B station marketing logic

Brands may have concerns about doing content marketing on Zhihu because the data cannot be seen immediately and it is difficult to establish a clear relationship with sales.

But when doing content marketing at Bilibili, we don’t need to ask why, we just need to take action.

Bilibili’s data feedback is very rich, and the effect of the campaign will be determined mainly in the first week. Convenient to report to the boss. And it’s easier to build connections with sales. More certain than Zhihu and Xiaohongshu. On the one hand, Bilibili’s content marketing can directly lead to sales conversions, and on the other hand, it can also be built into a place for potential buyers to refer to opinions.

01 Where does B Station’s traffic come from?

The main traffic is the first page you see when you open the APP: Homepage-Recommendation page .

The live broadcast page also has some traffic.

Search will also bring some traffic. The content of Bilibili and Zhihu may appear in Baidu search results, but Xiaohongshu will not.

02 The core logic of B station's recommended content

In the past two years, many people have interpreted the core algorithm that is suspected to be recommended by B Station.

The recommendation index of a video = coins × 0.4 + collections × 0.3 + barrage × 0.4 + comments × 0.4 + plays × 0.25 + likes × 0.4 + shares × 0.6 . The recommendation index of a newly released video within one day increases by 1.5 times.

I saw that netizens verified this recommendation index, but the amount of data was very small, so I also verified it.

(The following data was collected on August 18, 2021)

First, I crawled the data of the default 160 popular videos at the top of the technology section of Bilibili. A scatter plot is drawn, with the horizontal axis representing the number of views and the vertical axis representing the total number of recommendations.

Among all the following data, the comment data is dynamically loaded by Ajax, which is a bit troublesome, so it is not crawled. My recommendation index does not include review factors. This omission did not significantly affect the results.

The relationship of the scatter plot is quite beautiful, strictly near a diagonal line, which shows that the relationship between the number of views of popular videos and the recommendation index is very clear.

The data of popular videos is the last stable data of the video. So does this rule also apply to newer videos as data grows?

However, most of the data in the above figure has a reading volume below 5,000, which is not clear enough, so I removed the data with a reading volume above 5,000.

Such data still does not reassure me, because in the recommendation index, the slope of the relationship between the number of views and the recommendation index alone is 0.25, and the weight of these interactive actions only accounts for 20%. So I calculated the data from the recommendation index excluding the playback volume factor. The following figure is obtained. When the recommendation index does not include the playback volume factor, the relationship between the recommendation index and the playback volume is not a straight line.

The following conclusions were drawn from the above data study:

The playback volume of videos on Bilibili is strictly related to the recommendation index: playback volume = 10/3 recommendation index .

Recommendation index = coins × 0.4 + collections × 0.3 + barrage × 0.4 + comments × 0.4 + plays × 0.25 + likes × 0.4 + shares × 0.6.

The recommendation index treats new and old videos equally, and does not provide 1.5 times the playback volume for new videos .

There is no streaming test of new videos. This is different from platforms such as Zhihu, Xiaohongshu, and Douyin. Other platforms will first give 500-1000 exposures to test the data, but Bilibili does not.

The number of views has nothing to do with the completion rate. Your video will not get a higher recommendation weight just because you have a high completion rate.

The number of views also has nothing to do with the open rate. Your video will not get a higher recommendation weight just because it has a high click-through rate in the information flow.

This brings up a problem. The playback statistics of Bilibili do not count the playback just by clicking on the video.

The playback data of Bilibili requires that if a video is watched multiple times by the same account within a period of time, only one playback will be counted. In addition, the viewing time of a video must reach a certain ratio. If the viewing time is too low, the playback data will not be calculated.

This calculation rule is almost as strict as the reading volume calculation rule of WeChat public accounts. It can be understood as the number of views of the video on Bilibili, that is, how many people actually watched the video.

Judging from the data in the above figure, the effective playback ratio of this calculated playback volume is less than 17%, and is expected to be between 10%-17%.

Why doesn’t Bilibili consider the completion rate when recommending videos? Because if the completion rate significantly affects the video's playback volume, short videos will have a clear advantage, and the traffic obtained by medium and long videos will be significantly reduced, making users no longer willing to produce medium and long videos. Tik Tok’s recommendation algorithm takes the completion rate into consideration, which makes bloggers reluctant to produce long videos.

Why doesn’t Bilibili conduct some streaming tests on new videos? The recommendation algorithms of other platforms test the streaming of new content based on data such as open rate and completion rate. The open rate is based on the header image and title. For medium and long videos, these are not the signs of whether the video is popular, interaction is.

Bilibili does not have a streaming test, which means that new videos released by new users may have zero views if there is no initial interaction. Unless the video is pushed up by a user who watches it through search.

Videos from UP hosts with a certain number of fans are not easily buried, but high-quality videos from UP hosts with a low number of fans may be buried. In order to prevent their videos from being buried, UP hosts with a low number of fans need to attract people to create popularity for their videos in the early stages.

How does Bilibili’s recommendation index algorithm work?

During the cold start phase of a video, the recommendation index prompts UP to bring initial traffic to the new video and gain popularity.

When the video's interactive data is good, and the number of views is <10/3×recommendation index, it will be recommended, and the number of views will increase until the number of views = 10/3×recommendation index.

When the video playback volume is too high, that is, the playback volume > 10/3× recommendation index, it will not be recommended until the interaction data catches up.

Therefore, the strategy of increasing the number of views does not work on Bilibili. It will only reduce the actual traffic of the video.

For Bilibili, the number of fans is important, and fans with high stickiness are important. This recommendation algorithm also explains why the commercial value of big UP hosts on Bilibili is higher than that of Xiaohongshu and Zhihu.

To summarize :

The core logic of B station's recommended content is to follow the recommendation index = coins × 0.4 + collections × 0.3 + barrage × 0.4 + comments × 0.4 + plays × 0.25 + likes × 0.4 + shares × 0.6. Number of views = 10/3 × recommendation index.

03 How does Bilibili push content to users?

The way B station recommends content is similar to Xiaohongshu and Zhihu. The recommendations mainly come from the following perspectives:

Content related to the searched keywords

Content posted by the UP masters you follow

Content related to content you have viewed and interacted with

Trending Content

Channels you follow and content related to channels you browse

04 How to implement content marketing strategy for Bilibili?

Suppose we want to do content marketing on Bilibili for a new product with an estimated annual sales of 100 million yuan and a price of 200 yuan.

The value of establishing an official brand account on Bilibili is greater than that of Xiaohongshu and even greater than that of Zhihu.

When brands advertise on Bilibili, they mainly look for UP hosts with a certain number of fans to make product promotion videos, guiding them to e-commerce platforms to generate purchases. This has been the normal pattern in the past.

Our content marketing strategy is to establish a virtuous cycle of content marketing on Bilibili.

Let each delivery action empower subsequent actions, rather than each delivery being independent.

Continuous action :

Through the official account and corresponding communities, a highly sticky fan group based on Bilibili is established to continuously cold-start the brand's content .

Anchor one or more keywords, and focus content delivery on this keyword , while also drawing traffic from other keywords. This way, users who search for this keyword or pay attention to this keyword will be more likely to pay attention to the brand's content.

05 How to evaluate the content marketing data of Bilibili?

The data of Bilibili first focuses on the number of views.

The number of fans or community users in a circle is also a very valuable indicator.

You can refer to the average playback volume corresponding to the average advertising price of the UP master on Bilibili, and calculate the average playback price = average playback volume/average price. As the break-even point for investment.

06 Summary

The traffic of Bilibili is mainly concentrated on the recommendation page, which is estimated to account for more than 90%. The live broadcast page also has a certain amount of traffic, which is estimated to account for less than 5%. The dynamic page has less traffic, which is expected to be less than 1%. Therefore, we must pay attention to the maintenance of fan stickiness in order to enhance the value of fans, otherwise the fan value will be low.

The playback volume of videos on Station B follows the following formula:

Number of views = 10/3×(coins×0.4+collections×0.3+barrage×0.4+comments×0.4+views×0.25+likes×0.4+shares×0.6).

In order to ensure that the video passes the cold start phase, efforts should be made to improve interactive data in the early stages. You can consider building a fan community to enhance interaction. Pay attention to gathering fans through official accounts and enhancing stickiness through activities and other means.