Google's largest source of revenue is advertising, and the largest source of advertising revenue is search recommendations, while the core of search recommendations is the ranking of relevant web pages.
The reason why we engage in bidding ranking is because the user's attention is limited, just like wealth, the smallest number of top pages will capture more than 95% of the user's attention.
The number of searches an average user can do and the number of pages they read every day are limited. How to push the most effective ads among the limited number of pages is the core issue of the recommended advertising algorithm.
All advertisers know that at least 70% to 80% of the advertising money they pay is wasted.
But the problem is that no one knows which 20% to 30% plays a role.
The amount of money that should be paid cannot be less.
For Google, how to better plan and distribute these display interfaces is a cost-effective way to generate revenue.
To take an extreme example, if Google could push ads to everyone, they would be exactly what they need and they would click on them.
Then the theoretical upper limit of recommended advertising is achieved, and all the money is earned with the smallest display.
Unfortunately, most manufacturers currently cannot achieve this ratio of 1%, and even Google is only around 2%.
The purpose of the "Mobile Optimization Strategy" and "Fred's Algorithm" provided by Meng Fanqi is to further eliminate those web pages that appear inconvenient on mobile phones, as well as various traffic-draining spam websites.
It's a process of elimination. As long as I eliminate more junk content, the remaining content will be more likely to be clicked.
Combining it with the language AI model is additive, providing more relevant results by better understanding the user's search intentions.
In this way, the interface users see after each search is more relevant and there are fewer junkyard websites, which in turn results in more ad clicks.
Based on the experience of the original timeline, Meng Fanqi estimated that after these three axes were swung, the improvement in effect should start at 20%.
The more Xi Xiaohu listened to the details of the idea, the more interesting it became. He originally planned to chat for only a few minutes, but before he knew it, nearly an hour had passed.
The two of them did not find a meeting room specifically, but found a sofa near the reception desk and chatted.
Many Google employees passing by at work were quite surprised after seeing it.
Xi Xiaohu's rank is still quite high, and many Google Shanghai employees rarely have direct contact with him.
A few interns have never even met the boss.
"Oh! Zhou Huan, Chen Hui, just in time. Come here." Xi Xiaohu caught a glimpse of two passing Google employees from the corner of his eye and called them over directly.
Zhou Huan and Chen Hui had just returned from making coffee when they saw the senior leader chatting with a young man who looked like a student. They were very curious.
Since Baidu announced the product immediately, Robin Li personally conducted two publicity campaigns, and many people knew Meng Fanqi.
On Google's side, Jeff and Hinton did not affect too many unrelated people except for their introduction inside Google Brain.
Therefore, Zhou Huan and Chen Hui, who were focused on the advertising department, didn't know Meng Fanqi's identity at the moment. Seeing that he was so young and could chat and laugh with senior leaders, they thought he was an intern who came in through connections.
"Leader! Don't worry, the two of us will definitely take good care of him!" Xi Xiaohu had just introduced a name, and Zhou Huan took the lead in expressing his position without waiting for Xi Xiaohu to issue an order.
"Okay, okay, please prepare local accounts, permissions and tools. He is mainly working on our search and recommendation section here during this period." Zhou Huan and Chen Hui are both experienced engineers in the group. , is technically sound and knows how to lead people.
Xi Xiaohu felt that although Meng Fanqi's ideas were quite thoughtful and outstanding, this was still his first time writing core content in the advertising algorithm department.
Apart from anything else, it takes a lot of time to become familiar with these existing things.
After all, there is a huge difference between understanding the principles and understanding the code so that you can develop and create on your own.
He saw that the two men had misunderstood Meng Fanqi's identity, but he had no intention of reminding them.
First, because Meng Fanqi is indeed like an intern and needs to be familiar with a lot of content.
The second reason is that Zhou Huan and Chen Hui usually rely on their own cleverness to cause him little trouble.
The two of them had made many meritorious deeds before, so he couldn't say much.
"Hmph! This time, I'll let you see what it means to be alone." Seeing Meng Fanqi follow the two of them to the work area, Xi Xiaohu grinned and sneered.
"How old is my little brother? Where did you read it?" Zhou Huan started to ask on the way to the workstation.
"I'm in my early twenties and will be a senior in college next year." Meng Fanqi told the truth.
"Okay!" Zhou Huan cheered loudly in his heart. When he saw Meng Fanqi chatting with the leader, he had already made a private bet with Chen Hui.
The bet was whether Meng Fanqi was an intern, and now Zhou Huan thought he had won a big victory.
While Zhou Huan continued to walk forward with the two of them, he put his left hand behind his back and moved his four fingers, indicating that Chen Hui was willing to admit defeat.
Chen Hui rolled his eyes, took out his mobile phone and transferred a 200 yuan red envelope to Zhou Huan.
"Has your internal account been processed?" Zhou Huan saw that Meng Fanqi logged into Google's internal account with ease shortly after he sat down, and suddenly felt something bad.
"Yes, now I mainly need you to briefly introduce to me the process of launching the advertising algorithm. I need to obtain permissions for all relevant data and scripts in this process."
Meng Fanqi probably knows this testing process. It is usually tested offline on different scales first, and then partially online. After the test is normal, it will be promoted to the whole situation.
Backups will be kept at each stage so that problems can be rolled back at any time.
"I'm super, Google Brain?" When Zhou Huan assigned permissions, he looked at Meng Fanqi's email suffix and was shocked.
Although Google Brain at this time was not the giant team it would become later, it was still a somewhat famous department due to its leadership by star engineer Jeff.
Chen Hui raised the corners of his mouth and raised his chin at Zhou Huan proudly. According to the rules, if you don't understand the situation first and ask for money, you will have to pay back double the amount.
Zhou Huan shouted that he had made a mistake, so he reluctantly withdrew 200 yuan and transferred 400 yuan.
Meng Fanqi glanced slightly sideways, probably understanding what these two guys were doing.
Capitalist companies are just not wolf-like enough and spend all their time focusing on the things that are available and not available.
After Meng Fanqi took over the authority, the first thing he wanted to implement was Fred's algorithm.
The part involving AI requires a period of preparation and training, and the mobile optimization method may not be obvious enough in short-term testing, and the effect will be better later in 2014.
But the Fred algorithm’s ability to screen spam websites on the Internet is amazing.
After this algorithm is launched, nearly 70% of current websites may be greatly affected.
