Qiu Guangcheng and Wu Cheng couldn't help but nodded, and the two of them were silent, listening to Fang Hong's discussion: "These three ways are from the perspective of the information recipient, so think about it from another perspective, when you are an information creator, what will it be like to look at the three types of information distribution methods from the perspective of it?" ”

Fang Hong continued: "The first attention mode, for example, if you want to do an official account, the most important thing is that you have to increase your fans, if you enter the game late, it is difficult for you to make a popular big number, this model is very difficult for creators who enter the game later, and this is also the dilemma that the official account must face in the future." ”

Qiu Guangcheng was surprised when he heard this, the current WeChat public account is very popular, and its success is being studied by peers from all walks of life.

Fang Hong looked around at the two and said: "It's the same with search, when you're an information creator, you want your target audience to search for your content, you have to guess what users are searching for? What do users search for? You need to do a lot of search engine optimization, the same logic, if you haven't been a webmaster in the early days, and you haven't entered the search engine news source, your content can only be relied on at the back station, which is also very unfavorable to the late entrants, so Baidu Baijia can't become a climate. ”

"And the third algorithm recommendation mode, after you publish the content, as long as your target audience is accurate, the algorithm will automatically identify your content and recommend your content to people who may like it, then simply put, you only need to publish the content and wait for him to become a hit."

At this time, Wu Cheng immediately said: "Then the question is, after Douyin is launched, a large number of short videos will be uploaded every day, maybe tens of thousands, how to make your video become one in ten thousand or even one in 100,000 recommended by the algorithm?" ”

Fang Hong couldn't help but smile when he heard this, and after a while, he said with a smile: "You don't have to worry about the problem of algorithm recommendation, when you led the group to go to buy Toutiao, I had already greeted Chen Yu, and I will provide you with technical support for algorithm recommendation, this guy's technical development ability is stronger than his investment in stocks, maybe it has been done at this time, just waiting for you to get it." ”

"Generally speaking, it is the four words 'judging things and recognizing people'. Breaking objects, that is, what does the algorithm need to identify your content? Knowing people means that the algorithm needs to identify the person who is traffic content on the platform, what kind of preferences he has, that is, how much user portrait do you master, which we are in the entire domestic Internet system, who else can I be? ”

Qunxing Capital can be said to have mastered the most complete user portraits, because the data of its major subsidiaries are under the adjustment of the parent company, and all data links are opened.

It is no exaggeration to say that based on the big data at its disposal, Qunxing Capital knows a user even better than he does.

This is where the power of big data lies, and it is also where the unparalleled value of big data lies, and it has super strong monetization capabilities.

Based on the algorithm technical support provided by Chen Yu, knowing that a person likes to play basketball, and the algorithm judges that there is an article about basketball, then it will recommend this article on basketball to people who like to play basketball.

The logic of algorithm recommendation is actually as simple as that.

Fang Hong said methodically: "There are two main ways for the algorithm I asked Chen Yu to develop. The first is a content-based recommendation algorithm, what is a content-based recommendation algorithm? For example, after a person enters Douyin, he likes an article A about playing basketball, and the algorithm finds that article B is also related to playing basketball, then the algorithm will recommend article B to him, which is a recommendation based on content. ”

"The second is algorithmic recommendation based on collaborative filtering, the word collaborative filtering may be obscure, in other words, it is algorithmic recommendation based on user behavior, and this algorithmic recommendation based on user behavior can be subdivided into two categories, one is the user behavior algorithm recommendation based on items, and the other is the algorithm recommendation based on the user's user behavior."

Fang Hong further deconstructed the algorithm recommendation of these two types of subdivisions for Wu Cheng: "Item-based user behavior algorithm recommendation. For example, if you are user A, you like item 1 in the system, and the system recognizes that the user who likes item 1 also includes user B, user C, and user D, and then the system also recognizes that user B, user C, and user D also like item 3 at the same time. In terms of the user's liking of items, can we guess that item 1 and item 3 are similar? ”

So the system recommends item 3 to user A. This set of logic is based on the recommendation from the perspective of liking the item, so it is called the user's behavior algorithm recommendation based on the item, because its recommendation logic is the user's liking of the item. ”

"The second kind of user-based user behavior algorithm recommendation, the logic is the same as the first, but the difference is that people and items are reversed, in this user behavior recommendation method, the target of the system recognition is no longer the object but the person, based on the user behavior and divide the user into different categories, so as to carry out similar recommendations to them."

At the end, Wu Cheng couldn't help nodding: "I understand, this algorithm recommendation model is really powerful, supplemented by the support of big data user portraits owned by the parent company, I have a hunch that Douyin will definitely be able to become a phenomenal hit on this track." ”

Wu Cheng immediately saved the recordings on his mobile phone and encrypted them, which is the essence of Douyin's development, and it can be called a big boss to give a set of "nanny-level" development strategies, just follow this idea to do it, at most do the optimization of the details, it is not an exaggeration to say that it is a lie to win.

Fang Hong said neatly: "Just follow this line of thought, Douyin will become a hit." ”

After a while, Qiu Guangcheng, who basically did not speak, said: "In other words, after the founder of Toutiao and his team left, what will Mr. Zhang do?" He took a lot of money, and there was no non-compete agreement. ”

Wu Cheng listened thoughtfully, and immediately looked at Fang Hong puzzled: "This is also a doubt in my heart, why did you specifically instruct you when you went to acquire Toutiao, if Mr. Zhang's team left and did not sign a non-compete agreement with him?" ”

After all, Mr. Zhang is the founder of Toutiao, and he also understands the potential of this track, and he must also know how to play, so it stands to reason that he has to sign a non-compete agreement or something, restricting him from entering this track for a period of time, preventing him from becoming a potential threat when he gets the money and rises again.

Hearing this, Fang Hong said with a smile: "Two reasons, if you don't sign the dedication agreement, Mr. Zhang will sell more neatly, and the other reason, I just want him to re-enter the game, if he signs a non-compete agreement, how can he re-enter this track?" ”

Qiu Guangcheng and Wu Cheng looked at each other, and it was even more difficult to understand when they heard the words of the big boss, but from these words, both of them heard that the big boss obviously had a deeper backhand layout, and the two of them were also curious.

……

(End of chapter)