Tik Tok’s 3 loops to retain users and its commercialization trends!

This article is mostly the author's own thoughts and speculations. Due to its length, it is divided into three parts:

1. TikTok's three loops to retain users: Interpret TikTok's current model and summarize TikTok's magic of attracting users with the help of the AARRR model;
2. Douyin commercialization and its trends: boldly speculate on future trends based on Douyin’s existing business model;
3. Several problems that Douyin may face: Several current problems of Douyin are raised and possible solutions are given.

01. TikTok’s three loops to retain users

Tik Tok seems to have "magic", which attracts users step by step into the "Garden of Eden" it has built and allows them to "live happily", and behind this magic are solid product, operation and marketing basics.

Here we use the AARRR model to try to analyze the mystery behind the "magic" of Tik Tok through "3 loops".

Douyin uses three loops : "user loop", "content loop" and "user-content loop" to firmly lock in users, and uses retained users for commercial monetization and acquiring new users.

1. User loopback

Douyin continuously acquires users through the "user loop" of "acquisition-activation-retention-reacquisition" to form its own user pool - it acquires users through relationship chain diversion, cross-industry cooperation, ASO and other methods;

Then use excellent product experience to guide and activate users, so that users can quickly experience the value of the product;

Then retain users through personalized high-quality content recommendations to form your own user pool;

Finally, users in the user pool will recommend high-quality content to their friends, helping Douyin acquire new users and forming a user loop.

The following is a brief summary of the main actions of each link:

1.1 User Acquisition

• The early V1.1.0 version provides a relationship chain import function, which supports importing address books, QQ, WeChat, and Weibo friends, providing product support for viral transmission.
• In the early stage, we used Toutiao, the “Dear Father”, to divert traffic and complete the cold start.
• In the early days, thanks to the promotion of KOLs such as Yue Yunpeng and Wu Yifan, the number of users increased dramatically
• Cooperate with many variety shows, such as Happy Camp, Day Day Up, and The Rap of China, to drive user growth
• …

1.2 User Activation

• Excellent preparation: In terms of product, we should prepare for a large influx of users while ensuring a complete experience; in terms of operation, we should introduce early creative users and high-quality content in advance.
• After the user enters the product, the full-screen high-definition video will be automatically played without any operation, instantly attracting the user's attention and allowing the user to feel the core value of the product.
• It is easy to use and provides operation guidance such as swiping up to play the next video and double-clicking to like, which teaches users the basic actions of exploring the product in a very short time.
• Guide users to create short videos at the right time, further explore products, and become content producers
• Through props, music, demonstrations, etc., the cost of short video creation is greatly reduced, and the possibility of user experience of the core value of the product is greatly improved.
• …

1.3 User Retention

• Through popular content screening + personalized recommendation algorithm, we recommend content that users are really interested in, gaining users' recognition of product value while increasing the possibility of users returning to the product after exiting.
• Through easy-to-use functions such as likes and comments, we promote interaction between content consuming users and content producing users, and encourage users, especially content producing users, to stay
• Through interesting shooting functions, and constantly updated props and music libraries, users are attracted to become and continue to become content production users and retain them.
• By making the video itself “popular” and having a high number of likes, users are inspired to think “I can also shoot a popular video”, which in turn stimulates users to continue to create and retain videos.
• Through the "See Music Plan", original musicians are attracted to contribute music content on Douyin and retain it
• Recall users through push notifications
• …

1.4 User Recommendations

• Through personalized recommendations and interesting creation methods, users' recognition of video content and products is greatly improved, making it possible for users to spread the content spontaneously.
• Add logo watermarks to videos distributed outside the website to increase customer acquisition channels
• In order to gain more attention, content producers may spontaneously spread their own videos to help Douyin gain more users.
• …

Through the above four links, Douyin has successfully built its own user loop, allowing new users to stay and play, freely consume or produce content; at the same time, further acquire new users through the accumulated user pool.

2. Content loop

Douyin promotes the continuous production of content through product design, community atmosphere, external stimulation and other means, forming a content loop.

Based on the dimension of content supply and demand, Douyin users can be divided into two categories: content production users and content consumption users.

Content producing users contribute content and gain income; content consuming users consume content and gain pleasure.

The content loop is the dynamic loop of the platform's content production, which generally includes the following aspects:

1. "Popular" videos with high number of likes make users feel that "I can also shoot popular videos", which stimulates their eagerness to create.

2. The endless #challenge activities, combined with the emotions created by the first one, constantly stimulate content production users to create new content, and through the low threshold of event participation and creation tools, convert some content consumption users into content production users.

3. "Take the Same Look" is similar to the challenge, and provides a solution with a lower threshold for creation.

4. Based on the first three points, some users are successfully converted into content production users, and the benign interaction between content production users and content consumption users constantly stimulates the former to create new content, forming a content loop.

5. The first four points mainly discuss ordinary content generators. In addition, there are some internet celebrities and self-media big names who are for profit. Their desire for traffic or income directly forms external stimuli, becomes a continuous driving force for their creation, and constitutes a content loop.

6. A community atmosphere that encourages creation also has a positive impact on content recycling.

7. In addition, with the surge in traffic on Tik Tok, external public opinions such as “how to make a viral Tik Tok video” also have a positive impact on the content loop.

8. Finally, after the content loop is formed, Douyin continues to convert content-consuming users into content-producing users, and continuously converts content-producing users into higher-frequency and higher-quality content-producing users, continuously optimizing the content loop, making it spiral upward and entering a virtuous circle.

Through the methods mentioned above and those not mentioned yet, Douyin has built its own content loop and provides support for user loops and user-content loops during continuous optimization.

3. User-Content Loop

Tik Tok uses artificial intelligence algorithms and other technical means to continuously optimize the matching effect between users and content.

As the number of users increases, user relationship chains and video content also increase, the data available for machine learning becomes richer, the accuracy of personalized recommendations is improved, and user satisfaction with recommended content increases, thereby stimulating users to consume and produce more content, forming a user-content loop.

The user-content loop is to optimize the matching efficiency of the two. The main methods are as follows:

1. Through the content screening mechanism, in the continuous testing of each video, the high-quality content that users are most interested in is screened out to form a content pool, thus ensuring the content quality from the source. There are a lot of speculations about TikTok’s content screening mechanism on the Internet, so I won’t go into details here.

2. Match the content in the content pool with the user and provide personalized recommendations for the user. I guess Douyin uses the same “collaborative filtering + content recommendation” algorithm as Toutiao for recommendations.

There are two types of collaborative filtering algorithms:

One is based on the "people" dimension , which first finds people with similar interests to the user through user behavior (such as likes), and then recommends the videos they like to the user;

The other is based on the "video" dimension . It first obtains the similarity between videos through the video preferences of all users, and then recommends similar videos to users based on the user's historical favorite video records. However, this algorithm cannot solve the cold start problem, that is, it cannot recommend videos to users that have not yet generated user behavior data (such as likes).

Another algorithm complements this well - the content recommendation algorithm, which finds similar videos through video tags and recommends similar videos to users based on their historical preference records. The tags can be manually tagged or abstracted from the video content itself.

In addition, through the prompt "People you follow liked this video" that appears in the recommended videos, it is speculated that the algorithm also integrates social relationship networks.

3. Recommend non-matching content in the content pool to users, relieve users’ aesthetic fatigue, expand users’ interest tags, and dynamically update the user-content matching relationship.

4. Recommend people who may be of interest to users, enrich the relationship chain, and optimize the matching effect. The recommendation categories visible in the product are roughly: friends in the social chain, people with common friends, people you may know, people who people you follow also follow, people who may be interested, etc.

The user-content loop accomplishes the important task of user retention by continuously optimizing the matching relationship between the two, allowing users to have a content consumption experience that they cannot stop scrolling.

In summary, we can see that the three loops have a "you in me, me in you" relationship. The optimization of any one loop may lead to the improvement of the effects of the other two loops. This is what I understand as the magic of Tik Tok that makes me unable to stop.

02. Douyin’s commercialization and its trends

For commercial products, monetization is an unavoidable issue. Based on the summary of Douyin’s current commercialization behaviors, this article will make bold guesses about Douyin’s possible future commercialization path.

1. Monetization methods

First, let’s summarize Tik Tok’s current visible monetization methods.

1.1 Feed Stream Advertising

Advertising is Toutiao’s strong point. In addition to matching users algorithmically, Douyin’s feed ads also have some interesting small attempts that I would like to briefly discuss.

(1) Determine the length of the ad button through A/B testing. I wonder if you have noticed that the length of the conversion button in the Douyin advertisement has undergone several changes - from the initial size to the full-column length, and finally to the current style. The idea behind this should be the same A/B test as the content screening mechanism, that is, to find the button length with the highest click-through rate through testing.

(2) Arouse users’ curiosity by randomly changing the color of buttons. The same is true for advertising buttons. When you first enter the advertisement, each button is transparent, and will change to a different color after a few seconds. For example, the button in the first advertisement will turn blue, and the button in the second advertisement will turn orange. This takes advantage of the user's curiosity about the color of the button, while providing fun for the product and extending the viewing time of the advertisement.

1.2 E-commerce traffic diversion

Merchants can use short videos to directly direct users to their product sales pages.

Although Toutiao launched its own e-commerce platform "Zhidian" some time ago, most of the e-commerce traffic visible to the naked eye is still directed to Tmall and Taobao. On the one hand, it may be because the user profile of Douyin is quite different from that of Zhidian. On the other hand, it may be because Zhidian is still in the trial stage and is not suitable for large-scale traffic diversion.

1.3 Customized Brand Advertising

Utilize challenge activities and materials to help big brands customize advertisements and profit from them, such as customized advertisements for #challenge - OPPO Challenge "So you are like this", customized advertisements for stickers - Pizza Hut stickers, etc.

1.4 Live streaming revenue sharing

It is speculated that it is just a general function follow-up, which directly monetizes by leveraging existing users whose attributes and purposes are similar to those of live broadcast users. There is not much difference between it and other live streaming platforms, so I will not go into details.

1.5 DOU+ quick push

DUO+ Speed ​​Push is an advertising format recently launched by Douyin. It can be delivered to a specific video and can be operated on mobile devices.

Delivery methods include:

1. Basic targeted delivery, targeting people of limited gender, age, and region;
2. Specify accounts with similar followers and distribute ads to the fans of the specified accounts;
3. Place it in nearby business districts and conduct regional placement within a specified diameter range.

The above are the commercialization methods of Tik Tok that are currently visible to the naked eye. Next, let’s speculate on where Tik Tok will focus its monetization efforts in the future.

2. Monetization focus

Personally, I believe that Douyin’s future monetization focus will still be “advertising” and “traffic diversion”.

2.1 Advertising

It is reported that Douyin’s advertising revenue may exceed 10 billion yuan in the next year [1]. It is hard to imagine that such a big piece of cake cannot become the focus of Douyin in the future.

In addition, the advertising skills honed by Toutiao are also a good support for Douyin's advertising journey.

For example, based on the delivery of Toutiao, Douyin continued the idea of ​​testing experiments and provided "intelligent delivery".

Under this delivery method, Douyin will first use a portion of the delivery funds to conduct a small-scale test, and then conduct large-scale delivery after finding the group with the best delivery effect, so as to achieve the best promotion effect (Note: please refer to the first option of DOU+ Speed ​​Push, the way to call out DOU+ is: just click the "Forward" button on any video).

Finally, TikTok ads do not seem to be that offensive to users, providing TikTok with the possibility of making money without hurting users.

Among them, "less disgust" may be attributed to the following :

(1) The advertising style fits the product attributes

Tik Tok is an entertainment product, so the advertisements in Tik Tok are almost all entertainment-style.

1. On the one hand, according to the report provided by App Growing[2], the proportion of “game” advertisements on Douyin is very high, and the entertainment nature of the game itself is very consistent with the product tone of Douyin. Therefore, the effect of placing games on Douyin may even be better than that of the information-based Toutiao.

2. On the other hand, the production of non-game advertising content on Douyin is also trying to fit in with the entertainment attributes, and some are funny, creative, and even weird. For example, Sophie’s creative advertisement received rave reviews at the time [3].

(2) Mark the words “Advertisement” to reduce the cost of skipping

All ads in TikTok are marked with the word “Advertisement” and the cost to skip is very low. Just like seeing a video you are not interested in, you can just swipe up to skip it.

(3) The push frequency can be artificially reduced

After clicking "Not Interested" on the ad multiple times, the push frequency dropped significantly.

I did an experiment with a friend. I served as the control group and did not click on the ads to show "not interested". My friends served as the experimental group and frequently clicked on the ads to show "not interested". After 7 days, ads would appear after I scrolled through less than ten pieces of content, while my friends would scroll through dozens or even hundreds of pieces of content without seeing any ads.

In summary, "advertising", a means of monetization with high income, strong capabilities and no harm to users, should be the focus of Douyin in the future.

But at the same time, Douyin's ads seem to have room for improvement. Here are two small suggestions for matching ads with users:

1) Improve the ability to analyze “video advertising content”

Although advertisers can currently choose ads and user types when placing ads, this may not be the best matching method. After all, "people" may not understand users better than "machines" in matching algorithms.

If the "video content" can be understood by machines, optimized using methods similar to those used for graphic ads, and reliable suggestions can be provided to advertising video producers, the advertising effect will surely be even better.

2) Create and further enrich user "commercial tags"

User commercialization tags may be different from user interest tags. For example, people who like cute pets do not necessarily have the need to buy pets or pet supplies.

Perhaps similar tags already exist on TikTok, but based on the current user experience, there seems to be a lot of room for improvement.

If the above two points are improved, combined with Douyin's existing screening and distribution capabilities, I believe Douyin's advertising will have breakthroughs in experience and revenue.

2.2 Traffic diversion

E-commerce traffic is also part of Douyin’s revenue source. According to a report provided by App Growing [2], Tmall and Taobao currently account for a large portion of traffic.

But Douyin’s ambition does not seem to stop there. Does the classification module under the “Same City” section look familiar? Food, attractions, fun, etc.

Huh? Isn’t this what Dianping is doing?

Therefore, it is speculated that Douyin’s focus of “traffic diversion” in the future will not only be limited to e-commerce, but will continue to expand to life services.

So how feasible is this thing?

It seems not low:

(1) Videos are more expressive

Compared with the text and picture format of Dianping, the video format of Douyin is more conducive to arousing emotional resonance among users and stimulating consumption.

Think about yourself every time you drool over the delicious food on TikTok at night. Have you ever silently noted down the name of the store in the comments and went there to try it?

(2) Excellent community check-in culture

Do you still remember the "check-in internet celebrity stores" that can be seen everywhere in video titles and comments? Bowl-smashing wine, ice cream, and sky realm. In the fun, delicious, and beautiful atmosphere rendered in the videos, many users traveled to various parts of the country to consume and posted "check-in" videos, while some users who had no conditions to travel far could only silently like them.

However, after Douyin launched the function of recommending "Internet celebrity stores" near users, has it further encouraged this atmosphere and made the distribution of life services possible?

(3) Good external environment

I don’t know if you have noticed this, but after a long trip, I have started searching for other life service rating software besides Dianping.

As the marketing atmosphere spreads, Dianping's credibility seems to be gradually declining. Does such an external environment create an opportunity for Douyin?

In summary, backed by its own large active user base, Douyin's ability to distribute life services by directly directing traffic to merchants is likely to constitute another monetization point.

However, Douyin still has a long way to go to distribute life services and directly connect with various businesses. Here are some suggestions:

1. If you want to play the "credibility" card well, you will inevitably need to do some "heavy work" of manpower excavation in the early stage, including the discovery of high-quality merchants, manual field exploration and review, etc. Maybe recruiting "special" merchants is a good idea. On the one hand, it avoids competing with Dianping in the "mass merchants" dimension; on the other hand, its characteristics are more in line with the tastes of Douyin users.

2. Corresponding functions will be steadily followed up, which may not necessarily be "ratings". Equally persuasive functions such as check-in, popularity, and even "video collections" can help users screen high-quality businesses.

3. Control the rhythm of monetization and move forward steadily while striking a balance between credibility and commercialization.

03. Several problems that Douyin may face

In the final article, I will raise several problems that Douyin may face within the scope of my limited knowledge and conduct divergent thinking.

1. Content supervision

With the surge in users, Douyin's social influence has grown, and the risks of the legality and compliance of its content have gradually increased.

Since April 2018, my country has strengthened supervision over short video content, involving a series of issues such as youth values, counterfeit goods, and copyright; Douyin has actively cooperated, adding risk warnings and reporting functions to its products, and using a mechanism of joint machine and manual review to screen risky content.

Since pure machine review is not possible at present, in terms of content supervision, Douyin still needs to train manual reviewers corresponding to the user growth rate; on the other hand, increasing the guidance of the community atmosphere can also alleviate the problem to a certain extent, such as adding more detailed guidance in the product aspect, and repeatedly conveying the correct community concept to users.

2. Content quantity and push quality

Many users have reported that Douyin’s videos are very attractive to them in the initial stage of use, but as time goes by, they will have problems such as “getting tired of watching” and “the videos are basically the same types they have seen before”.

There are two main reasons for this problem:

2.1 There are too many homogeneous videos and no fresh blood is flowing in

In order to solve this problem, Douyin has experimented with activity content and props, launching new #challenges for users to participate every now and then, and props and stickers are also constantly being innovated.

However, there is still room for improvement in product design - such as optimizing the path to participate in the #challenge: the latest challenge entrance can be directly provided at the shooting function to guide and inspire more users to participate in the challenge; a shooting entrance to directly enter the challenge can also be added to the challenge message, and the props required for the challenge can be directly provided.

On the other hand, Douyin may also consider providing incentives to users who produce "high-quality content that triggers a certain amount of imitation" to drive the creation of high-quality original content.

2.2 There is still room for improvement in content distribution algorithms

The reason for content duplication may not only be due to the limited content inventory, but also due to the efficiency problem of the content recommendation algorithm. That is, although there is content that users are interested in in the inventory, it is not recommended to the user as expected.

This problem has occurred in the music field. For example, in the use of NetEase Cloud Music, there may be a problem of recommending a single style of music.

Nowadays, NetEase Cloud Music occasionally pushes some very "strange" songs, such as ghost videos, nostalgic classics, and game BGM...

Tik Tok also often pushes videos that have never been liked. These changes may be due to the addition of a "social network" dimension to the algorithm, which recommends content that other users in the user's social network are interested in to the user after being screened and sorted by the algorithm; it may also be due to the addition of a "random" dimension, which gradually recommends tags that the user has never been exposed to to the user according to their weights and dynamically updates the user's "preference pool".

In addition, there is another possible optimization space for videos - that is, analyzing the video content itself.

Just like the analysis of music tracks, if the content in the video can be identified and labeled, or the video content can be better restored based on the correlation between the tags in the video, then the user-video matching effect will be very worth looking forward to.

3. Addiction may cause user churn

Douyin has captured the attention of a large number of users through its immersive product design and high-quality content. It is said that the average daily usage time of Douyin has exceeded 1 hour.

However, behind the encouraging data, some users have begun to complain: "TikTok kills too much time", "TikTok is poisonous, uninstall it!", "I watch TikTok before going to bed, at least two or three o'clock"... This makes us start to think: Is it really a good thing to extend the user's usage time as much as possible?

If a product cannot continue to provide value to users, it will eventually be abandoned by users in various ways.

For most products, the user abandonment trajectory tends to be:

Normal use - Rarely used - No longer used

For TikTok, perhaps the abandonment trajectory will be:

Normal use - Excessive use - No longer used

As knowledge is upgraded, more and more users realize that they should spend time on things with a "long half-life", that is, things that will bring higher returns in the future, such as systematically learning knowledge, practicing a skill, etc.; and they should not spend time on things with a "short half-life", such as watching Douyin.

So why doesn’t WeChat, which occupies as much of people’s time as TikTok, suffer such criticism?

From my understanding, WeChat is more of a tool-based product for communication among acquaintances. Although it has a very long average daily usage time like TikTok, people tend to gain value from WeChat, such as communicating problems, actively reading an article to improve cognition, and enhancing relationships with close people through video calls. Therefore, the time spent on WeChat is "rewarding" - it actually helps us deal with problems or increase our intimacy with important people.

In contrast, TikTok is a community product with entertainment attributes, and its value is more inclined to allow users to gain short-term benefits through emotions such as relaxation, pleasure, and sharing. The time spent on TikTok is "low-return". Users originally just wanted to regulate their emotions in a shorter period of time, but ended up investing more time and even developing a "sense of guilt".

Therefore, we will not blame WeChat for “killing time” because most behaviors are triggered by us actively and can obtain long-term benefits; however, the product attributes of Douyin determine that it is difficult to achieve the above two points.

Therefore, I personally think that what Tik Tok should do is to fulfill its own value and make users feel "just right" during their entertainment time, rather than blindly pursuing user usage time.

So how can we get people to relax properly on TikTok without staying there for too long?

Douyin has tried a "time lock", where users can set a viewing time, after which they need to enter a password to continue watching; at the same time, Douyin will also display a banner at the top of the video to remind users that the viewing time is too long if the viewing time exceeds 90 minutes.

The above two methods are both "external" constraints in the time dimension, and their functions are relatively hidden and last a long time. They cannot really help users feel "just right", but are more like a compromise to public pressure.

If we change the perspective, is it possible to try to solve the problem from an "internal" constraint?

For example, starting with content to explore solutions.

One idea is to provide content with greater long-term value.

Although at first glance it seems to contradict the "relaxing" entertainment demands of Douyin users themselves, recently more and more knowledge-based self-media seem to have provided a good compromise: providing entertaining light knowledge and packaging knowledge in an interesting form.

If the ratio of "entertainment" and "knowledge" content is reasonable, is it possible to reduce the anxiety of users after long-term use to a certain extent?

Another idea is to start with "anti-addiction" from the content, drawing on gaming experience and reducing rewards after users have used the game for too long. That is, the proportion of "knowledge-based" content can gradually increase with usage time, and you can even take the opportunity to push some content that users may not be so interested in, allowing users to leave on their own, while enriching their own user tag system.

In summary, possible solutions to addiction problems are:

1. External constraints, providing an obvious entry point to limit usage time, or even pre-setting the process;

2. Internal constraints: provide content with “long half-life” value and reduce the “reward” after the viewing timeout, so that users will leave on their own.

Of course, the time to solve the addiction problem may not be now. In the current competitive landscape of short videos, it is indeed too difficult to make such a bold attempt.

4. Search capabilities

Douyin currently seems to have not yet made any efforts in terms of "search" capabilities. Here are two small points that can prove it:

4.1 Poor search experience when selecting music for short videos

It currently does not support searching by lyrics, emotions, music recognition, etc. It is difficult to find the music you want when you have a clear goal.

However, the current processing methods can solve most of the problems:

• Various charts, even if you can't find what you want, you can use other equally good music instead
• While watching the video, you can directly choose "shoot the same style" and use the same BGM to create
• Collect music and search directly from your collection next time

4.2 Poor search experience for short videos

Currently, searches seem to be limited to titles and other text content, so when you have a clear goal, it's hard to find the video you really want.

But the solution currently provided by Douyin is also excellent:

Providing rich content such as rankings and popular challenges below the search can instantly attract users' attention and easily cause them to engage in endless scrolling and even forget their original purpose.

It seems understandable that Tik Tok does not prioritize search at this stage. The main form of consumption for users is to "passively" accept personalized videos that have been screened, and the demand for search is very weak.

But if, as speculated in the second part, Douyin really wants to become a distribution portal for life services, search capabilities still need to be supplemented in the long run, especially the deconstruction and search of the video content itself.

This concludes the series of articles on thoughts on TikTok.

During the writing process, I sorted out the questions that confused me. Although the expression may not be accurate enough and I did not use a combination of pictures and text to take good care of the readers, I still benefited a lot.

Source: Zhou Zhang Jiawen (lilyblood)