Return to site

Interview with Founder and CEO of DiDi, Cheng Wei. Part I: Understanding DiDi and Data

· Founders and VCs,DiDi,Big Data,Transportation,Internet
broken image

The Harbinger translated's recent interview with Cheng Wei, founder and CEO of DiDi. Looking back at his experience, Cheng Wei says, “I believe that what I've been doing stands for the future. The future is on our side.”

DiDi has now acquired tens of billions in funding, as well as the transportation data generated by hundreds of millions of users. Cheng Wei’s creation has the potential to change the way future generations get around.

In the interview, the post-80s entrepreneur points out that China’s rapid urbanization led to a huge problem: people buy cars only to drive them for 5% of the time, and then leave them in parking lots 95% of the time, reshaping urban space for an inert object.

Cheng Wei considers the possibility of drones taking over urban transportation, as three-dimensional cities are best served by three-dimensional transportation. Meanwhile, planes and high-speed rail trains traveling at three times the speed of sound will cover long distances, and the time it takes to travel among major cities will be measured in minutes instead of hours.

DiDi may be the bridge connecting the era described in science fiction and today’s Chinese society. In the interview, Cheng Wei also talks about the combination of traditional taxi services and online car-hailing services, the impact of self-driving cars, pricing, and his plan to promote DiDi’s model in other parts of the world.

Translated by Shaolong Lin, Tyler Xie, Yang He, Celine Ding, Juzhi Zheng, Larry Liang, Frank Huang; edited by Jordan Schneider, Frank Huang, Glen Meyerowitz, Yamini Bhandari.

[Editor's note: this interview has been edited and condensed for clarity]


Taking the Long View

Cheng Wei: Let me begin with timing. In the past two millennia, most people weren’t lucky enough to experience technology revolution, which offers opportunities to solve existing problems with new technologies and models. The application of new technologies and new models is where innovation and industrial revolution stem from, and only during industrial revolutions are new businesses likely to stand a chance.

Today, the largest banks, retailers, car makers, and FMCG companies are multinational businesses that were founded approximately 100 years ago. They should all thank the industrial revolution and globalization that started around that time. Were it not for the right timing, they would not be where they are today.

Then came the Internet, and the world started to reshape around technology.

Location is also important. If I were not born in China, but in Japan, Europe, India, or other Southeast Asian countries, I wouldn’t have had the opportunity to create an online start-up. Even in China, if I were not born in Beijing, in Zhongguancun, I would be much less likely to succeed.

Everything has to be just right.

In my own case, joining Alibaba in 2005 was also an important decision. I ended up working there for eight years, and am very grateful for that experience. For a long time, I did not know what I was good at. And then one day, I asked myself if I had to choose something, what would it be? I answered that I would choose the Internet, even if I would fail.

So I looked up Alibaba’s address online, and went there at once. When it first took off, Alibaba focused on B2B foreign trade, and our customers often asked us: “Do you sell barbecues?” The process of going from 0 to 1 is always the most difficult, but the most memorable as well.

Then I witnessed how Alibaba completely reshaped the B2B foreign trade, retails and finance in China. As businesses continued to expand, I realized that the Internet would change all industries, and lead us into a new era.

Ma Ping: What else, besides the Internet, do you think could change the world? You mentioned that the Internet was a groundbreaking technology that could disrupt all existing industries, and offer young people opportunities to start their own companies. Based on your own experience, what other important opportunities are there for people to take advantage of?

Cheng Wei: Starting your own business is about creating value for your customers. Technological advances will create massive value for users in every industry. Most of the past 2000 years were the agricultural era, and the development of technology and society was so low that it was impossible for the average person with life expectancy of 50 to 70 years to witness any significant advancement in technology, so it is only reasonable that there have been no waves of start-up in the past.

Few events in history had as much an impact as the Industrial Revolution and the Internet Revolution. So, it is no wonder that both eras witnessed a spate of innovations and start-ups. Right now, there is nothing as influential as the Internet. Artificial intelligence is dominating the second round of the Internet Revolution, and in the future we might see more advancements in life sciences and space technology.

The Internet is meant to connect. Through connections, efficiency is improved, and value is created and distributed through the industries. Taobao stands for connecting people with products, Baidu for connecting people with information, and Tencent for connecting people with one another… And I hope when people talk about connecting people with cars, they will think of DiDi. Connections form platforms, which then collect big data, and as a result we look to artificial intelligence to be more efficient in utilizing these data. This is why AI is the second round of the Internet Revolution.

DiDi is fortunate to have started building online platforms before it was too late. In the second round of the Internet Revolution, we hope to build a super transportation AI and gradually transition into an AI company.

Ma Ping: You are a post-80s entrepreneur who succeeded in this start-up era. If you had the chance to choose when you were born, would you rather be born in the 70s to be able to start your plans sooner, or would you prefer to be a millennial faced with more opportunities?

Cheng Wei: I think I was born at the right time. If I had to choose, I might go with 200 years ago when the Industrial Revolution kicked off in Europe, or 500 years ago in the Age of Discovery. Both were ages full of opportunities. However, in the industrial age, logistics were expensive, and opportunities were scattered, so it was unlikely for a large-scale platform to emerge. The Internet, on the other hand, connects the entire world, integrating it into one huge market, allowing visionaries to grasp opportunities from all kinds of different places. To be honest, I find myself in the greatest age of all, so I won’t go anywhere else.


Lessons from Kings of Glory, China’s Most Popular Mobile Game


Ma Ping: Do you play games with your friends?

Cheng Wei: Yes, I play Kings of Glory [ed. a League of Legends-type game played on mobile]. I am actually very good at games in general. On weekends my colleagues will come to my house to play Street Fighter and Pro Evolution Soccer. I started playing Kings of Glory because I wanted to find out why it’s so successful, also because some of my friends kept recommending it to me.

Ma Ping: You sound like a hardcore gamer. As I was coming upstairs, I even saw a poster for a Kings of Glory competition within DiDi. Why do you think it has achieved such success?

Cheng Wei: Kings of Glory is extremely successful, and there is a lot that we can learn from it.

First, we know from the theory of Yu Jun (designer of Baidu Tieba, now at DiDi) that product value = (new experience - old experience) - cost of substitution. King of Glory is easy to play and therefore attracted a wide range of non-traditional users. Though there were already similar games in the market, Tencent adapted the concept for mobile phone platforms, and simplified it to the extent that even newcomers can enjoy it in their fragmented free time.

Second, the operation of Kings of Glory is based on years of experience that Tencent has cumulated, and is almost the best that you can find today. It includes level-system, teamwork, commercialization, task systems, and so forth. The logic behind the design is comprehensive, and the team also works well with data. A great example for the industry.

Last but not least, the strong distribution channel provided by Tencent is also crucial. King of Glory is a phenomenal game, and could not have succeeded without the right company backing it up.


How DiDi Applies its Data


Ma Ping: DiDi collects data from users all over the country. DiDi Research Institute sometimes publishes reports that use big data to analyze business districts in major cities, which I find quite insightful. Which aspects of DiDi's data do you think best reflect the pace of this era?

Cheng Wei: We process 200TB of data every day. For each order, we carry out 57.6 billion times of computation for route planning. These numbers alone are pretty amazing.

Our vision is to make getting around easier and more comfortable. The first step is to set up an all-in-one platform that connects all means of transportation. The core of our platform is an artificial intelligence brain, called DiDi Brain. It studies each individual user’s pattern of commuting, collects data on road and traffic conditions in real time. As such, it optimizes the efficiency of the whole community and maximizes the personal experience.

For instance, in a Beijing suburb called HuiLongGuan where more than half a million people live, on an average morning, 10% of the population goes to ZhongGuanCun for work, 10% goes to Guo Mao Business District, 80% goes to several other business districts. There will be tens of thousands of people going each direction leaving in two hours. This pattern of transport demand can be discerned by machine learning.

Nowadays, people tend to choose their own means of commuting, including driving their own car, which causes congestion and lowers overall efficiency due to the lack of the systematic knowledge of traffic conditions. Through the deep learning algorithm of DiDi Brain, we aim to optimize travel planning of the whole society through, for example, recommending bus, carpooling, car hailing or bike share services accordingly, avoiding peak time travel and encouraging carpooling, helping drivers avoid congested road sections. A 20% increase in efficiency would be considered extraordinary.

There are big gaps between the supply and demand in commuters, vehicles and road space. This is the root of all the traffic problems. Smart transport and the sharing economy have solved the problem to a certain extent. However, solving the problem of matching people with cars places big challenges on computation speed as the road condition changes in every two seconds which requires real time data collection and processing.

Second, we need to consider the optimal allocation of resources across the whole system. Optimization is not only in terms of timing, but also in terms of how it might impact the traffic in the next 30 minutes. Also, it needs to consider the service quality and safety of the drivers. You may know that DiDi scores every driver according to their service quality but you may not know that we also score each driver in terms of safety. We monitor a range of behaviors such as sudden brakes and fatigue driving in the aim of lowering incident rate. As a result, vehicles with DiDi have a 40% lower accident rate than traditional cabs.

We hope our platform can fully utilize DiDi Brain. In essence, we are a data company and an artificial intelligence company. We established DiDi Research Institute when the company was at its very early stage. Our research center at Silicon Valley has just had its first year anniversary. DiDi finds traffic problem in China both challenging and interesting. We’d like to invite the best scientists to join us in our quest to find a solution.

Ma Ping: So you think, DiDi’s computation volume is unimaginable, compared to other companies in the history?

Cheng Wei: Yes. And the data we have could have multiple applications in the future.

Take traffic light as an example. Traditionally, traffic lights’ signal cycle either based on a certain interval, like 30 seconds or 50 seconds, or based on vehicle detection loops and sensors. None of them can realize accurate and real-time control. Now every intersection constantly has vehicles with the DiDi app passing through. With the data sent back from those vehicles indicating traffic volumes in every intersection day and night, we can make reasonable predictions and give suggestions to the transport department for better-coordinated traffic light cycles. And we’re actually doing this already.

The first batch of big data-based traffic lights has been installed on one road in Jinan City. Many intersections on that road have seen a traffic decrease of 10-20%. And through the coordination of the dozens of traffic lights in the most congested area, the overall traffic congestion has been reduced by 20-30%. We have improved the traffic efficiency this much merely leveraging the data without any extra cost on the infrastructure. That’s why we call it smart transport.

Another example, we have 60% match rate for people who are seeking a pool mate using DiDiPOOL (similar to UberPOOL). Plus, we found that 70% of this kind of orders can be matched together. This helps solve the congestion significantly.

Traffic congestion has been one of the biggest problems of urbanization. On the one hand, people live in the multiple layers in the city, but we only have one or two layers for vehicle and trains. On the other hand, we have some many vehicles that only have the driver in it. Actually, statistics show that 80% of vehicles on the road have empty seats, many of them going in the same direction. That’s why we need to utilize data to optimize the traffic and encourage people to carpool.

There are extras from utilizing data. One example is that when a cab driver becomes a DiDi driver, his service quality usually will improve a lot. Why does this happen? Because we keep a record of his service, the quality of service impacts his future income. Data changes his incentives and behavioral patterns.

Today we are developing statistical models for predicting transport safety. It is the internet that makes everything transparent and makes things easy. We have data on every driver’s every driving behavior: how often he drives over the speed limit, how frequently he takes a sharp turn, etc. All driving behaviors are used to predict a driver’s rate of accidents.

For those drivers at a higher risk level, we do personalized education rather than simply emphasize following traffic rules. For example, we are able to specifically warn a driver that he/she should not be too close to the inner side of the road when turning.

Also, DiDi has a restrictive policy on driving hours. No matter you are a DiDi driver or DiDi chauffeur (premium drivers), if you have been diving for too long, we’ll force you to go offline. We hope to ensure the driving safety, avoid fatigue driving.

Safety challenges vary across counties. In China we have traffic safety problem while in Brazil, we’re facing more issues in crimes. We all need big data to identify the behavior of riders and drivers, in order to prevent the issues or risks from happening.


Ethics and Government Access to Data

Ma Ping: When we think about self-driving cars, their value is not just a question of efficiency. There are also legal and ethical issues. One common problem is the so-called ‘Trolley problem.’ A self-driving car is approaching a motorcyclist who is not wearing a helmet but can maneuver and hit a motorcyclist who is wearing a helmet. To minimize injury, an AI may decide to hit the driver with a helmet on. However, this case is not so simple, and we may think that the motorcyclist with a helmet should be rewarded for obeying the law. Clearly, it is not fair to endanger someone for being a law-abiding citizen. DiDi Research has invited sociologists and anthropologists to contribute to this research. What insights have they offered about these situations?

Cheng Wei: Why does your hypothetical only have two inevitable crashes as outcomes? In our daily lives, drivers are always distracted by phones or other stimuli, or there are objects in our blind spots.

These extreme hypothetical problems will not even exist with self-driving cars. Self-driving cars will not be distracted by phones. While cameras and sensors on the car may be limited, the real-time data analysis can pull data from anywhere. Our top priority is to develop our technology properly, so that many of these ethical questions are unnecessary.

Finally, it is not just the responsibility of a company to determine its social and economic value. That should be determined by users and consumers, but we are not there yet.

Ma Ping: DiDi often has more information about traffic and transportation issues than the government. The amount of data you collect is increasing, and the government may ask companies like DiDi to affect transportation in cities. How would DiDi balance their social responsibility with the political power of the government?

Cheng Wei: DiDi is a technology provider, while the government is the operator of transportation systems.

But the Chinese government is always looking for new private partnerships. Elements like infrastructure monitoring; determination of pricing; and providing services, like buses, can be handled by the private sector. Private companies can provide the best services at the best prices to meet customer needs. Administration is one area the government should be in charge of.

Ma Ping: With the development of technology, it is possible that the need for transportation will decrease as work can be performed remotely. How will DiDi respond to these changes?

Cheng Wei: Transportation today is far from perfect. We have been working hard and have significantly improved China’s transportation systems over the last five years. There are better products, but it is still far from perfect. Perfect is calling a car and not having to wait. Perfect is not having traffic on the roads. Perfect is customers being satisfied with our services.

Customers still need to wait when they call a car. There is still traffic on the roads. I hope that DiDi can promote the idea of ridesharing and increase the utility of each car, and decrease the number of cars. I hope we can increase the quality of our service to provide a better experience for our customers. Looking at all this development, there is still a long way to go until we reach a limit.

We recently talked with the government’s Industry Association and they indicated that transportation expenses make up 14% of an individual’s expenses. This has been a very stable percentage, even with changes and improvements in technology. Based on this, I do not think we need to worry about a shrinking market.


Looking Beyond China


Ma Ping: China is implementing the Belt and Road Initiative to increase connectivity and facilitate development between China and Eurasian countries. Does DiDi have any plans to take advantage of this opportunity? Will DiDi look to replicate its successes in Eurasian countries through this initiative?

Cheng Wei: I do want to promote DiDi as a global brand. Over the past five years, we have been trying so hard to solve China’s transportation problems and have gained much experience and made many technological improvements. Many innovations relating to ridesharing have originated in China, and these should be passed to global customers. I am sure China will be at the center of transportation revolution and the center of sharing economy.

But how? DiDi wants a collaborative, and not competitive, position in a new market. Transportation is a very localized industry, and if DiDi wants to be a part of a new market, for example in either Brazil or Southeast Asia, it is not wise to compete against the local companies. What we want is to find ideal local partners from all different markets to help export our model and technology.

DiDi has already invested in local companies in all BRICS countries. We invited companies like Ola and Taxify, local market leaders, to China to discuss collaboration and growth opportunities. As far as I am concerned, there is huge value in these discussions and it will allow us to provide better services to customers.

Our long-term goal is to open the whole world to our users, so that they may go to any country and use DiDi to facilitate ease of transportation. We are looking to simulate airline partnerships that cover huge areas. Regardless of where our users go, they will still be able to experience the same level of excellent service.

This article was originally published by Ma Ping via Guancha Syndicate on September 26, 2017.