Tik Tok restructures its recommendation algorithm, 3 major changes need to be noted!

Recently, Douyin restructured its recommendation algorithm, and this change has far-reaching implications for creators who want to gain more exposure on the platform. This article will explore the three major changes in Douyin's recommendation algorithm and what these changes mean for content creators, helping creators better adapt to the new algorithm and seize traffic opportunities.

It’s been a long time since I talked about Tik Tok’s algorithm. Recently, many fans and friends have been asking about it, so today I’m going to write an article to talk about the latest changes.

I believe that those who work on Douyin should have clearly felt the difference in traffic in the past six months. It is difficult to have explosive traffic according to the traditional shallow content approach. All this comes from the reconstruction of Douyin's recommendation algorithm. The specific changes are as follows. Those who need it can pay attention to it:

1. Diversification of recommendation algorithms

Douyin’s previous recommendation algorithm was basically evolved from Toutiao. The content creator side would identify the creator’s field and extract key tags of the content.

On the user side, we will recommend a few initial interest-mining contents at the beginning, judge the user's preferences through interactive feedback, and then slowly tag the user with corresponding interest tags.

Finally, match the content tags with user interest tags to achieve interest distribution and recommend content that users like to watch.

However, this simple interest-based distribution can no longer meet current user needs. As user needs become diversified, recommendation algorithms also need to become diversified.

At present, Douyin has been implementing a diversified combination recommendation algorithm, adding other elements based on the original interest distribution, such as interest intersection, viewing frequency, etc.

Interest intersection is easy to understand. For example, if a user likes to watch fishing videos and is a fishing enthusiast, he or she may also like to watch outdoor videos about catching fish in the sea or catching fish in the water. There will be some overlap between the two. Of course, it is not limited to the intersection of content points, such as background music and video special effects.

The other is the same frequency of viewing and blogging, which means that the audience and the blogger have the same interests and hobbies. For example, if you are a beauty and fashion blogger, the system will recognize that you are a beauty blogger, fashion blogger, etc., and have many interests and hobbies. Then users with similar interests and hobbies may not have seen your video based on interest distribution before, but now they may be recommended to them based on the same frequency of viewing and blogging.

In addition, the new algorithm also has some new changes in the assessment of work data. For example, the new algorithm pays more attention to the user's playback time, so work hooks and user retention must be worked on extra. The hook is to attract users, and "slowly chewing" content is what retains users.

2. Merging content and e-commerce traffic

In the past two years, Douyin's e-commerce has developed very rapidly, starting from Luo Yonghao's "Make Friends". However, the rise of e-commerce business still requires corresponding e-commerce traffic to support it.

As a UGC platform and a community focusing on content, Douyin's sudden introduction of live streaming or short video sales would cause some users to be disgusted, so Douyin previously divided its traffic into content traffic and e-commerce traffic.

As the name suggests, content traffic focuses on content, and e-commerce traffic focuses on bringing goods. Obviously, content traffic is greater than e-commerce traffic.

But with the rapid development of Douyin e-commerce, there are more and more merchants, so Douyin's e-commerce traffic is not enough to support it. It is like there are more online car-hailing drivers on the Didi platform, but not so many people taking taxis. What should we do?

There are two methods, one is to tap traffic and the other is to cultivate traffic.

The combination of e-commerce traffic and content traffic just meets these two methods. On the one hand, it can tap into e-commerce users from the content traffic, and on the other hand, it can drive conversions through content, thereby cultivating users.

Some time ago, the president of Douyin e-commerce also officially announced the reconstruction of Douyin e-commerce traffic distribution mechanism, encouraging content creators to achieve some conversions and merchants to create some content, and cited Dong Yuhui as a typical case.

3. The search and push linkage mechanism is quietly launched

In the past, Watangren has talked about the importance of Douyin search traffic. On the one hand, search traffic is very accurate, because users search with needs, and it may be more accurate than recommended traffic.

On the other hand, the search traffic of Douyin will become larger and larger. At least it has not reached the ceiling yet, because the content on the short video platform is very rich, and it is an application that occupies a relatively long time of users. You can search directly when you need it. The search results are sometimes better than those of a certain search. One application has closed the loop of search traffic (WeChat is also doing this, you can check out WeChat Ask or the new track for increasing fans!).

So when many people are still studying the recommended traffic of Douyin, others have already been making a lot of money by planning the search traffic of Douyin.

Based on the development of Douyin search, Douyin has quietly launched a search-push linkage mechanism, claiming that good content can be seen through both recommendation and search paths.

So where is the search-push linkage mechanism specifically manifested? Actually, there are two aspects: one is search and the other is push.

One is the user's regular search, which is when you open the magnifying glass in the upper right corner of Douyin and enter a keyword to search. When you search for a keyword, there will be search results at that time, and the content you swipe to next will also include related content you searched for. In other words, when you search for a keyword, there will be related recommended content, commonly known as search and see.

The other is content recommendation search, commonly known as search after watching. That is, you are watching a video, and there are some points in the video that you don’t know. At this time, you will want to search for further information. In order to facilitate your search, Douyin directly shows you the popular search words of users after watching the video at the top of the comment area. You can search by clicking.

The above is the performance on the user side. On the creator side, Douyin also provides you with a good method to help you better perform "search and push linkage".

Now, in the data analysis of Tik Tok works, there is an additional search keyword, which shows the words that users often search for after watching your work. You can further create content based on these words.

In addition, your content may also appear in users' regular searches, and you can also check how much traffic this type of search brings you. You can also further optimize your work keywords based on the data.

How should creators respond to the reconstruction of TikTok’s recommendation algorithm?

Regarding the reconstruction of Douyin's recommendation algorithm, we think that those who create deep content will benefit, while those who create shallow content will be disadvantaged. So how should these creators deal with it? Here are a few suggestions:

1. Be able to create high-quality content

Douyin has always emphasized high-quality content. Of course, it is not the only one to do so. All UGC platforms do so. Although their definitions of high-quality content are different, I think it is still the three points summarized by me before, namely, touching, useful and interesting.

2. Don’t piece together content

Tik Tok's algorithm for identifying reposted content has reached a new level. It no longer uses the MD5 code of a video to determine whether it has been reposted. More elements have been incorporated into the algorithm, including script content, video stickers, mosaics, and so on.

For example, the technology for converting content into text is already very mature. Even if you don’t copy it directly, the remake may be judged as homogeneous content.

3. Content optimization

Based on the evaluation mechanism of Tik Tok's new algorithm for works, you may need to set more demand points for your work. On the one hand, this operation can make it differentiated, and on the other hand, it can attract users with different interests.

The above are some of the views of Watangren on Douyin’s latest recommendation algorithm. I hope it will be helpful to you.