Product Analysis Report | Bilibili, from a two-dimensional community to a comprehensive video community

As a bullet screen video website in the early days, Bilibili had a strict registration system, and its users were mainly ACGN culture enthusiasts. With the development of Bilibili itself and the gradually lowered entry threshold, more and more users are pouring in.

Today, Bilibili has become a pan-two-dimensional cultural and entertainment community with young people as the main users. From the perspective of product positioning, it has two main aspects: one is a platform for two-dimensional fans to watch and communicate; the other is a comprehensive barrage video community with UGC content as the main body.

The author will analyze Bilibili from two dimensions: the second dimension and the comprehensive video community, and put forward corresponding optimization suggestions.

This article consists of the following aspects:

1. Product Architecture
2. Market Analysis
3. User Analysis
4. Functional analysis
5. Summarize

Experience environment:

• APP version: 5.37.1
• Experience device: iPhone 8
• Mobile phone system: IOS12.1

1. Product Architecture

The main functional modules of Bilibili can be divided into video, live broadcast, community and commercialization.

The functional architecture diagram is as follows:

2. Market Analysis

1. Product Positioning

Bilibili was originally positioned as a two-dimensional vertical community. In the ACGN cultural circle, "two-dimensional" is used to refer to the virtual world.

The main carriers of two-dimensional culture include: animation, comics, games, novels, virtual idols, special effects films, some movies, some TV series, as well as their derivative fan creations and peripheral products.

Today, Bilibili is a comprehensive video barrage community and two-dimensional community targeting fans of general two-dimensional culture.

1. Market Space

Bilibili's target users are the pan-two-dimensional group, including: core two-dimensional people and other pan-two-dimensional people.

The core 2D population has a strong demand for anime and is willing to spend on 2D peripheral products. The general 2D crowd has no obsession with anime and otaku culture, but at least they have watched cartoons on TV, read comics, or played video games in their childhood, and have some understanding of current popular anime. They are influenced by the two-dimensional culture and have a temperament that matches the B station community.

According to Aurora Big Data, the majority of pan-two-dimensional users are young people, with those aged 25 and below accounting for about 65% and those aged 35 and below accounting for about 86%.

Data source: Aurora

In terms of user scale, the number of pan-two-dimensional users will reach 350 million in 2018, of which 110 million will be core users.

Data source: Sutu Research Institute

According to the sixth national census of the People's Republic of China: China currently has a population of approximately 500 million aged 10-35, and the penetration rate of smartphones among people in this age group is extremely high, reaching more than 90%.

Given the rapid growth trend of pan-two-dimensional users, which can reach at least 20% growth rate each year, people aged 10-35 who use smartphones are potential pan-two-dimensional enthusiasts - that is, Bilibili's potential user group, and the number of this group is 450 million.

1. Commercialization

Bilibili’s promise not to include advertisements in the opening credits resulted in insufficient advertising revenue for the company, and in the early days it was highly dependent on revenue from game joint operations.

In 2017, game joint operation revenue accounted for 83% of total revenue.

In 2018, the year-on-year growth rate of Bilibili's revenue in the fields of advertising, live broadcasting, and membership value-added services expanded, while the year-on-year growth rate of game joint operation revenue declined, and its proportion gradually decreased.

According to Bilibili’s 2018 financial report: In fiscal year 2018, Bilibili’s total revenue reached 4.13 billion yuan, a year-on-year increase of 67%, of which the revenue from game joint operations accounted for 70%, a decrease from 83% in 2017.

Source: Futu NiuNiu

The target group of mobile games is the core users of the two-dimensional world.

Most of Bilibili's mobile game revenue comes from the two mobile games Fate/Grand Order and Azur Lane. Due to their early release time, they have a high penetration rate among core two-dimensional users and have little room for future development.

In addition, the proportion of core two-dimensional users in Bilibili is limited, and most of the new users are non-core users. It can be predicted that Bilibili's dependence on mobile games in terms of future revenue will gradually decrease, and it will do more exploration in advertising, live broadcasting, value-added services, e-commerce and other aspects.

Here are a few examples:

• In terms of advertising, Bilibili has currently increased the proportion of information flow advertising, and has also added advertisements recommended by UP hosts below the videos.
• In terms of live streaming, Bilibili recently introduced a well-known virtual anchor from Japan to build IP and convert more two-dimensional users.
• In terms of e-commerce, Bilibili has cooperated with Taobao to support Bilibili's contracted UP hosts in establishing certified Taobao expert accounts, helping UP hosts increase their creative income.

1. Competitive product comparison and analysis

The core of Bilibili is still the two-dimensional community, so its competitors are other two-dimensional community apps.

In June 2018, based on monthly active users (the number of users who subjectively opened the app at least once in a month), the top four two-dimensional communities were: Bilibili, First Bullet, Fuciyuan and AcFun.

Looking at the following data, we can find that Bilibili has an absolute advantage in most scale indicators, daily average indicators, and per capita indicators.

Since Bilibili is in a leading position in the field of the two-dimensional community, from the perspective of making suggestions, the author does not do much analysis on the advantageous data, and therefore mainly selects a few data points where Bilibili is not as good as other competitors for analysis.

Data source: Analysys Qianfan

4.1 Comparison between B Station and First Bullet

In terms of next-month retention rate, Bilibili ranked second with 55.7%, and the first batch ranked first with 69.2%.

Compared to Bilibili, Yidan is a more vertical two-dimensional community, while Bilibili also has a large number of three-dimensional videos in addition to the two-dimensional community, such as: life, technology, ghost animals, etc.

As a purer 2D community, Diyidan has more core users who truly love 2D, thus having stronger user stickiness.

It can also be seen from the download data that the daily download volume of the first wave is on a downward trend overall. It can be inferred that this APP does not have any advertisements, so the increase in users mainly depends on recommendations from two-dimensional enthusiasts. Most of the new users are target users and are more likely to stay in the next month. As a more well-known product, Bilibili has more non-target users who try it out of curiosity, so its retention rate in the next month is relatively low.

Source: Qimai Data

In terms of the product itself, the first wave has some subdivided sections that are not available on Bilibili, such as the pure dating section "CP Dating".

The first batch of "CP Friendship" section

There are also some sections whose buttons are more prominent in the first wave, but the same sections are relatively niche in Bilibili and the buttons are deeper and unknown to users, such as: dubbing, American comics, etc.

The first part of the "Dubbing/Voice Control" section

The first batch of "American comics" section

The CP dating section has over 2.63 million members and 40 million posts, and the dubbing/voice control section has over 1.76 million members and 20 million posts.

It can be said that the interest circles similar to Tieba adopted by the first bullet can indeed gather users. This is because each interest circle has a clear entrance - it can be found through the first-level button or search.

Within the interest circle, the way content connects users is not only "popular" but also "latest", so you can see a lot of content posted by ordinary users that no one likes or comments on. The high exposure rate of ordinary users helps to form small communities (refer to Baidu Tieba) and create a strong social atmosphere. Small communities with stronger social orientation are conducive to user retention.

In fact, Bilibili also had an interest circle. It started testing in December 2015, entered maintenance in December 2017, and has not been open to this day.

The reason for this, in addition to personnel changes, was that most users in the interest circle at the time were not keen on sharing interests, but on sharing resources, such as software, emoticons, movies, etc., many of which were not suitable for children, and there was very little high-quality content.

Bilibili focuses on community atmosphere and hopes to present high-quality UGC. Therefore, interest circles require a lot of supervision and support, otherwise they will become like Tieba and destroy the community atmosphere of Bilibili. In addition, the development of interest circles also faces competition from Baidu Tieba, and users who have been using Tieba for a long time may not be willing to transfer to Bilibili.

For various reasons, Bilibili has given up on interest circles, and its community model is still mainly based on "UP host-audience".

Bilibili is a community oriented towards content quality. Taking videos as an example, videos will undergo strict review before being released. For example, only very excellent ghost videos will pass the review in the ghost area.

Take updates as an example: the updates that users see often come from people they follow and influencers. Similarly, in the comment section of a TV show, users tend to look at comments with a high number of likes.

All these indicate that it is difficult for ordinary users to see each other’s content, the interaction is very limited, and their communication mostly comes from comments on the same popular content. In contrast, Diyidan opened up social interaction among ordinary users, which is one of the reasons why Diyidan has a higher retention rate.

Another reason for the high retention rate of the first batch is: the sign-in system, including sign-in at the entire APP level and sign-in in each section.

In the check-in system at the APP level, checking in can increase experience. The higher the level, the more permissions you have, and you also have cooler names and icons. Checking in within the forum can increase the user's influence and improve the ranking within the forum.

The definition of next month retention rate is: the percentage of newly installed users who still use the app in the next month, where opening the app is counted as usage.

Even if some users do not browse the content in the APP, they will open the APP because of signing in, resulting in a higher first-round retention rate, but the average usage time and average number of times per person are much lower than those of Bilibili. For the same reason, the sign-in system also caused the number of active users of Diyidan to be higher than that of Bilibili on a month-on-month basis, but the number of launches and usage time were lower than those of Bilibili on a month-on-month basis.

4.2 Comparison between Bilibili and Fuciyuan

In terms of indicators related to growth (including month-on-month data on the number of launches, month-on-month data on usage time, month-on-month data on average daily active users, month-on-month data on average daily launches, month-on-month data on average daily usage time, month-on-month data on average number of launches per person, month-on-month data on average usage time per person, etc.), Bilibili lags behind Fuciyuan in all aspects.

In terms of per capita indicators such as the number of launches per person and the average usage time per person, Fuciyuan is close to Bilibili, and has a certain gap with the first batch.

Fuciyuan focuses on BL comics, but Bilibili does not have a specific BL section. BL content mainly occupies a small part in the "Pure Love" section, so it is impossible to form a community of BL enthusiasts, resulting in this part of users being diverted to Fuciyuan. The next month retention rate is the lowest among the four apps, which shows that users have polarized views on Danmei. Users who are not interested will uninstall the app, while users who are interested will spend more time on it.

Due to the specificity and niche nature of its business, Fuciyuan will not pose a fundamental threat to Bilibili.

However, the development of the Fu dimension can explain some problems.

In the field of danmei, Fuciyuan is the leader. Compared with BL novels, BL comics have greater visual impact and are more in line with the reading habits of today's young people - they prefer to look at pictures.

Compared with other comprehensive comic apps that include Danmei sections, the pure Danmei Fudiyuan can better create a community atmosphere. Behind the development of Danmei is the appreciation of female users (including a few men) for handsome men, mainly two-dimensional images, but also including real people whose appearance is close to the two-dimensional ones.

Judging from the prevalence of young and handsome men in the current entertainment industry, the public has a strong demand for appreciating handsome men. With the development of the two-dimensional world and the combination of the two, danmei comics have long-term development space.

However, Bilibili, which focuses on subculture, has missed out on such a subculture field. In addition to regulatory factors, it is also attributed to the fact that with the increase in the number of users, Bilibili has gradually become a pan-two-dimensional community connected to real life, but is lacking in the core field of the two-dimensional world. Therefore, there is still a lot of room for exploration in Bilibili's two-dimensional business.

4.3 Competitive product comparison summary

Source: Analysys Qianfan

As a giant in the two-dimensional community, Bilibili has maintained a fluctuating upward trend in the number of active users since 2014, which has indirectly led to an increase in non-two-dimensional content on Bilibili, and gradually developed into a pan-two-dimensional entertainment community.

By analyzing the competitors, Bilibili can make some new attempts in the following two aspects:

1) There should be more exploration in the core business of the second dimension

As an APP with the two-dimensional world as its core, Bilibili can strengthen the atmosphere of the two-dimensional world when there is no shortage of traffic. It can make more subdivisions in the two-dimensional field, provide more choices for two-dimensional fans, and cultivate more core users. Bilibili should also increase its investment in anime copyrights. Richer resources are the basis for user retention and usage time.

2) Provide more social channels between ordinary users

Although most Bilibili users are just viewers and not content producers, this group also has a strong desire to communicate. Currently, the only methods that Bilibili can provide are barrage and comment areas, which mainly provide one-way input from a single user to the entire user group. Some users will not post content because they cannot get a reply. Therefore, Bilibili can increase the ways of interaction between a small number of users within a small range.

3. User Analysis

1. User Profile

According to the independent application-side data from Aurora Big Data, in terms of gender, there are slightly more male users than female users on Bilibili.

In terms of age distribution, users aged 20-24 account for the largest proportion, and users under 30 years old account for 92%. Bilibili's user base is very young, and the trend is towards becoming even younger.

According to Chen Rui, chairman and CEO of Bilibili, the average age of platform users is 21 years old, and the average age of new users in 2018 is 19.8 years old.

From the age distribution of users, it can be inferred that most users are students or newcomers to the workplace.

Source: Aurora

In terms of the geographical distribution of users, from the perspective of provinces: users mainly come from the developed eastern coastal areas and some populous provinces in the central and western regions. These regions have developed economies, high smartphone penetration rates, material needs in life, and people have more pursuits for spiritual culture. In addition, developed regions are more tolerant of non-mainstream cultures, which is conducive to the development of two-dimensional culture.

Source: iResearch

From the perspective of cities: the 10 cities with the largest number of users are all first-tier or second-tier cities.

It is worth noting that users from third-tier and lower cities account for nearly half of the total, which shows that Bilibili and the two-dimensional culture are not only popular in big cities, but also in urban and rural areas.

Source: Aurora

1. User usage scenarios

User Type 1:

Xiao A, male, 20 years old, lives in Guangzhou, is a college student, and is a registered member of Bilibili (informal member).

I am not a fan of the 2D world, and my exposure to the 2D world is limited to the cartoons I watched when I was a child.

In my spare time, I watch various videos in the "recommendations" of Bilibili to kill time. I watch whatever is recommended and I don’t follow any UP hosts. Sometimes I use the search function. Recently, I have been addicted to Honor of Kings. After searching for some commentary videos and guides, I found more content related to Honor of Kings in the recommended videos, so I watched them in the recommendations.

User Type 2:

Xiao B, male, 14 years old, lives in Shanghai, and is a junior high school student.

Ghost animal lovers initially looked for videos in ghost animal channels, but now more than half of the recommendations are ghost animal videos.

When I see an interesting video, I will watch other videos with the same material below the video. If you are not satisfied with what you see, just search for keywords and continue reading. I often laugh until my stomach hurts, and I recommend funny videos to my friends.

User Type 3:

Xiao C, female, 24 years old, lives in Hangzhou. She stayed in the local area to work after graduating from university. She comes from a wealthy family and has no pressure in life.

She has a sweet and cute appearance. Initially, she posted videos about beauty, fashion, and travel on Weibo, Bilibili, and other platforms, and had tens of thousands of female fans. Now she has started vlogging to record her life, and her number of male fans is gradually increasing.

I opened an online store selling cosmetics and clothes that I recommend, and I also help people take artistic photos. He is now a full-time internet celebrity, increasing his influence on Bilibili to generate income for his online store.

User Type 4:

Xiao D, a science and engineering man, is 25 years old and lives in Chengdu, where he is a graduate student.

A member of Bilibili and a fan of the two-dimensional world, he often watches anime on Bilibili to relieve academic pressure. Since he is single, he began to like cute two-dimensional girls and has a tendency to fall deeper and deeper into it.

I like the two-dimensional idol project Love Live, so I browsed the peripherals of Love Live on Bilibili. I will also try B station's two-dimensional games such as Azur Lane.

User Type 5:

Xiao E, female, 18 years old, studying in the United States, a Japanese girl.

If you like Japanese idol group Arashi, you will not miss any videos about Arashi on Bilibili, including: programs hosted by Arashi, concerts, various clips, etc., and post barrages to interact with other fans.

I followed a series of subtitle groups that translated Arashi's videos, followed Arashi's topics, browsed users' updates about Arashi in the feed, and participated in discussions. I often open Bilibili when I feel lonely, and I feel healed.

User Type 6:

Xiao F, male, 22 years old, based in Shenzhen, college student, LOL master.

I often watch game videos and live broadcasts on Bilibili. Later I decided to stream League of Legends myself.

Because he doesn't like the atmosphere of other live streaming platforms but likes the community environment of Bilibili, and doesn't care about the number of viewers and doesn't ask for rewards, Xiao F chooses to live stream on Bilibili at irregular intervals. Interact and get to know a small number of viewers, and often play games together.

4. Functional Analysis

The functional analysis takes Bilibili's two positionings as a comprehensive video community and a two-dimensional community as the entry point. Based on in-depth interviews with users, some needs are summarized and corresponding optimization suggestions are put forward.

1. Comprehensive video community

The increasing amount of three-dimensional content on Bilibili has made it gradually become a comprehensive video community similar to YouTube. The author will analyze it from the perspective of the audience.

1.1 Video Recommendation

One of the most common behaviors of Bilibili users is to watch the videos recommended on the homepage.

Recommendations are mainly based on users’ viewing history, the UP hosts and tags they follow, and the hottest topics at the moment.

In interviews with users and based on my own experience, I found that the proportion of recommended videos that match users' interests is relatively low, with respondents believing that on average only 1-2 out of 10 videos are of interest.

As the number of videos grows exponentially, the quality of videos varies greatly, and the proportion of videos that match your interests in recommendations has decreased. Since most users have limited energy, if there are a large number of uninteresting videos in the recommendations, over time, some users will give up looking for videos in the recommendations and prefer to watch the videos of the UP masters they follow in the dynamics, resulting in a decrease in the usage rate of recommendations. Therefore, there is a lot of room for optimization in the video recommendation module.

1.1.1 Optimization of recommendation mechanism based on video tags

Tags are used to categorize videos. When a user watches several videos of the same type, the system will recommend more videos with the same tags.

Each video may have 3-4 tags at least and 10 or more tags at most.

Take the following video as an example: a news clip about Liu Xiao Ling Tong has 11 tags, including tags with low relevance such as "Stephen Chow" and "A Chinese Odyssey". This is because in order to increase the possibility of the video being seen, UP hosts usually add the maximum number of tags as much as possible, especially for non-original UP hosts, such as porters and marketing accounts, the phenomenon of abusing tags is very common.

Although the content of the video is an interview with Lu Xiu Ling Tong and has nothing to do with Stephen Chow, this video was recommended to users with the label "Stephen Chow".

If the user feels disgusted and wants to block the video, he will click on the "Not Interested" column. The user can select the reason for not interested to make the feedback more accurate.

As shown in the figure below, from the user's perspective, you can find that there are no suitable options!

Since the user is not interested in news about Lu Xiu Ling Tong, but not in Stephen Chow, the option "Channel: Stephen Chow" is not suitable.

“Partition: Stars” has too wide an impact. Users who just like Stephen Chow will not choose it. Users may also follow other stars, so the possibility of choosing it is even lower. As for the UP host's options, users see the video due to the imperfect recommendation mechanism, but they are not disgusted by the video itself, at most they are not interested in it. Therefore, the probability of users choosing to "ban" the UP host is very low.

On the other hand, if users "ban" UP hosts simply because of problems with the recommendation mechanism, the possibility of UP hosts being blocked by users will greatly increase, which is something that neither UP hosts nor audience users want to see.

In summary, the user's best option is "not interested", but feedback without a clear reason will cause all tags of the video to be downgraded in the recommendation priority, which may include tags that the user is interested in, further causing inaccurate recommendations.

Optimization suggestions:

1. Reduce the maximum number of tags per video. The above problem is caused by the fact that video recommendations are based only on a single tag while the video itself has multiple tags. Therefore, reducing the number of tags can allow UP hosts to select the most relevant tags for the video, so that the recommended videos are more consistent with their tags, and tag-based recommendations can be more in line with user interests.
2. Recommendations are made by matching multiple tags of a video. For example, a video is recommended to a user only when it contains more than two tags that match the user's interests.
3. In the Not Interested column, users can select any number of tags of the video that they are not interested in. Clearer feedback can make the recommendation algorithm converge faster and produce an accurate user profile.
4. Enhance the review of video tags before publishing. Tags with low relevance and hype should be deleted.

1.1.2 Optimization of recommendation mechanism based on user feedback

Bilibili builds user profiles through user feedback on videos. User behaviors include but are not limited to: "Did you click to watch?", viewing completion, "Are you 'not interested'?", likes, coins, and following UP hosts, etc.

As the number of users increases, some non-core users use Bilibili just to watch videos, mostly watching them in recommendations or using the search function. For example, busy working people will use fragmented time to watch some short videos on Bilibili to relax themselves.

These users open Bilibili multiple times a day, but the duration of each visit is relatively short, usually less than 10 minutes. This group of users often do not have the time or interest to provide feedback on videos. They will not use the like and coin functions to follow the UP host for videos they like, and they will not click "not interested" for videos they do not like.

Based on interviews with users and my personal experience, I concluded that the recommendation mechanism of Bilibili based on user feedback is as follows:

The current recommendation mechanism is more suitable for users who provide active feedback, but is not perfect for users who provide less feedback.

Optimization suggestions:

In order to increase the relevance of recommendations for users without strong feedback, a weak feedback mechanism can be added:

Regarding the current situation of repeated recommendations on Bilibili, if a user ignores the same video three times, it is likely that they have seen it but are not interested in it, so the video should stop being recommended.

If a user searches for the same keyword multiple times, even if he or she does not watch the video or follow the corresponding tag after the search, Bilibili should add recommendations for related videos. However, this function does not yet exist.

1.1.3 Video evaluation system optimization

When users see a video they dislike, they can “dislike” the video (Android) or give it a “bad review” (IOS) (hereinafter collectively referred to as “bad reviews”).

The number of negative reviews is not counted on the user side. A user cannot know how many other users have given the video negative reviews, and the number of negative reviews will not offset the number of positive reviews. Users cannot understand the content and quality of a video without clicking on it, which results in that once a video has a traffic entrance, such as being recommended, even a low-quality video can maintain an increase in clicks. Users need to watch the video for a while, read the barrage content, or check the comments first to know the quality of the video, whether it is a clickbait title, etc.

For example: The users interviewed said that if they saw comments such as "Marketing account warning" or "You will waste xx minutes of your life" at the beginning of a video, they would close the video immediately. Even so, users spent a small amount of time on videos they were not interested in. Therefore, a more intuitive way should be provided for users to determine whether a video is what they want to watch.

Android interface: "Dislike"

IOS interface: “Bad review”

Optimization suggestions:

For recommended videos, users can learn about their praise rate without clicking on the video, where "praise rate = number of praises/(number of praises + number of negative reviews)".

It can also be replaced by a positive review index, where "positive review index = (number of positive reviews - number of negative reviews) / total number of clicks". This way, users can understand the quality of the video more intuitively and efficiently. For videos with low scores and uninteresting titles, users can directly skip them.

It should be emphasized that the praise rate only needs to be applied to videos in the recommendation interface, because the recommended videos have not been screened by users and are most likely to be a mixture of good and bad. Videos found under other usage paths do not need to display the praise rate because they have been screened by users. For example: videos searched by users, videos posted by UP hosts followed under dynamics, etc.

The rate of positive reviews can also be used as one of the bases for recommendations to improve the quality of recommended content.

1.1.4 Optimize the number of video refreshes

Taking the usage of iPhone and Xiaomi mobile phones as an example, there will be 10 new videos in the recommendation each time you refresh, but one screen can only display 6 videos, and the other 4 videos need to be scrolled down to see.

Although users can refresh the video by clicking "Just saw this, click to refresh" below the 10th video, this button is rarely used and most users still prefer to go back to the top page and pull down to refresh. Because pull-down to refresh provides a better experience and is more in line with user habits on other apps.

The result is that most users only watch the first 6 videos before refreshing, and when they are bored, they will constantly pull down to refresh in order to seek pleasure.

First 6 videos Next 4 videos

Optimization suggestion: The viewing rate of the last 4 videos is relatively low, so change it to refreshing 6 videos each time you pull down.

1.2 Video Tags

The video tags of Bilibili are equivalent to channels, which contain recommendations and topics. Recommendations can provide users with the latest and most popular videos under relevant tags. Topics are users' creative areas in the form of text, reprinted articles, pictures, original videos, reprinted videos, etc. Users can also create groups and organize activities here.

Topics is a section with relatively rich UGC, and is also one of the places with the highest exposure rate for ordinary users on Bilibili.

In addition, users can also see the active user rankings under the tag and follow them to discover more high-quality videos. In the above comparison between Bilibili and First Bomb, it was mentioned that Bilibili has cancelled the interest circle function. Tags are currently the closest function to interest circles in Bilibili.

B station's tag function

1.2.1 Tag entry optimization

During user interviews, it was found that the usage rate of tags was not high, and most of the interviewed users had never clicked on a tag.

In fact, the tag usage rate of users across the entire site is relatively low. Taking the most representative ghost animal channel of Bilibili as an example, there are 757,000 users subscribed to the ghost animal channel, and the most popular ghost animal video has tens of millions of views. Videos with millions of views are already very common.

There is an order of magnitude difference between subscriptions and playbacks. One important reason is that the entry of the tag is not obvious or too deep, making it difficult for users to find it.

There are currently two ways for users to follow a tag:

1) There are several tags under each video, and users can click on the tags and follow them.

In actual use, it is found that when users finish watching a video, they will more often choose to watch other videos below, read the comment section, like and follow, or exit.

Clicking on a tag is the least frequent operation because the tag bar only occupies one row of space, and more tags will be folded. Visually, users are more likely to be attracted by the video below and ignore the existence of tags.

Even if you click on a tag, users will see the video under the tag. Since the video is similar to other videos below the video, users will find the tag useless and it is more convenient to directly watch other videos below the video. In summary, the tags under the videos are less attractive to users, and it is difficult to increase tag subscriptions through this channel.

Channels - Discover more channels - Find/search channels (channels are tags), as shown below:

The entrance to this path is too deep, and usually only users who specifically look for tags will use it. The general user usage path cannot reach it, so the frequency of use is extremely low.

Why increase label usage:

1. The use of tags expands the dimensions of content that users receive, improves the accuracy of content, and provides a better user experience for subscribers of tags.
2. High-quality UP hosts can get more attention through the rankings, which increases the user stickiness of the UP hosts.
3. The use of tags can gather users with the same interests, which is the basis for carrying out related operational activities. In addition, the number of banners on the entire B site is relatively small compared to the number of users, but banners can be added within tags (as shown in the figure below). The content of the banner, whether it is an advertisement or an operational activity, is conducive to user conversion.

Optimization suggestions:

In order to increase the usage of tags, a shallower entry should be added to tags under common user paths: when users search, if there are tags that are the same or similar to the search content, the tag will be included as part of the search results and placed at the top , as shown in the following figure:

When users use search, it means that they are interested in the search content. If the tag can be searched and placed at the top, the probability of users paying attention to the tag in this scenario is much higher than other usage paths.

1.2.2 Label content optimization

Currently, video tags can only be added by UP hosts.

Although the UP host is the producer or porter of the video, his personal perspective is relatively single after all, and the labels he gives to the video may not be the most in line with the audience's ideas. Inaccurate labels will cause problems such as decreased accuracy of label-based video recommendations. If tags based on audience feedback can be added, the video will have a higher audience rate and will be more in line with the ecology of the interactive video community.

Optimization suggestions:

1) Since viewers do not have the authority to modify or add tags, Bilibili can collect users’ barrage content and recommend appropriate video tags to UP hosts. Ultimately, it is up to the UP host to decide whether to add them.

After a video has been online for 10 days and has more than 300 comments, 1-2 tags will be recommended to the UP host based on the frequently appearing comments. The UP host can choose to add these tags to the video.

The feasibility of doing this lies in the fact that the content of the barrage reflects the content of the video to a certain extent, especially since many videos have a large number of repeated barrages. Apart from the most common ones such as "Hahahahaha" and "66666", most of the repeated comments are memes. For example, "Who can resist this?" often appears in videos about cute pets, curiosity, and beauties; "Wu Yifan, come in and get beaten" often appears in funny and music videos; videos with a large number of comments like "Welcome back" and "Treasure of the station" are often of higher quality.

The tagging mechanism based on barrage has the following advantages: since it comes from the audience, such tags are often highly relevant to the video content, and more accurate tags can improve the accuracy of homepage recommendations.

2) Provides richer interest classification.

Let’s take the common comment “Who can resist this?” in videos about cute pets, curiosity, and beauties as an example: it is rare for a normal tag to cover these three types of videos, but “梗” can. Nowadays, memes are more likely to exist in video titles. Once the memes are tagged, meme enthusiasts can find more content when searching for videos using memes as keywords.

3) Memes contain more information than general labels. With meme-based labels, the accuracy of the labels can be guaranteed even if the number of labels is reduced.

Add tags based on bullet comments

1. Second dimension community

The core two-dimensional functions of Bilibili mainly include: following anime, virtual anchors, membership purchases and mobile game joint operations.

Since membership purchases and mobile game joint operations are mainly market-oriented and focus on operations, the author will mainly analyze the two modules of following anime and virtual anchors.

2.1 Optimization of the drama comment area

It has been mentioned in the competitive product analysis that B station’s social interaction is mainly one-way communication between “UP host-audience”, and lacks social channels between ordinary users; and if the social interaction of ordinary users has no interest basis, it will lead to low-quality content and become a platform for sharing resources.

You can try to open the social network of ordinary users in the anime-chasing section, because the two-dimensional group is the core user of Bilibili and their quality is relatively high. Having interest in the same anime series as the basis for communication can produce relatively high-quality UGC.

Currently, each anime’s discussion area currently has long reviews, short reviews and comments. Comments is the function with the shallowest entry and the largest user traffic among the three. Comments are often short and casual, have no topic, and have a low response rate. Once there are updated comments, the original non-popular comments will sink to the bottom. For ordinary users who cannot be included in the popular comments, the comment area cannot bring about interaction.

Non-popular comments

The content of a short review is similar to that of a comment, without a theme and relatively brief. In addition, short comments cannot be commented on, so the interaction is zero and users do not need to read other people’s short comments. Its main purpose is to provide ratings, and users must write a short review in order to rate it.

Although long reviews have topics, authors need to write long paragraphs, which poses a certain threshold for ordinary users. Long reviews are often subjective, and readers may not have the interest or time to read them. Even if they do read them, they may not be willing to participate in the discussion. Therefore, whether from the perspective of the author or the reader, the communication threshold for long reviews is relatively high, and it belongs to the interaction of a small group of people.

Long review Short review

Therefore, the following conclusions can be drawn:

1. The anime-chasing section of Bilibili lacks a social channel between long reviews and comments, which can allow users to initiate a topic with a low threshold and arouse discussion among a large number of users.
2. The short review function that exists for the purpose of rating is rather useless.

Optimization suggestions:

1) Separate the rating from the short review: Users do not need to attach a short review when rating, they only need to give a rating.

Cancel the short comment function and merge the original short comments into the comments. Users can still post comments multiple times and reply to, like and dislike other comments. When posting a comment, their rating will be displayed by default, but they can also hide the rating through settings; there is no need to display the rating when replying to a comment.

Before/After

2) A new "Discussion" section has been added, which will be placed alongside the long review section, where users can post discussions on topics related to the anime.

To reduce spam, the publisher must include a description of no less than 15 words. In order to increase the activity of the discussion area, the topic posts are sorted by the most recent reply time. In addition, in order to increase the exposure of the discussion area, the discussion label is placed on the left and the long comment label is placed on the right, and users will see the discussion area first.

In order to increase the exposure and usage rate of the discussion area, you can also add an entrance to the discussion area: since most users will browse the comment area, the top 3 hot topics in the discussion area can be pinned to the top of the comment area as "hot discussions".

Compared to comment sections where messages are easily overlooked, users who want to get a reply tend to communicate in the discussion section. Placing the discussion area at the top of the comment area can attract a group of users with social needs.

Revised comment section: Top up popular discussions

Compared with the comment area, the discussion area can gather more core followers; compared with long comments, the atmosphere in the discussion area is more relaxed and can trigger more interaction. The discussion area can create conditions for operating activities and advertising targeting crowds.

2.2 Virtual anchor optimization

The virtual anchor is a section in the live broadcast area of ​​B station. It is unique in that its anchor has a two-dimensional image.

The concept of virtual anchor was first established in Japan in 2016. In August 2017, Xiaoxi, the first virtual UP host in China, met with everyone. At present, the industry is in its infancy in China. As a leading second-dimensional platform, B Station recently put the virtual anchor section on the homepage of the live broadcast area and introduced a series of virtual anchors from Japan. It can be seen that virtual anchors are one of the important businesses of B Station in the future.

Currently, the virtual anchors on B station can be roughly divided into two categories:

1. One category is: virtual characters created by professional teams. They are highly similar to real people and have their own names, backgrounds and personalities. The live broadcast content is to record one's own life, such as studying at home during the day, sleeping at night, etc. Some anchors also have the ability to interact with the audience. This type of virtual anchor requires professional animation teams, voice engineers, and artificial intelligence assistance. It takes tens of millions of investment to create such virtual anchors, so it is mostly operated by companies, and B station also introduces this type of anchor.
2. Another type of virtual anchor is: ordinary users track their expressions and actions through third-party animation software and map them to the animation image, and the sound comes from the user himself. Due to the low cost, this type of live broadcast accounts for more than 90% of the virtual anchors on B station.

Since the first type of anchor is a competition of capital and technology among teams, and there are few samples, the author will analyze the second type of live broadcast .

Currently, the virtual anchors have relatively single expressions and actions, mainly blinking and turning their heads. Therefore, the interaction with the audience is mainly through sound, which has great limitations.

Since the image of a virtual anchor is a decoration, the audience can only listen to his explanation, so the live broadcast format is mainly game live broadcast , and it is difficult to start with show live broadcasts, which is very attractive to two-dimensional users.

Common virtual anchors and their game live broadcasts

Optimization suggestions:

The virtual anchors created by the team tend to be idolized, while in contrast, the virtual anchors of individual users are more entertaining. Because compared to real people facing the audience, virtual anchors have greater room for performance in character creation and interaction with the audience; and most virtual anchors take the cute route, and the audience has a higher tolerance for them, and their image is not easy to collapse.

Since the bottleneck of virtual anchors is ultimately technology, B station can provide some technical interfaces to add more gameplay to virtual anchors and drive the popularity of the virtual anchor area.

Enrich the expressions and body language of virtual anchors, and gamify the interaction between the audience and the virtual anchors, mobilize the audience's initiative towards paper people. If the show live broadcast can be made, the number of viewers and retention rate of the second-dimensional live broadcast will be greatly increased; and the show live broadcast will have more audience rewards than game live broadcasts, and Bilibili can get more commissions from it.

There are several specific aspects:

1) Virtual anchors need to have more types of expressions and actions. This does not require the live anchor behind the scene to show the corresponding expressions and styling, but allows the live anchor to express the emotions of the second-dimensional anchor he controls through the function buttons, such as: calm, joy, sadness, concentration, etc.

2) Users can interact with virtual anchors on the audience. In the independent interactive area of ​​the virtual anchor, the viewer clicking on the interactive area will not affect the live viewing, that is, it will not cause problems such as adjusting brightness and sound, and pausing of the live broadcast.

For the audience's "flirting", the virtual anchor will respond accordingly based on the current state.

As shown in Figure 1, when the anchor is in a low mood, if the audience clicks on the anchor's head, a 2-second "head touch" animation effect will appear.

As shown in Figure 2, in order to avoid confusion, only viewers who interact with the anchor can see the anchor's response, and users who do not click on the interactive area can only see the anchor's default status. When clicking on the anchor head, the user can see the situation in Figure 2; at the same time, the user without operation still sees the situation in Figure 1.

Figure 1

Figure 2

If the anchor clicks on the anchor's head when the anchor is in the "focus" state, the anchor will show "angry", as shown in Figure 3 and Figure 4:

Figure 3

Figure 4

3) The dress of the virtual anchor can be determined by the audience who rewards it, and B station needs to provide the virtual anchor with an image choice.

The first user on the Golden Gourd Seed List, Gift List, and Fan List, as well as the Captain of the Great Navigation, can vote for their favorite outfits, and users who have rewarded the anchor can participate in the vote, with a deadline of one week. After the voting, the anchor changed into the clothing with the highest vote rate, which lasted for a week.

Dressup diagram

V. Conclusion

Since its establishment in 2009, Bilibili has gradually become a comprehensive barrage video community where young people gather.

As the scale of the pan-2D group becomes increasingly large, Bilibili needs to consider how to explore at the forefront of the two-dimensional industry, cultivate more core users, and improve the conversion rate from core users to consumers.

As a comprehensive video community, B station needs to improve the way content connects users and distribute the most suitable content to users. In addition, how to open up the social transportation routes for ordinary users is also something that Bilibili needs to consider.

Source: FMR