In early December 2016, under the auspices of Andy, Yuandian Technology's A-round financing was officially completed, with a post-financing valuation of up to US$8 billion.
In this round of financing, Origin Technology invested a total of 15% of its equity and raised US$1.2 billion. Sequoia, CEFC, CICC, Temasek, Shengxuan... a total of 7 large investment funds participated in the financing.
With a valuation of US$8 billion and a financing amount of US$1.2 billion, Yuandian Technology began to gain a reputation in China and became a well-known unicorn company.
At the end of the A round of financing, in addition to the 15% of the additional equity issued through financing, Li Mu also took out 10% of the equity to establish an employee stock ownership platform, which serves as the option pool of Yuandian Technology to attract and retain more talents.
In this way, the equity structure of Origin Technology is as follows:
Li Mu holds 75% of the shares and is undoubtedly the largest shareholder; the employee stock ownership platform holds 10%; 7 investment institutions hold 15%.
As for Origin Technology, which was founded just over a year ago, its valuation is as high as US$8 billion. Countless people are amazed and talking about it.
But this is just the beginning. As the country vigorously promotes speed increases and fee reductions, traffic becomes cheaper and cheaper, and short video platforms will explode.
With the receipt of US$1.2 billion, Origin Technology's cash flow has once again become abundant. With sufficient ammunition, a new round of cash-burning plans has been launched again.
…
In the blink of an eye, another few months have passed, and the time has entered 2017. With abundant cash support, Origin Technology has developed rapidly during this period.
Its two softwares, Toutiao's total users are as high as 450 million, and Douyin's users have just exceeded 180 million, making it the largest comprehensive media platform and short video platform in the country.
During this period, Happy Five Sisters Online Media Company also developed rapidly and launched a round of financing. Li Mu and Andy jointly invested 50 million, with a post-investment valuation of 200 million.
Suddenly, Fan Shengmei, Qiu Yingying, and Guan Ju'er instantly became rich little women with a net worth of tens of millions.
Qu Xiaoxiao was especially happy. After the Qu family fell, she stood up again with hard work and opportunities. In the future, she might not be able to start from scratch like her father.
…
As Douyin users officially exceeded 200 million, some problems began to appear. Based on real-world experience as a reference, Li Mu began to further optimize Douyin's algorithm.
The current algorithm mechanism of Douyin can be summed up in one sentence: "You are browsing Douyin, and Douyin is also staring at you."
We use the user's browsing traces, preferences, etc. to create a portrait of the user, and then use a set of mechanisms to finally push the content that the user likes to watch, making the user addicted to it. Douyin's algorithm is actually a complex "for you" algorithm
The system focuses on and distinguishes different groups of people.
For example, for video publishers, when a video is successfully released, this algorithm system will first automatically allocate an initial playback volume of 0~1k, and the satisfaction of these users after watching, "comprehensive clicks, completion rate, likes"
Volume, comment volume and other indicators" will determine whether the work can enter the next-level traffic pool to obtain more traffic.
If the user is satisfied, then the second wave of traffic pool (1k ~ 5k playback volume) will be given, followed by the third wave of traffic pool (1w ~ 10w playback volume), and so on.
The videos launched in this way are high-quality videos that users really like to see. Under this algorithm mechanism, on the one hand, it will encourage video creators to continuously create better videos. On the other hand, high-quality videos can also attract more and more people.
users and traffic.
Currently, Yuandian Technology has different traffic pools for accounts with different weights. The main evaluation criteria are: number of likes, number of comments, number of forwards and completion rate.
Volume of views > Volume of likes > Volume of comments > Volume of retweets.
At the user level, the design of this mechanism is much more complex. According to the different preferences of each user, the system can rank videos through a variety of factors to recommend content.
In fact, from the moment you register an account, your series of user behaviors are already continuously feeding data to the "for you" algorithm.
These user behaviors basically cover all aspects: such as account settings and device models, regional settings, device models, etc.
Categories of interest selected during registration: news, travel, entertainment, etc.; user usage scenarios: user's geographical location, video viewing scenarios, etc.; published video information: topic tags, video content, etc.; various user interaction information, clicks
, like, collect, comment, and forward the content.
Finally, after comprehensive weighing and continuous revision, the user can draw an increasingly accurate self-portrait. Sometimes the software understands you better than you do.
Although the mechanism itself seems complicated, the entire recommendation process is completed in the background "moisturizing things silently".
Users don’t need to do anything. They only need to click on the video or click a few likes. Douyin’s background can gradually determine your preferences. Moreover, the algorithm recommendation is tracked and updated in real time. The more traces of your browsing, the better the software will understand you.
However, this current algorithm seems to be advanced, and it is indeed very advanced. It is also the best in the world, but there are still many problems.
Li Mu discovered some signs in advance based on real-world experience.
For example, it is not obvious yet, but in two years more and more people will find that this algorithm mechanism will form a huge information cocoon by recommending based on user preferences, segmenting various types of users and allowing them to immerse themselves in their own
I cannot extricate myself from the small circle and cannot be exposed to more external information.
Since the number of Douyin users exceeded 200 million, the information cocoon effect has begun to appear. At present, not many people pay attention to this issue. If no changes are made, it will definitely cause greater controversy in the future.
In response to this, based on real-world Byte experience, Li Mu has recently begun to carry out large-scale optimization of Douyin’s current algorithm.
It will be more humane and flexible, allowing the software to not only automatically screen out errors and deceptive browsing traces, but also not to push a large number of homogeneous videos in a "retarded" manner.
From time to time, some related, similar or even opposite videos will be pushed to test, induce and stimulate users, so that users can escape from the information cocoon and have more interests and concerns.
In this way, you can ensure that high-quality videos continue to flow out and are continuously pushed to people who are truly interested.
Of course, this is not just an algorithm problem, the underlying logic behind it is actually the decentralization of traffic distribution.
As long as your video is high-quality and interesting enough, even if you are a newly registered novice account, your video can get tens of millions of clicks; conversely, even if you are a big V with millions of fans, but your video is mediocre and boring, it will
It’s hard to win by lying down.
At present, Li Mu has recruited and poached hundreds of algorithm experts to build and optimize this algorithm system. When it is hatched, in the next few years, Douyin will be the leader in the short video field, even on a global scale.
It is a unique existence.
As long as nothing goes wrong in other aspects, the future will be fine!