Tik Tok new media’s replicable content methodology!

On Tik Tok, only entertainment is the popular content, such as young ladies, funny jokes, music and dance, young men, etc. However, for most companies, the relevant content is not in the entertainment field, but in unpopular fields.

The health field in which Dingxiang Doctor is located is also a niche field. How do we do it?

On March 1, 2018, our new media team applied for a TikTok account. At the beginning, more than a dozen friends from the entire Dingxiang Doctor new media team could log in to this account, record the little things around them, record the beauty around them, and then upload them to Douyin, but the results were very poor.

On April 25, we officially started operating Douyin. Ten days later, the number of followers reached 100,000, and 45 days later, the number of followers reached 1 million.

The Douyin team we initially set up only had 2.5 people, including me, the video producer (videography + post-production), and another young man who appeared on Douyin. He could only be counted as half because he was also responsible for Weibo operations.

2.5 How can an individual achieve 2-3 updates per day? Our process is as follows: 3 people hold a topic selection meeting in the morning, shoot in the afternoon, and edit immediately after shooting. Upload the editing the next day, review the previous day's content on the third day, and propose new topics immediately after the review. Repeat this cycle and iterate quickly.

After operating in this way for half a year, our number of fans reached 5.8 million and the total number of views reached 650 million. There are many popular videos among them. There are videos with millions of views every day, more than 5 million views every week, and tens of millions views in less than a month.

1. Enterprise Douyin needs to have clear goals

Because TikTok is very popular and has a good growth momentum, and short videos are a trend, many companies want to do TikTok, but the content is often not about good-looking boys and girls. In this case, how should companies do Douyin?

In order to answer this question, I want to first ask a very important question: Why do you want to do TikTok?

Let me first explain why Dingxiang Doctor wants to do Douyin.

1) Validate our content production methodology

First, we need to validate our content production methodology.

Dingxiang Doctor’s new media group has media attributes, which means that the content we produce must be able to reach as many users as possible and be recognized and liked by as many users as possible.

Dingxiang Doctor started to provide health science popularization since 2014, mainly providing graphic and text content. Among the four most classic forms: text, pictures, audio, and video, Dingxiang Doctor's graphic content has basically been fully developed in the health field.

We have summarized a complete set of content production methodologies. Next, we need to verify whether our understanding of content and our content production methodology are only applicable to pictures and texts, and whether they can also be applied to videos.

2) Verify our ability to quickly and successfully enter a new platform

For new media, we need different content distribution platforms and use these platforms to reach our users.

Previously, we had a relatively large fan base and exposure on WeChat, Weibo, and Zhihu. Now that a new platform has emerged, with our existing content production capabilities, can we immediately enter a new content distribution platform and get the results we want? We need to verify capabilities.

This includes 2 abilities. First, whether Dingxiang Doctor’s content production methodology supports the production of video content; second, whether Dingxiang Doctor has the ability to enter new and popular platforms.

3) The company’s communication needs

Secondly, there is a more fundamental reason. As a media department, our job requirement is to go to places where the public is more concerned and where the public attention is more focused, so Dingxiang Doctor has to do Douyin.

By analogy, if a company wants to do Douyin, it first needs to think about: What does the company want to achieve? At the same time, what can Douyin bring to enterprises?

2. Douyin’s content operation – research platform

After thinking clearly about why you want to create TikTok, the next step is to do content operation.

In fact, I worked as an obstetrician and gynecologist for 10 years. It seems that I changed my career from doctor to operations, but in my opinion there is not much change. I still do the same thing, which is operation.

Because the clinical thinking model, or thinking process, of doctors in evidence-based medicine can be very well grafted into operations.

For example, Dingxiang Doctor’s content production methodology 1.0 was created entirely using the clinical decision-making thinking method of clinical medicine. Now the content production methodology has been upgraded to version 3.0. It is the same to operate short videos using this method, except that doctors’ tools used to be scalpels, and now new media’s tools are content.

So how do we proceed specifically? First of all, we need to analyze what this platform is like and how our content production should match the new content ecology and content platform.

The new platform mainly studies three points: first, the characteristics of platform users, second, the characteristics of platform content, and third, the platform's content distribution mechanism.

1) Analysis of platform user characteristics

Take TikTok as an example. When we entered the market in April, we read a lot of articles about TikTok, including some official data provided by TikTok. Finally, we came to the conclusion that the characteristics of Douyin users are that they come from first- and second-tier cities, are mostly female (accounting for 60-70%), most of them are born after 1995, and are generally students or newly employed.

In addition, by observing videos, we can understand the time, scenario, and purpose of users’ content consumption. To give a classic example, there is a type of video shot in a university dormitory, where people crawl through a ladder between bunk beds. The scene of this video is a university dormitory, and the user is a girl. She crawls through the ladder to see if she is fat, or to show off her figure.

Through these videos, you will find that the purpose of users watching Douyin is not to gain practical knowledge, but to see interesting things happening in certain scenarios. On a deeper level, these users hope to be thinner and have a better figure.

Through these user portrait analyses, we can determine what content users want to obtain on this platform and what needs the content can meet.

2) Platform content characteristics

Next, let’s take a look at the characteristics of the platform content.

For example, most of the videos on Tik Tok are short videos in vertical format. How short are the short videos? There is only 15 seconds for the first entry, and 1 minute after the number of fans increases.

Douyin also has another feature: it has no operating background for creators, no PC version, only a mobile version. In other words, the data seen by ordinary users is basically the same as that seen by operators, except for the number of views.

This means that we have very few metrics to evaluate our own content.

ps: Currently, Douyin Enterprise Accounts can see homepage data (total playback, homepage visits, etc.), video data (playback completion rate, likes, comments, fans), operation data (POI store visits, shopping cart/product showcase visits, etc.), competition data (industry-level operation data reference), expert data (traffic diversion to merchants), and fan data (fan activity, fan traffic contribution)

What are indicators used for? We need to observe the connection between the indicators and the results we want to achieve, and find the connection to help us achieve our goals. In addition, it is necessary to analyze the popular content and analyze what kind of content can become a popular product on this platform.

3) Content distribution mechanism

In addition, we also need to know how the platform's content is distributed. Different platforms have different content distribution mechanisms.

For example, WeChat public accounts are relatively closed platforms, and after the content is released, it is triggered to spread through a forwarding mechanism.

Because all fans can receive the article after it is published, and more of the dissemination comes from fans forwarding it after reading it. There are two types of forwarding: one is point-to-point, sent to a certain person or to a group; the other is sent to the circle of friends for spread. The version released by WeChat yesterday also requires us to think about a new distribution mechanism.

Weibo and Zhihu also have their own content distribution mechanisms.

Tik Tok belongs to Toutiao. The content of Toutiao is based on algorithm recommendations and directly distributes traffic. The impact of user interaction on distribution is not direct but indirect.

Through the analysis of others, as well as our own testing and verification, we found that the distribution mechanism of Douyin is a hierarchical traffic pool. Each video will have a small traffic pool, which will be distributed based on the performance of the video in the small traffic pool.

For example, after you release a video, 1,000 people may watch it. After watching it, they will take action on your video: 300 people may delete it without finishing it, the remaining 200 people may like it, 300 people may write comments, 100 people may forward it, and so on.

We speculate that these users who provide feedback will be labeled by the platform, and these labels are based on the users' past behavior.

For example, 80% of the users who like your video are those who like cats and dogs, so your video will be liked by users who like cats and dogs. Next, your video will be placed in another traffic pool. If most of the users who like cats and dogs like your video, then your video may be placed in a higher-level traffic pool for users who like cats and dogs. Keep jumping up like this.

Based on this speculation about the matching of content and tags, if we want the content to be spread more, we must first determine what kind of tags the content will be labeled with, label the content in advance, and then guess the keywords of the tags, search in the Douyin search box, and see how the videos under the keywords perform.

Next, we will observe the content of the videos under this tag, determine what kind of content will be liked by users, and use this to evaluate whether the content we create will be popular.

Based on our understanding of content distribution mechanisms, if we can become the most favorite video of users under a large tag, then the video has a greater chance of becoming popular. Our criterion for judging whether a topic is approved is whether the topic can become the top one under a big label.

3. Douyin’s content operation – content production

After studying the platform, we need to produce content. We have a formula for content production, which applies not only to video production, but also to other forms of new media content production.

As we said before, one of the goals of creating TikTok is to verify whether our content production methodology is applicable to video production, that is, to verify our ability to produce video content.

So how do we demonstrate our abilities? Our team is conducting OKR management (Objectives and Key Results, a set of management tools and methods for defining and tracking goals and their completion status), so we need to determine what our KR value (KR refers to what we must do to achieve this goal) is?

Our KR value is that users like us, so on the Douyin platform, the size of the fan base can best reflect everyone's liking for us. Because following means that they want to watch your videos for a long time, which is a true liking.

Now we want to study how to increase attention. We think it means high-quality content X sufficient quantity, and we get the formula: high attention = high-quality content X sufficient quantity.

High-quality content and sufficient quantity are in a multiplicative relationship, that is, both variables must be good enough to increase attention, and it will not work if either variable is 0.

For example, the content quality of a video is very high and it is a hit, but if an account only has one hit, it will not be able to attract enough users to like it and it will be difficult to expand the fan base. Therefore, it is not enough to have high-quality content but not too much quantity.

Likewise, having enough quantity alone is not enough. Just like what we did in March, everyone in the group posted more than a dozen short videos a day, but the quality was not good enough and the effect was very poor.

1) What kind of content is considered high quality?

What kind of content is considered high-quality? The definition of high-quality content also applies to all new media content. There is also a multiplication formula here: high-quality content = credibility × sense of value.

What is credibility? That is, after consuming the content, users are willing to believe in our content. Credibility consists of four dimensions, all of which come from social psychology research.

The first is authority, which is easy to understand; the second is conformity. When users see that everyone is doing something, they will be willing to believe it; the third is visualization, which means seeing is believing. Users are willing to believe what they see in concrete form, and users should see the results they want; the fourth is cuteness, which means making users feel interesting, like this person and thus like the content.

What is a sense of value? When users consume content, their needs are met, which is a sense of value. There are two dimensions of sense of value.

The first is the sense of involvement. When users are reading content, do they subconsciously feel that the content is relevant to them? The most classic example is when watching a movie. Sometimes you can't finish it because you are not involved in it. For example, when watching "Titanic", you feel involved not because you are Jack or Rose, but because everyone has experienced love.

The sense of involvement is correlated with each person's individual experience. The stronger the correlation and the more people involved, the higher the degree to which content consumers can be involved.

Next is conflict. Conflict does not refer to contradictions, but the impact of content on users. Good content should be like this: after users see the content, their lives may improve; but if they miss the content, their lives will be negatively affected.

The greater the sense of gain from seeing the content and the sense of loss from missing the content, the greater the conflict.

I would like to recommend two books to you. We have summarized a lot of content production methodologies from these books. A book called "Rumors" was introduced into China in 1991. The author is Jean-Noël Kapferet, a Frenchman. He systematically organized rumors from the 1940s and 1950s to the 1970s and 1980s, and summarized why rumors were spread.

When false information becomes a rumor, it actually has value. Why? Because rumors are spread widely and those who spread them believe the information is true.

If you want to dispel rumors and spread true information more widely, you need to learn how rumors spread.

We can learn some methods from the spread mechanism of rumors to pass on our scientific health concepts and knowledge to more people, so that users can gain credibility and sense of value from the content.

Another book is "Story", written by the famous playwright and screenwriting coach Robert McKee. This is a screenwriting book that is regarded as a bible in the film and television industry. There is a very important point in it. It tells us what value a movie can provide if it is made well enough, that is, the so-called sense of value and conflict.

2) How to ensure sufficient quantity?

So how do we ensure sufficient quantities? Two conditions must be met.

First, content can be copied. If I have a great video that gets 30 million views, but it's just one video and there's no way to get more than that, then it's useless.

Second, the cost is low enough. If a video can be copied, but it takes two people three weeks to make a video, then the cost is too high and there is no way to get enough samples.

Therefore, the cost must be low enough and the content must be replicable, so that a sufficient amount of content can be generated, thereby obtaining enough samples, and better optimization and iteration can be carried out to gain more user attention.

4. Create content like a clinical trial

What has been said before is all theoretical content, and what comes next is the more critical step of how to do it.

Before I talk about how to do it, I want to emphasize again the thinking mode of evidence-based medicine and clinical decision-making. That is to say, when we create content, we must design experiments and analyze data like clinical experiments.

What should we do specifically?

First of all, it should be made clear that the theories proposed in the previous article are not conclusions, but hypotheses based on retrospective analysis.

So is the assumption correct? Prospective experiments are needed to verify the hypothesis.

Everyone knows how to analyze the so-called hits. This young lady’s dancing is very popular, and the kid’s somersaults are also very popular. So is it possible that some children do somersaults without exploding? It is possible.

One advantage of Tik Tok is that you can keep scrolling down to follow the content. The content before the somersault is all popular, but the number of likes for the videos further down is in the single digits. So how can we judge that the factor that makes a hit is the somersault?

Therefore, the things summarized based on the hot products, including the experience I am passing on to you now, are just hypotheses and need to be verified through experiments.

1) 4 experimental conditions

What should we do specifically?

First, there must be a control group. Only with a control group can we obtain results that are closer to the truth. Assuming that the reason why the video became popular is because of the young man, then next time don’t let the young man act and replace him with a young lady. Will the result be different?

Second, there must be measurement indicators. Judging the quality of a video is not based on your or my feelings, but is verified through operational data, including likes, reposts, playbacks, etc.

In addition to the data provided by the platform, there is also data from video production. For example, let’s assume that after I adjust my glasses, the shorter my words are, the better. But is shorter really better? We need to quantify the length of time after pushing the glasses, and then design experiments to verify the conclusions and find the optimal duration.

Third, there must be control variables. Let’s still take the example of the time after adjusting the glasses to prove that the time after adjusting the glasses affects the video. Other factors need to remain unchanged, including the selection of similar topics. The only variable is the time after the glasses are pushed, so as to prevent interference from other influencing factors. Through different control variables, we can gradually find out what factors affect the results.

Fourth, there must be a sufficient number of samples. Because there are many factors that need to be tested, sufficient sample support is required. If the sample size is too small, such as only one video per week, then the information that can be obtained will be too little. Therefore, we need to publish 2-3 pieces of content every day, or more than ten to twenty pieces of content a week, for our review and optimization.

2) Two statistical methods

Once we have controls, measurable data indicators, control variables, and enough samples, we can then use clinician methods to perform statistical analysis.

There are two methods provided here: chi-square test and correlation analysis.

The chi-square test is a statistical method, generally used in retrospective analysis, to help us verify whether an element really has an impact on the results.

For example, if I say that a video will go viral if it has this element, then will a video without this element also go viral?

Next, I need to detect, for example, 3 out of 10 videos with this element went viral, and 15 out of 20 videos without this element went viral, so I can make a four-grid table, do statistical analysis, and then draw a conclusion.

After reaching our conclusion, we still need to conduct further tests.

Correlation analysis is used to study what factors affect the results.

For example, do the various button functions of Tik Tok help with the number of views? We can analyze 100 videos and observe the correlation between the number of reposts and views, the number of comments and views, the number of likes and views, and the number of likes of comments and views.

As a result, we found that the number of likes is the best correlation with the number of views, and the number of comments is very weakly correlated with the number of video views. In other words, even if there are a lot of comments, the number of views will not increase significantly. However, if the number of likes on a comment is very high, it will help the video playback volume. Perhaps these are all analyzed through statistical methods.

How many samples should we use to complete the experiment? For a video, is it 10 or 20? There is also basic theory of statistics here.

The probability can be obtained by taking the square root of n and dividing the sample size by one. If 4 consecutive videos produce the same result, then the result is most likely correct.

The theory is a little hard to understand, so I’ll just tell you the conclusion: if a video goes viral, you need to make another similar video immediately. If the result is good, make another one. If three videos in a row are all good, then the pattern of this video will most likely be fixed. On the contrary, if three consecutive bets fail, then there is no need to verify this pattern again.

5. The Successful Journey of Dingxiang Doctor’s Douyin Account

After talking about all the methods, let me now show you what Dr. Dingxiang tried and finally achieved success.

1) Various failed attempts

At the beginning, we found that videos that showed off their skills were very popular, such as children doing a dozen somersaults. So I also started to show off my skills, showing off how I could sew pig's trotters, suture them intradermally, and sew them up after cutting the trotter as if it had not been cut at all; I also showed off my knot-tying skills, and I could tie dozens of knots in a minute.

As a result, these videos all flopped. Why? This is because users know that somersaults are difficult, but they have never experienced sewing pig's trotters or tying knots, and have no idea of ​​the difficulty level, so they do not feel involved after watching.

In addition, we tried to use short videos to tell everyone some life tips, such as how to drink yogurt without licking the lid and how to better use a Band-Aid. However, we found that 15 seconds was not enough time to convey the sense of value.

We also tried sharing little-known facts, but users were not interested.

There are also ways to induce users to click "like". For example, tell users the Guinness record for how many times a person taps on the keyboard in one minute, and let users test how many times they can tap in one minute if they click "like" crazily. The idea is to increase the number of "likes". But users are not stupid. They cancel their "likes" after the test.

Later, a video we made with a PPT became popular, with 5 million views, bringing us our first 100,000 fans. What is this video like? We can look at the title, which is called "There is a 20-year-old girl".

Why did this video go viral? Because it fits the user characteristics of the platform: girls, 20 years old, that is, women born after 1995, and users immediately felt involved. So we thought it was great at the time, how easy it is to make a video with PPT, just put the text directly on the PPT.

It turned out that this success was just luck, as we failed many times after that. Because the method was wrong, PPT is not a very good video format and cannot give full play to the advantages of video. Its content capacity is too small, so it was eliminated.

We also tried short videos about food, but that also failed.

2) Finally succeeded

Finally, we succeeded. How did we do it?

At the beginning, we thought that doctors could introduce health knowledge to users, but Dingxiang Doctor paid special attention to one point, that is, as a doctor, when providing popular science to users, he should not be condescending, but should convey and introduce information with the users on an equal basis.

If there is a doctor who looks like an old professor sitting in the video, the user will feel very depressed. We cannot lecture to users in the video, but we have to perform for them. But how should we perform?

We thought of doing a talk show because we had all received talk show training and could put the talk show in the video. At first, everyone thought it was not feasible because users could not judge the authenticity of the content. Later we thought of the fact that I like to push glasses, and we can tell users that the content before pushing glasses is real, and the content when pushing glasses is a reversal and talk show, so pushing glasses became a replicable form of our video.

For example, we wanted to publish a topic "What are the harms of sleeping with wet hair?", and we needed to think of a talk show joke, because sleeping with wet hair is related to hairstyle. Then we searched for a related one on the Internet, and the content was "Madam, your hairstyle is dangerous." In our team, only one intern who was born after 1995 knew that this meme was related to the ACG culture, the rest of us didn’t know.

But it doesn’t matter if we don’t know. There are over 3 million results for the search on Baidu, so enough young users know the meme. As a result, this video went viral, and some users asked in the comment section what the joke was about. Someone immediately answered the question, and the answer received thousands of likes.

So, it doesn’t matter whether we like the content or not; the key is whether users like it.

Finally, many chicken soup articles say that failure is the mother of success, but I want to say that is not the case. Without a good method, failure will always be failure and will never lead to success. A successful mother should be successful. But success requires continuous iteration of methodology, because sometimes success is just good luck.

Therefore, we need to constantly summarize good methodologies, not to prevent you from making mistakes or failing, but to enable you to fail faster, verify your failures faster, and then achieve 0-1 as quickly as possible, and usher in success after experiencing all the failures in the shortest time.

Author: Operation Research Society

Source: Operation Research Society