Reborn in 2014, getting rich from the World Cup

in the office.

"Mr. Yang."

After Chen Yan came in, he nodded and greeted.

Yang Zheng also nodded and said, "Sit down."

After Chen Yan sat down.

Yang Zheng continued, "I called Mr. Chen here this time because I wanted to discuss Huoduoduo's recommendation algorithm."

Chen Yan was stunned when he heard this: "Mr. Yang also understands recommendation algorithms?"

I can only say that he is worthy of being a science and engineering man.

Speak really straight.

Yang Zheng didn't care and just said, "I definitely don't understand the specific technical stuff, but I know what kind of recommendation algorithm Huoduoduo needs."

Recommendation algorithms are, of course, not only needed by short video platforms.

E-commerce platforms, live streaming platforms, even news and information websites, search engines, etc., all require recommendation algorithms.

The so-called recommendation algorithm is that the platform labels users with various labels based on their previous behaviors.

Then, through these tags, we can portray the user’s profile and recommend content and products that they may be interested in.

A more basic one is that the user has bought a certain product, viewed a certain content, and stayed there for a long time.

After that, we will continue to recommend similar products and similar content to users.

The advanced method is to infer the user’s purpose through the products he/she purchased, and then recommend products that the user may need.

For example, if a user buys a piece of clothing, then he will be pushed pants, shoes, and even socks and underwear.

So many times when I shop online, I will encounter a situation where the platform pushes to me what I want to buy.

The same is true for information content.

Whatever I wanted to watch, he would push it to me.

The more I watch, the more I can’t stop, and I can play with my phone all day long.

This is where recommendation algorithms come into play.

Why is Douyin’s algorithm praised so highly?

Because it's complicated enough.

The algorithms of general platforms simply analyze users' basic behavior.

For example, if you buy clothes, we will push pants, shoes, etc. to you.

And what about Douyin?

It is said that it will label users with hundreds of tags!

What are the consequences?

When you buy clothes, it will be based on the style, price and other information of the clothes you buy.

Infer your purpose of buying clothes, as well as your age, status, etc.

If you buy cheap home clothes.

In addition to pushing similar pants and shoes to you, it will also tentatively push some household gadgets to you.

It positions you as a housewife.

But you have no interest in those household gadgets and just swipe past them.

Then it will change to push other relevant content, continue to label you, and create a profile for you.

Until your portrait takes shape.

Then it may know you better than you know yourself.

Then it will become a normal thing that it will push whatever you think of in your mind.

You even suspect it can read minds.

Under this mind-reading algorithm.

Who wouldn’t be addicted to Douyin?

It really understands me!

How to do this algorithm specifically.

Yang Zheng definitely... doesn't know.

But the boss didn't say that we need to know how to do this algorithm.

He just needs to state his request.

The rest is naturally up to the employees.

Yang Zheng briefly outlined his requirements and concluded:

“I need each user to see a different page, which is what they want to see most.”

This is a core point of Douyin’s algorithm: personalized recommendation.

Chen Yan's brows were furrowed when he heard this, but he didn't respond.

Yang Zheng also frowned slightly: "Can't do it?"

Chen Yan glanced at him and nodded frankly: "It can't be done in a short time."

He paused and said, "To achieve the effect that Mr. Yang mentioned, it is not only a technical problem, but also requires a large amount of data models for deduction and training, which will take a long time."

Yang Zheng nodded in understanding.

It takes time, for sure.

Douyin's algorithm was not developed overnight.

Douyin was launched in September 2016.

However, Zijie Group launched Jintiao as early as 2012 and began to research and train machine algorithms.

It was not until the end of 14 that the user tag cluster streaming computing system was launched, which truly distinguished it from the ordinary algorithms of other platforms and established an algorithmic advantage.

So it takes time, which is within Yang Zheng's expectations.

"I don't need you to do it in a short time, just move in this direction."

Yang Zheng said something.

Chen Yan remained straightforward: "But I don't think e-commerce platforms need such a complex recommendation algorithm."

“We just need to make recommendations based on the user’s purchasing behavior and browsing history.”

"An algorithm that is too complex will place high demands on servers and technical personnel, and the cost will also increase significantly."

Definitely.

It is said that the technical staff responsible for the recommendation algorithm of Douyin is a large team of more than a thousand people!

And the average monthly salary is 3 to 5.

15 salaries per year.

A rough calculation shows that the annual salary for technical personnel alone is as much as 6 million!

Just a little more.

Chen Yan didn't know Douyin.

But he understands the industry salary levels.

So he also knew that it would really cost a lot to implement the algorithm that Yang Zheng mentioned.

Yang Zheng did not refute Chen Yan's words.

Others must know more about this than him.

He simply asked, "Do you think live streaming e-commerce can work using the recommendation methods currently used by major live streaming platforms?"

This time Chen Yan did not respond directly, but fell into deep thought.

What are the recommended methods of major live streaming platforms now?

Both.

The first is the recommended position assigned by the platform itself;

The second is the ranking-style recommendation position based on competitive data.

Chen Yan didn’t think there was anything wrong with this before.

Because everyone has always been like this.

But as someone who works in Internet technology, he certainly knows that just because something has always been the case is not necessarily right.

Internet technology focuses on innovation!

"I know there must be something wrong here, but I can't think of where it is. Could you please clarify it, Mr. Yang?"

Chen Yan paused and asked.

Yang Zheng didn't keep anyone in suspense and said bluntly:

"Entertainment live streaming is free, so users are very tolerant of it."

"It's like eating at a free cafeteria. You eat whatever they have and you won't complain."

"The purpose of e-commerce live streaming is to make users pay."

"It's like if you go to a normal restaurant and you have to pay for the meal, would you accept not being able to order and just eat whatever is served?"

"If this kind of restaurant continues like this, will anyone still be willing to go there?"

Chen Yan nodded suddenly, "I understand what Mr. Yang means."

Zhou Wei, who was standing by, also said with a sense of understanding: "Actually, the recommendation method of entertainment live broadcast is not good either."

"Nowadays, traffic on major live streaming platforms is basically concentrated on big anchors."

“It’s almost impossible for new anchors to make a name for themselves without the support of a platform or guild.”

"The platform has gradually become stagnant, and it's difficult to make new breakthroughs."

Yang Zheng nodded and said, "So we need a more refined personalized recommendation algorithm."

"understood."

Chen Yan responded again without making any rebuttal remarks.

Yang Zheng looked at him and asked, "Mr. Chen, do you know anything about this?"

Chen Yan also looked at him with a look of joy on his face: "Boss Yang, are you also paying attention to the short video field?"

The two looked at each other and laughed.