar to the second principle.

How should we understand this point?

Just imagine that in our real society, as long as a person is popular enough, there will definitely be controversy, because everyone has different ideas and there will always be different voices. But these controversies do not mean that he is a bad person, he is just popular enough.

Therefore, when the number of approvals for a piece of content is high enough to allow it to be seen by more people, someone will definitely raise objections. At this stage, the score of the content will drop rapidly until it forms a stable proportional relationship with the approvals.

However, it should be noted that only when we start to receive opposition will there be a significant change in the score. In the subsequent development process, the more oppositions we receive, the smaller the impact of each opposition will be.

Let’s go back to the second formula, which is: s = Likes * Favorites * Likes * Initial Weight * Comments * Dislikes

From this formula and long-term operating experience, we can get the following six important principles:

1. All users see the same ranking;

2. When other conditions remain unchanged, approval will increase the ranking of the answer, while disapproval will decrease it;

3. The influence of the content created by users in a certain field will increase the weight of the user in this field. In other words, the initial weight can affect the content score, and the final content score will in turn enhance the weight of our current field;

4. The votes of high-weight users in the field have a greater impact on the ranking. This vote includes likes and dislikes. Of course, when high-weight users answer related questions themselves, their answers will be ranked higher at the beginning due to the blessing of the initial weight;

5. When voting or answering questions anonymously, the user's weight will not be calculated;

6. Just because a piece of content does not currently receive good feedback does not mean it will not have the potential to become highly praised in the future;

According to the Wilson formula image, it can be inferred that a good content will eventually be recognized, but the time may vary. It may be affected by the initial weight or unstable factors at the beginning, so that it does not get more exposure, but there will always be a time point when it is discovered and recognized. This is also one of the regulatory functions of Zhihu's distribution mechanism.

Therefore, content is very important, and the saying “content is king” is by no means groundless.

After fully understanding Zhihu's algorithm, what important information can we obtain from it?

In other words, how should we make the most of these official rules?

To summarize briefly, there are two points:

1. After understanding Zhihu's algorithm mechanism, before creating a new answer, we must not answer questions at will, but first think about what our vertical field is, and then look for questions selectively.

2. In the process of creation, we must learn to add some interesting and easy-to-complain points on the original basis. This is what I often say, self-media people must constantly touch the users' "pain points" and "pleasure points" in order to actively interact with more users.

Of course, there are also some proactive behaviors, such as directly reminding readers to like the article at the end of the article, or even inducing readers to like the article. These are common ways to increase interactive behaviors.

I have told some friends before that there are three realms for self-media people:

The first is to express oneself, which is a common mistake made by many junior self-media people. They regard Zhihu as a platform to express themselves, or even an emotional tree hole.

This is a wrong way of operating. There is nothing wrong with expressing yourself, but if you are thinking about exposure, traffic, and subsequent monetization, you should not express yourself blindly.

The second is to serve users. Many self-media people have already experienced this. When writing self-media articles, they are not writing for themselves, but for users. Self-media people act more as a medium, better expressing and listing the opinions that users want to see and want to say.

The third is to trigger the platform rules. Why should self-media people write articles for users?

In essence, of course, it is not to satisfy users, but to trigger the platform rules through user satisfaction, so that one's answers can get higher exposure and enter a larger traffic pool.

This illustrates very well that self-media has never been a battle between creators and readers, but a mutual game between creators and platforms.

Author: User 637299

Source: User 637299