APTnovel Chapter 487 The Game of Speed and Shake
Chapter 487 The Game of Speed and Shake
Lu Ran received feedback from Kuaishou on a Wednesday afternoon.
The call was made personally by Zhang Yifan, the founder of Kuaishou and Douyin. His voice sounded younger than Lu Ran had imagined, and he spoke at a moderate pace, with the composure characteristic of someone who has been working on products for many years.
He said he had read our proposal and found it interesting. He suggested we visit Hangzhou sometime and he would treat us to a meal.
Lu Ran agreed.
After hanging up the phone, he sat in his office for a while, pondering Zhang Yifan's words over and over again.
The phrase "something's interesting" might be a polite expression elsewhere, but when used by the founders of Kuaishou and Douyin, it carries a different weight.
With the size of Kuaishou and Douyin today, Zhang Yifan receives at least dozens of cooperation proposals every day, but only a handful that he personally calls to say "this is interesting" over the course of a year.
He picked up his phone and sent a message to Zhou Mingzhe and Chen Mo: "Going to Hangzhou next Monday to meet someone from Kuaishou."
Zhou Mingzhe replied with an "okay," and Chen Mo asked another question: "Who saw him? Zhang Yifan himself?"
"Yes, he made the call himself."
Chen Mo sent an ellipsis and then said, "Then this trip is more important than I thought."
The three met at Hongqiao Railway Station at 8:00 a.m. on Monday.
Lu Ran wore a light gray jacket, Zhou Mingzhe was still wearing the same dark blue jacket he had worn for a long time, and Chen Mo was wearing a suit, but without a tie, with one button undone at the collar.
The high-speed train takes forty minutes to reach Hangzhou. When I got off the train, a car from Kuaishou was already waiting for me. The driver was a young man wearing Kuaishou's uniform, and he smiled, revealing a set of white teeth.
After driving for twenty minutes, we turned into an area that didn't quite look like an office area.
Camphor trees line both sides of the road, their branches and leaves intertwined, filtering the sunlight into a shower of golden fragments.
Several low-rise gray buildings are scattered among the trees, the tallest being no more than five or six stories high. The exterior walls are covered with light-colored stone, and the window frames are made of dark gray aluminum alloy. The entire park is quiet and peaceful, not like the headquarters of a company with hundreds of millions of daily active users, but more like a university research institute.
Zhang Yifan waited at the entrance of the park.
He wasn't tall, wore round-framed glasses, a plain white T-shirt and dark trousers, and sneakers. He looked like a programmer who had been sitting at his desk for a long time.
He walked over and shook hands with Lu Ran, his palm warm and dry. He smiled and said, "It's good that you're here. Don't rush to chat, let me show you around."
Lu Ran nodded in response.
Zhang Yifan led three people to start walking from the east side of the park.
First, you pass a low white building with floor-to-ceiling windows on the first floor. Inside, you can see a surveillance wall made up of dozens of screens of different sizes, with various colored data streams jumping on the screens.
Zhang Yifan said this is a real-time data monitoring center where behavioral data from all users across the country are aggregated. Tens of thousands of new data entries come in every second, and the system automatically performs classification and labeling.
Lu Ran stood outside the glass and watched for a while.
The data streams jump so fast, with colors and numbers constantly refreshing on the screen, that the human eye simply can't keep up.
But behind that data is a recommendation algorithm that has been running for many years. Every user's every swipe is broken down into feature vectors of dozens of dimensions by the system, and then fed into the model to be labeled, grouped, and predict the next action.
Zhang Yifan didn't let him look for too long, then turned around and led them deeper into the park.
He stopped as he passed a three-story building and pointed to a row of south-facing windows on the second floor: "That's where our recommendation algorithm team works. The core team of Kuaishou and Douyin is just a few dozen people. What they do every day is look at data, adjust parameters, and run experiments. Sometimes I pass by at 11 pm and the lights are still on."
Zhou Mingzhe chimed in, "I heard that your recommendation algorithm team has always kept its size very tight and doesn't expand easily."
"Yes. Having more people isn't enough; the key is whether those few can understand what the data is saying. With algorithms, third-rate people run the model, second-rate people tune the parameters, and first-rate people understand the data. We only keep those who can understand the data; the rest have gone to other groups."
Chen Mo stood next to Zhou Mingzhe without saying a word, but Lu Ran noticed that he hadn't taken out his phone much since entering the park, which was unusual for him.
Zhang Yifan then showed them several more places.
In a content moderation department, hundreds of monitors are lined up in rows, with various videos and images scrolling across the screens. The moderators wear headphones and their fingers fly across the keyboards.
A tech team working on audio and video processing had a "No Photography" sign posted at the entrance. Zhang Yifan didn't take them inside; he simply pointed to a closed door in the hallway and said, "The stuff inside won't be released until next year; you can't see it now."
Finally, they stopped in front of a small building at the very back of the park.
The building is not tall, only two stories, with light gray exterior walls and a small wooden sign hanging at the entrance that reads "User Behavior Lab".
Zhang Yifan pushed open the door and led them inside. Inside was a small exhibition hall with dozens of screens hanging on the walls, each displaying live footage of different users using Kuaishou.
It's not a real screen, it's a simulation. Every user icon is swiping, liking, commenting, and sharing, and the frequency and speed of these actions are all different, like dozens of virtual people are simultaneously scrolling through their phones.
Lu Ran walked around the exhibition hall and saw a screen in the corner displaying a real-time curve of the "recommendation accuracy".
The curve is very stable, consistently remaining above 90%.
He stared at the line for a while, mentally calculating the commercial value corresponding to that accuracy rate.
Zhang Yifan, standing behind him, said unhurriedly, "It took us five years to achieve this accuracy rate. For the first three years, it hovered around 70%, then we changed the model structure and broke through to 80%. It took another two years to break through to 90. Now, every percentage point increase requires a whole quarter's worth of work, but every percentage point increase adds one and a half minutes of user time."
Lu Ran turned to look at him: "So you've been pushing things up the ladder."
"If you don't keep moving forward, you'll fall behind," Zhang Yifan said. "There's no 'maintaining the status quo' option in the internet industry. If you don't move forward, others will. If you stop for three months, you won't even know where you are when you look back."
After the tour, Zhang Yifan led them to a teahouse on the east side of the park. The teahouse was small, with seven or eight tables, and the decor was more refined than the buildings outside, with wooden tables and chairs, white porcelain tea sets, and a waist-high green plant in the corner. The waiter brought over a pot of Longjing tea; the tea was clear and had a faint bean aroma.
Zhang Yifan poured a cup for each of the four people, put down the teapot, leaned back in his chair, as if he had finally finished the warm-up and was ready to get down to business.
"Mr. Lu," Zhang Yifan began, his tone slightly more formal than before, "I've reviewed the cooperation proposal you sent earlier. The application cases of AI technology in the e-commerce and food delivery industries are clearly described, and the data and logic are sound. However, the content of your proposal for Kuaishou and Douyin is somewhat different from that for JD.com and Ele.me. Your proposal to them was 'I'll help you optimize your existing business,' while your proposal to Kuaishou and Douyin is 'Our technology can help you further increase user time spent.' Both directions sound reasonable, but I need to make one thing clear first—Kuaishou and Douyin's current user time spent is already very high, so high that every second of increase requires a huge cost. The room for improvement you can bring us may be smaller than you anticipate."
He spoke very politely, but Lu Ran understood.
The subtext is that Kuaishou and Douyin are already strong enough, and the incremental growth that TUTU can bring is limited. Don't overestimate your own resources.
Lu Ran picked up his teacup, took a sip, put it down, and then spoke: "Mr. Zhang, I agree with your assessment. Kuaishou and Douyin have indeed reached a very high position. If they go any higher, the marginal cost will only increase. But I'm not here today to sell you the 'help you increase user time' solution. I'm here to discuss something else—the traffic ecosystem."
Zhang Yifan didn't respond, but picked up his teacup, took a sip, and gestured for him to continue.
Lu Ran leaned forward slightly: "TUTU has game products, an esports system, and a user community. Kuaishou and Douyin have the largest short video traffic pool in China. Putting these two together naturally amplifies each other. If a user plays a great game on TUTU, they need a place to share it with more people. If a Kuaishou or Douyin user sees an exciting game video, they might become interested in downloading that game. It's a closed loop, more efficient than any individual traffic purchase."
After listening, Zhang Yifan didn't respond immediately. He glanced down at the teacup in front of him, thought for a moment, and then said, "I agree with the direction. But how do you plan to implement it? If it's just a content collaboration, a framework agreement is enough; you don't need to come in person."
Lu Ran waited for him to finish speaking before continuing, "Content cooperation is just the surface. What I want to discuss today is something deeper. TUTU and Kuaishou/Doudou can collaborate on two levels—the first is content sharing: TUTU will open up its live streaming rights for its esports events to Kuaishou/Doudou, and Kuaishou/Doudou will prioritize TUTU's game content recommendations. The second is technological cooperation: TUTU's AI engine can be integrated into Kuaishou/Doudou's recommendation system, increasing recommendation accuracy by one to two percentage points without increasing server costs. In exchange, Kuaishou/Doudou will provide TUTU with a stable entry point within its app, similar to Ele.me's approach."
Zhang Yifan tapped his fingers lightly on the table twice, at a moderate pace, as if he were mentally going through every step of the cooperation process.
"One to two percentage points," he repeated the figure, "how confident are you?"
"Seventy percent. Whether it can be achieved, and to what extent, depends on your actual data. I need to run a simulation test before I can give you a precise number. But I'm very confident in the direction—the AI engine I have is different from the underlying logic of your existing recommendation system. It understands user behavior from a different perspective. The two sets of logic combined are more effective than running one set alone."
Zhang Yifan leaned back in his chair, staring at Lu Ran for several seconds before saying something that made Lu Ran's heart sink slightly: "I understand the direction you're heading. But the share swap ratio in your proposal is a bit vague. I roughly guess you mean to exchange TUTU's shares for Kuaishou's shares. So, my question is—how much are you planning to offer?"
The air in the tea room fell silent for a moment.
Zhou Mingzhe paused for a moment, his hand holding the teacup still.
Chen Mo's gaze moved from Zhang Yifan's face to Lu Ran's face, and then back again.
Lu Ran knew that this was the real topic of discussion today.
The preceding visits, exhibitions, and discussions about technology and the ecosystem are all paving the way.
Ultimately, it all boils down to this one question: How much are you willing to give up in exchange?
He put down his teacup, looked at Zhang Yifan, and said a number.
After listening, Zhang Yifan's expression remained unchanged. He picked up the teapot, poured a cup for each of them, said, "I understand, let's eat first," and then stood up and walked towards the door.
When Lu Ran stood up, he knew in his heart that this meal was just the beginning.
Kuaishou is different from JD.com and Ele.me. It's not a second-place company that needs saving; it's the industry leader itself.
When discussing cooperation with the top leader, you can't follow the same approach as the second-in-command.
...