Tencent, A Manufacturing Re-evaluation
*Note: These are Jeffrey Ding’s informal and unofficial translations -- all credit for the original goes to the authors and the original text linked below. These are informal translations and all credit for the original work goes to the authors. Others are welcome to share excerpts from these translations as long as my original translation is cited. Commenters should be aware that the Google Doc is also publicly shareable by link. These translations are part of the ChinAI newsletter - weekly-updated library of translations from Chinese thinkers on AI-related issues: https://chinai.substack.com/
Reporters: Wei Pang (微胖)
Source: 机器之能/jiqizhineng (Synced) — a long-time source for ChinAI translations, often features longform articles about China’s tech industry
Date: September 26
"Tencent does not have factories and does not understand manufacturing," Li Xiangqian, general manager of Tencent Cloud’s Industry Cloud, once told the media. Yet, this does not prevent Tencent -- born from social media -- from trying to find the most suitable slice of the entire manufacturing chain (research, production, supply, sales, and service).
Holding unparalleled connection capabilities for massive amounts of users, Tencent chose to go from sales and services to cut its way to the front end. And the clients who choose Tencent highly value user traffic. "De-linking" allows companies to directly face consumers, and it also provides new opportunities for the transformation of manufacturing services.
In addition, Tencent's core capabilities are the cloud and the bottom layer. It prefers to serve as the Internet's water and electricity, so it also needs third-party service providers as complements.
As more and more traditional industries join Tencent's ecology, Tencent's business will become ubiquitous and possess unlimited imaginative space.
Like cutting meat with a slow knife, factories are not going into operation, nor are they stopping production. But after toiling away for a while, you discover that the situation is becoming worse. This has already been the portrayal of China's manufacturing industry for many years.
From doing interviews last year, I still remember：
there was a safety shoe factory located in an industrial zone in Baoshan District, Shanghai. There was no one willing to take it over, and it was just waiting to die because of continuous rent increases, stop-and-start production, and thinner-than-paper profits.
A factory owner, who had worked in the educational toys industry in Zhongshan for twenty years, gritted their teeth and bought fifty robotic arms;
Although the senior technicians of a medium-sized cement plant questioned these Internet people who didn’t even know how to heat calcium carbonate, the declining cement prices and small profits forced them to do something desperate.
Finance writer Wu Xiaobo predicted in 2017 that in the next few years, 80% of small and medium-sized enterprises in traditional manufacturing will go bankrupt. Their elimination will be very tragic, but the industry will eventually force everyone to make changes.
For these manufacturing companies ahead of them are two paths: one is transformation via informatization, with high levels of technology and capital density, such as robots and industrial Internet; the other path is called the servicification of manufacturing.
This will be an unprecedented societal movement. Many years ago, Tencent Cloud, which was still a new company at the time, "foresaw" the future and planned to move millions of traditional manufacturing industries to the cloud.
Unlike Germany, for which Industry 4.0 was initiated and implemented by top companies within the (manufacturing) industry, the existence of major consumer Internet companies in China dictates that they will play an important role in the transformation of the manufacturing industry. These companies hold a large amount of user and business data and are eager to realize these potential resources.
Just two years ago, Tencent made its third major organizational adjustment since its establishment. If it is said that before this, Tencent's B-side services had not opened up and connected their internal structure, pushing forward on one’s “knees.” After the third structural adjustment, the obstacles in B-side services have been eliminated one by one.
1. Traffic: Marketing transformation, who else besides me?
Almost all Chinese can tell the difference between Tencent and other major Internet companies:
He allowed 1.2 billion mobile Internet users to install WeChat, as well as the WeChat Enterprise (WeChat Work) dream of allowing 43 million domestic enterprises to have their own instant messaging tools.
With these connectivity capabilities in hand, "Tencent is going from sales and services to cut its way to the front end." During the Tencent Global Ecosystem Digital Summit, Liang Ding’an, general manager of Tencent Cloud Intelligent Manufacturing, said in an interview with the media, "Tencent has made connections before. Where there is a connection, there will be business opportunities, and where there is traffic, there will be services.
A complete manufacturing chain includes research, production, supply, sales, and service. Tencent, from a background in social media, is trying to find the most suitable slice of this chain.
The essence of each company’s cloud services will not be much different, so clients that choose Tencent are still choosing based on user traffic. An industry insider told us that merchants who have marketing needs will seek Tencent.
After Huazhi Cloud discovered some of these paint points while cooperating with Linglong Tire, they went to find Tencent.
Linglong Tire is the second largest tire manufacturer in China, with nearly 2,000 types of tires, and wholesalers and agents all over the country and even the world. For a long time, they had adopted the traditional marketing model of going from manufacturers to wholesalers to retailers, and finally into thousands of households.
For each additional step in this circulation process, the average price increase is 5% to 10%. This model requires high gross profit to support. With the domestic tire market entering a stage of complete competition and overcapacity, the profit of tire wholesalers may only be 2%.
Another aspect is that, even if price increases at each level are eventually spread onto costs on consumers, it is very difficult to transmit their feedback on the product back to the manufacturers.
By a "two-end" approach — WeChat and WeChat Enterprise (WeChat Work), Tencent tried to build up Linglong Tire's ability to connect with its customers: the connection between Linglong and dealers, the connection between dealers and lower-level dealers, the connection between dealers and stores, and the connection between stores and customers.
"After constructing this loop, companies will discover new business models." Liang Ding'an said.
For example, when changing tires, customers will go to the store, and sales can connect customers to the company by scanning the code to retain user data.
In the past, Linglong Tire could only see dry data. For example, this car company bought my tires, that e-commerce platform sold 100,000 goods, and certain regional dealers sold very well. Now, the company knows who bought my tires, where they bought them, the type of tires, and even the reason for the purchase, model, etc.
This will not only help manufacturers to more accurately advertise and diversify the terminal stores and improve their promotional strategies, but also affect the back-end production links.
"Originally, we were targeting nearly 400 dealers and fifty or sixty automobile factories across the country, designing orders and products according to their needs,” Linglong Tire Chairman and President Wang Feng said in an interview with the media.
"Turning to new retail, we are now targeting hundreds of millions of users across the country, studying the needs of end users, designing products, and arranging manufacturing plans."
In the past, it was difficult for tire manufacturers to control channel information, and there were often situations in which marketing was not implemented, loyalty was not high, malicious billing, and merchandising. With the "two-end" approach — WeChat Enterprise (WeChat Work) and WeChat, Linglong can connect the upstream and downstream from the large B (large area distributor) to the small B (mom-and-pop shop) and realize the control of the distribution system.
As manufacturers will provide platforms and uniformly control prices, distributors are also changing to service providers. The business expanded to "combination of tires and non-tires", "combination of tires, non-tire products and services", and the profit is no longer tire price difference but service. Such as financial services, warehousing costs, last mile transportation fees, manufacturer service fees, etc.
After launching new retail, Linglong Tire’s performance rose, against the market trend, during the epidemic, and it ranked first in the “domestic tire original equipment list” in the first half of 2020.
In Liang Ding’an's view, for this leading company, rhetoric such as "strengthening core track capabilities", "investing ten million to save hundreds of millions" is not enough to move them.
"Can there be some new business models that will allow it to find new growth points and make more profits?" Liang Ding'an said, "It cannot be answered by selling tires alone." In fact, this is what Tencent is trying to export to more traditional companies that only focus on products but ignore services. "If I can’t even sell my products, what's the point of doing smart manufacturing?"
This is also exactly the difference between China and Germany’s manufacturing transformation. It can be said that if Germany is more focused on optimizing production processes, then, it is easier for Chinese companies to see opportunities such as expanding service types and improving services.
2. Platform: the "base" for doing a good job
In many more aspects, there is no essential difference between Tencent Cloud and other Internet giants. The core capabilities are still the cloud and the bottom layer.
Cloud storage, big data mining and computing are the "standard equipment" of their services. In past interviews, some people in the manufacturing industry have told jiqizhixin. They are also more willing to do common things, such as the construction of large data centers and middleware.
"Low profits and inability to scale all affect their continued investment and depth of cooperation."
Tencent Cloud sees huge business opportunities in the high-end equipment manufacturing industry. Over the past ten years, the global market share of domestic equipment manufacturing represented by construction vehicles has been steadily increasing. SANY Heavy Industry's market share increased from 10% to 26%, and the market share of other brands also increased from 27% to 36%.
"In recent years, there’s been a rise in the leasing of light equipment as well as other business lines, which indicates that the proportion of services is gradually increasing." Xu Zhixiong, deputy general manager of Tencent Industrial Cloud, said at the Tencent Global Digital Ecosystem Summit.
Construction machinery is different from those products that "as long you can sell them out the door, all is fine." After the products are sold to users, manufacturers still have to assume the long-term responsibility of providing users with parts replacement and other equipment maintenance services.
Due to factors such as price and professionalism, there are not many purchases of construction machinery in full cash, as mortgage purchases or leases constitute most cases.
The benefits of ICT applications on construction machinery to manufacturers and leasing companies have already been verified by the Japanese construction machinery giant Komatsu's "KOMTRAX System.”
For example, “bad debts” rarely occur when buying Komatsu machinery through mortgages in China; data can also be fed back into the product design process. But it was a Japanese leasing company that really mined the value of this system. They found that the various data related to construction machinery collected by the system brought a lot of convenience to how the leasing companies managed their vehicles.
For example, you can grasp useful information such as "a part of this equipment should be replaced" without leaving home. When the leasing company wants to send a vehicle to refuel the construction machinery, it can also confirm the current remaining fuel of each equipment through the KOMTRAX system in advance, and then make arrangements for the order and route according to the principle of highest efficiency.
As an important competitor of Komatsu, in order to tap the potential of data to provide better services to customers, SANY Heavy Industry is also actively creating innovative applications with regard to the "last mile," and these upper-level structures are inseparable from strong infrastructure support.
SANY Heavy Industry has collected a large amount of equipment data, with some high-frequency sensor data being accessed 24/7, and about 400,000 pieces of engineering equipment that need to be monitored. Whether it is data compression or storage and computing, many companies find it very difficult to be competent.
When we used to do consumer Internet, we accumulated rich experience in big data management. Liang Ding'an explained that in addition to the absolute cost advantage, Tencent's understanding of data compression and calculation is also very advantageous.
"Tencent has more than 1 million servers, and Tencent Cloud also has operated these 1 million servers for a long time and has the ability to provide services (in this domain)," a Tencent Cloud employee told us. "This ability is more valuable than having physical servers."
At present, SANY Heavy Industry can complete the monitoring of 400,000 engineering equipment, reaching early warning of equipment failures 6.5 hours in advance, and the early warning accuracy rate is 87%. At the same time, the frequency of sluggish inventories with regard parts that wear down is more than 40% lower than others in the same industry
Foxconn, which has a deep accumulation of automation, is another representative case of manufacturing service transformation.
Foxconn not only has the ability to do factory consulting and planning, but it also has rich experience in tool use, and has a relatively deep understanding of the models and mechanisms for predicting tool life. Whether it is inside or outside the enterprise, there is demand for this accumulated experience.
From an internal point of view, this means that the new plant can quickly replicate this set of technologies; as for the external huge industrial chain, it is far from enough to "go it alone." Only by attaching these capabilities to the industrial chain can we achieve higher product quality.
Tencent Cloud started from the Internet of Things, and transmitted the data collected by various sensors of each machine to the cloud via 2G, 3G, and 4G networks to solve the problem of massive data storage, analysis and utilization.
Tencent "honestly builds the base of the cloud platform, the entire data center, and some AI and other capabilities." Liang Ding'an told us, "Then, integrated with some innovations in the application layer of the enterprise, we build the platform from there."
3. Explore industrial vision opportunities in the production process
From the perspective of traditional manufacturing, it is difficult for the Internet giants to transform them at the critical segment of "production."
The bottleneck is mainly limited understanding of industry and technology, the aforementioned manufacturing professionals admitted frankly. For example, the process manufacturing process is continuous and cannot be stopped. Problems in any process will inevitably affect the entire production line and the final product quality. During this period, it relies heavily on empirical data, and the parameters and data analysis are complex. "Without the deep cooperation of clients, it is useless to rely solely on algorithms and models, regardless of equipment and technology."
As for the core links and core equipment, such as blast furnaces and steelmaking converters, even a small safety failure will have a huge impact, so Internet companies dare not try to control it easily.
However, Tencent Cloud is also making note of some industrial vision opportunities. In Guangming District, Shenzhen, China Star Optoelectronics, a subsidiary of TCL, has a huge factory, in which workers are rarely seen on the production line, and robotic arms are busy at work.
Though the level of investment isn’t particularly large, the first domestic AI recognition project for LCD panel defect types, ADC (Auto defect Classification, has been repeatedly brought up as a typical example of industrial Internet intelligent manufacturing.
Before the ADC project, AOI equipment was required to take pictures of glass defects and upload them to the NAS system. The staff downloaded each picture to determine the classification, and the identification results were uploaded to the system. All work was done manually.
By importing Tencent AI to determine whether the film had defects, a film recognition speed for the whole panel of 15 milliseconds has been achieved, compared to about five minutes for manual work. At the same time, the defect recognition accuracy rate has exceeded 90%, exceeding that of humans.
In the beginning, there were only 20 or 30 personnel replacements. The team took two years to upgrade the system. Now the number of replacements has reached 140, which can replace 50% of the labor power in the future.
In this project, Tencent Cloud provides the underlying technology, mainly doing two things. One is the vision algorithm itself, which trains the algorithm model to identify defects.
The other is the underlying support behind it. "This is very important. After storing the massive amount of data returned, the underlying support for AI rapid image recognition is also needed." The aforementioned Tencent employee told us.
Tencent has designed a stable and reliable automated identification for the panel defects which is suitable for industrial production lines. This is in order to ensure that nothing goes amiss in interactions between the 1.4 million pieces of image data and the MES system, hundreds of models and site data are accurate and correct, and GPU card resources are flexibly scheduled and load balanced.
In addition, because the technology and manufacturing process are also constantly adjusted, for example, changing the production batch or production model, the model must constantly adapt to these new problems and be stabilized through training before being put into production again.
"This process also requires a platform to help implement very complex AI image annotation and training," the aforementioned Tencent Cloud employee said.
China Star Optoelectronics began planning to launch this project in 2017, and after a year of discussions and negotiations, it officially launched in 2018. Tencent has tackled the most difficult work with regards to the vision algorithm. "Because Tencent has the strongest (computer) vision algorithm team in China, they definitely made the right decisions through the process," said Liang Ding'an.
Now, Tencent is also reusing some of the industrial vision intelligence capabilities it has accumulated in working with China Star Optoelectronics and Airbus to deepen its manufacturing. For example, in the manufacturing process of Linglong tires, there are many steel wires in the tires, and there are also industrial vision opportunities with regard to examining these steel wires.
"If the leading companies feel that they can handle this core algorithm, I think they have failed to identify their core competitiveness." Liang Ding'an said.
4. Ecosystem: When everyone adds fuel, the fire burns brighter
In addition to its pride in its traffic (connections to users) and the platform base capabilities of major Internet companies, even though it is a giant, Tencent also relies heavily on its ecosystem partners as complements.
In addition to Internet startups, offline companies also urgently need IT technical support, upgrades and transformations, which also constitute the largest stock market for cloud computing. In this market, the information architecture and direction of enterprises are often determined by system service providers or ISVs (Independent Software Vendors). If Internet companies want to develop users in this field, they must win system vendors.
"The service provider does something similar to a bricklayer, building the raw materials provided by Tencent into a house where businesses can live." Zhu Ning, the founder of Youzan, once described the relationship between the WeChat ecosystem and Youzan.
However, every link from retail to production is clear to the partners, and partners pay attention to the different situations for each company. Tencent does not have this condition. "Tencent has front-end capabilities, cloud and AI platform capabilities, and then they rely on partners to do the commercialization capabilities for the rest." EC founder and CEO Zhang Xingliang said in an interview with AI Finance and Economics.
This is the main difference between Tencent and other Internet giants in the development of industrial Internet, and it is also the motivation behind Tencent's ecological partner recruitment plan "511". Co-build with partners, a solution for the thousands of people in Chinese manufacturing companies.
In fact, in order to attract more ISVs to the enterprise WeChat, Tencent is also intensifying its opening up of its ecosystem, putting ISV tools into the workbench to enrich the existing application ecosystem. During the epidemic, many cities and regions launched a large number of small programs for epidemic prevention and control and resumption of work and production. The ability of a large number of users to use these services without any threshold is a manifestation of Tencent's open ecological integration.
Tencent wants to use team one-to-one services to create benchmark cases and demonstrate to more businesses what Tencent can achieve. Liang Ding'an told us that more landing work needs to rely on a large number of service providers.