The following article is from 36 krypton Pro by Wang Yutong Xiaoxi
On March 5, 2016, the weather in Beijing was fine. At 10 a.m., in a room ten meters long and three meters wide, several programmers in plaid shirts began to communicate with a white board.
This is pingcap’s first infra meetup. Among these people, there are nearly 10 founding teams from pingcap, and the only two guests are open source enthusiasts from Huawei and JD. It’s understandable that there are more hosts than guests. After all, the correct way to open it on Saturday morning should be “make up sleep”.
Today, infra meetup, with the theme of exchanging infrastructure technologies at home and abroad, has held more than 130 sessions. Pingcap has also become one of the largest open source communities in China today. As an underlying software company, the database product tidb independently developed by pingcap has won more than 25000 standard stars in GitHub (the world’s popular open source code hosting platform), and has gathered nearly 1200 contributors; Worldwide, tidb is also a well-known open source project in the field of infrastructure.
But the ambition of the founding team is far more than that. CEO Liu Qi said:“We still need to focus on tidb, hoping to achieve a market value of $50 billion, which requires revenue support of $1 billion.”
01 commercialization: hundreds of paying customers, 300% revenue growth
What is the concept of a market value of $50 billion? Oracle has a market value of $170 billion and snowflake has a market value of $70 billion. If you want to achieve a market value of 50 billion, you need a revenue support of $1 billion. How to do revenue?
For open source software, the primary concern of everyone is:Free open source products, can you make money?
Let’s start with the answer: Yes.
At present, pingcap has the highest revenue in the field of open source software in China. The company has more than 1500 households using tidb for online production environment. The customer unit price in China is in the order of one million yuan, and it is several times higher abroad.
China’s open source is still in the early stage of development, and most open source manufacturers have not yet begun to commercialize. The commercialization of pingcap has enabled more domestic open source companies to develop faster. In the field of open source, the head dollar fund is also actively layout. For example, Jingwei has also invested in pingcap, nebula and many other open source companies in recent years.
Pingcap had no revenue in the first two years of product development, but after the commercialization, pingcap had more than 100 paying customers, and the revenue growth rate remained at 300% on average。 Pingcap told 36 krypton that low customer acquisition cost, high retention, high additional purchase, stronger sales efficiency than traditional software companies and the acceleration of product iteration are important reasons for pingcap’s rapid growth.
The commercialization of most commercial open source software is to provide free version software first, and then charge for advanced functions or services later. The source code can be viewed and used by everyone, so many potential customer developers will also do projects based on open source products to make the core products more useful and valuable. This is also the biggest advantage of the open source model, so there is little resistance when it is adopted.
But the difficulty is also obvious: open source software may take years to develop successfully; Open source software word-of-mouth is more important than traditional marketing methods. Therefore, it takes some time to collect feedback on iterative products; The gap between open source users and paying users is very large. After all, it is not easy to change from “free collaboration projects open to all” to “commercial products open only to paying users”.
So,The best way for open source software to gain the trust of commercial customers is to be completely transparent, plan functions in advance and discuss them in the community. In conclusion, it is to use the “open source community” to achieve a high match between open source projects and the community.
So how can open source commercialization be realized in China with lower willingness to pay and weak digital awareness?
Compared with foreign models, pingcap also has two advantages: Communication Based on open source community and clear payment and free functions.
First of all, pingcap is based on the open source community, which can achieve one pass, ten pass and hundreds of customers.
Open source naturally has the same viral communication ability as TOC. In the engineer community, unhappy engineers will feed back and provide solutions; The better the product quality, the more people promote it, the faster the problem is solved, and the faster the positive cycle is. This is much more efficient than traditional software by selling door-to-door knocking.
Let’s see how pingcap defines the free function and paid function.
Different from the business model of the first generation of open source software such as red hat, which focuses on supporting services, pingcap is a third-generation product. In addition to services and hybrid licensing, it can also be paid according to the SaaS model.
The software customer acquisition process is divided into several steps: listening and speaking – Research – testing – Online – maintenance. Open source is the source of the funnel.
This funnel naturally contains the advantages of open source over traditional SaaS:High customer acquisition efficiency and low cost。
First, in the “listening and speaking” stage, open source can be trusted more easily because it can see the source code.
Second, in the “research” stage, customers need to compare and select types when using traditional SaaS services, but the quality of open source products can study the source code, and there are many community users’ ready-made test reports for reference, so as to save time for both customers and manufacturers.
Third, when customers encounter problems in the process of “testing” and “going online”, many open source developers can help, but in the state of closed source, it can only be solved by manufacturers.
Fourth, the open source community model is naturally very friendly to the “operation and maintenance” stage.
The first payment occurs in the online phase, that is, the purchase of tidb services. In the domestic market, the companies that choose to pay at this stage are often financial institutions; A few Internet companies will also pay at this stage, such as Gaia mutual entertainment, hungry, etc.
In addition, tidb is a standard OLTP (online transaction processing) database, which carries key production businesses such as bank core system. Such key businesses require very high business continuity requirements (SLA), so it is almost just needed for the original factory business support, which is also an important reason for the high conversion rate of pingcap open source users and customer unit price.
Finally, the difference of pingcap in Chinese soil is that it has identified the customer base: Internet enterprises and financial enterprises.
Internet companies are the early target customers of pingcap. First, early open source enthusiasts often worked in Internet companies. Second, the architecture design of tidb separation of storage and computing allows users to expand and shrink capacity on demand, which meets the business needs of Internet companies with rapid business growth.
Pingcap’s first single payment customer is also an Internet company. In August 2016, Gaia mutual entertainment, a game company, invited pingcap, which has no idea of commercialization, to test. At that time, Gaia mutual entertainment had an online database system, which was mainly used for the analysis of advertising. It needed to see the advertising results in real time. With the rapid growth of business, the system had reached the bottleneck and urgently needed to be transformed.
Tidb’s positioning can perfectly solve this pain point. It only needs to expand the capacity without transforming the existing business. However, tidb at that time was not stable enough and was in the transformation stage from “can see” to “can use”, so the internal test invitation for Gaia mutual entertainment was quite flustered. “We were so ungrounded that we didn’t even have a Chinese manual,” Liu recalled.
The real launch is also full of twists and turns. The first online time was 5 a.m., but it didn’t succeed. Gaia mutual entertainment also continues to encourage pingcap to let them find the reason. The next day, tidb went online again, stumbled and finally succeeded.
After that, Internet enterprises became one of the customer sources of pingcap. In June 2020, pingcap released the tidb cloud product to provide out of the box tidb cloud database hosting service relying on the public cloud. With the help of cloud, tidb cloud can expand horizontally, with almost unlimited storage capacity and computing power, so that users can focus on the rapid growth of their own business.
Compared with revenue, pingcap’s more important reason for choosing the Internet industry is that the application scenarios of many Internet companies in China are actually very valuable and challenging. Many scenarios have ranked first in the world, which is China’s unique advantage. In the process of serving these top scenarios, pingcap continues to polish its products and businesses and grow together.
The reasons for choosing the financial industry are not complicated. The financial industry is in the key window period of digital transformation. There are not only the time window of steady-state scenarios such as core business system migration and transformation, but also a large number of sensitive scenarios such as Internet services that tidb is very good at. In addition, the financial industry has a high willingness to pay. The net profit of the financial industry accounts for 70% of the domestic listed companies, which is almost the proportion of the financial industry in the domestic revenue of pingcap.
Of course, the financial industry has higher requirements for security and availability. When serving the financial industry, it is inevitable that they will fight hand to hand with big manufacturers such as Alibaba and Huawei. However, in Liu Qi’s view, although they have strong comprehensive ability and wide business lines, they may not invest much in a single point. In addition, considering the neutrality of the system, most CIOs and CTOs do not want to be locked by a single big factory.
Therefore, after winning the benchmark customers in the industry, pingcap’s financial road will be much smoother. At present, it has served dozens of financial institutions, including Bank of China, Everbright Bank, Shanghai Pudong Development Bank, Zheshang Bank, Bank of Beijing, Weizhong bank, China life, Ping An Life, Guotai Junan, Huatai Securities, lufax and so on.
**In conclusion, pingcap’s open source commercialization has been realized in China. First, it is based on the advantages of efficient communication of the open source community. Second, it is more efficient and low-cost than traditional software, which is more suitable for the needs of the current scene. Third, it selects the right industry, cuts in through the Internet industry, polishes and iterates products at a high speed, finds the right time, and then radiates to other industries.
02 Internationalization: overseas revenue has exceeded domestic revenue
Knowing how to make money, the next step is to clarify how much money you can make, that is, answer the question of “the best market”.
The market scale determines the ceiling of the company and how big a story the company can tell. Let’s take Oracle, which has the highest market share in the database industry, as an example. In 2015, when pingcap was just established, Oracle’s share in China’s database market was as high as 56%.
Oracle’s largest market is in the United States. This is also in line with the current situation of any software in the world, whether Oracle, salesforce or snowflake, the revenue of the U.S. market can basically account for half of the global revenue.
It is understood that the Chinese market accounts for less than 5% of Oracle’s total global revenue. The logic behind this is that the Chinese market has a large demand for databases, but at the same time, Chinese customers can not well support the rapid commercial growth demands of a global database company because of the objective reasons of human cost, payment habits and payment ability.
On the other hand, the overseas market is also further increasing.As the earliest and largest institutional investor of pingcap, Xiong Fei, a partner of Jingwei China, believes that for the overseas market, the popularity of the cloud also brings new ideas to the commercialization of open source infrastructure software. Compared with the traditional software sales model, providing services on the cloud is undoubtedly more efficient. Due to the high standardization and automation of cloud infrastructure, At the same time, with pay as you go, a more user-friendly payment and rental model, relying on cloud infrastructure is an important way to obtain large-scale revenue for highly standardized software products such as databases.
In a report by Morgan Stanley, the market of relational database will reach $25 billion in 2024, and pingcap, a new relational database relying on public cloud, will account for 46% by 2024, an increase of 19% over 2019.
In other words, the Chinese market is suitable for accumulating product scenarios and solution needs in the early stage and constantly polishing technology and service capabilities. The overseas market is suitable for making profits after the solutions are mature.
This is also pingcap’s strategy.
China, which has experienced the digital transformation of mobile Internet, finance and logistics, has cultivated many of the most abundant and challenging scenes in the field of basic software: meituan is the largest takeout trading platform in the world; Sitong Yida is the world’s top logistics company; Shell is the world’s largest housing trading platform… Serving them well means that it has the ability to serve similar overseas companies.
Therefore, starting from serving shopee in 2019, pingcap has moved to the overseas market.At present, the revenue in overseas markets has exceeded pingcap’s domestic revenue, and there are leading customers in different countries, such as paypay, the largest online payment company in Japan, square in the United States, vng unicorn in Vietnam, zomato in India, shopee, the largest e-commerce in Southeast Asia, Dailymotion, the largest online video company in France, and so on.
One of the reasons for globalization is that the Chinese market is not big enough. On the other hand, from the perspective of the best products, good products should pass through the world. Therefore, on the first day of its establishment, pingcap took internationalization as a standard configuration.
At present, pingcap has more than 300 overseas community contributors from Google, twitter, square and other companies. By participating in the activities organized by CNCF and publishing popular articles on cutting-edge technologies, the company attracts technology masters interested in tidb in Hacker News, dzone and other media. Finally, many people will settle on some professional social platforms, such as slack or interest groups in the community, and hatch a series of potential business opportunities.
03 product: separate structure, reducing labor cost by 75% for enterprises
The foundation of open source commercialization is still products.
In 2015, Gartner, a consulting firm, predicted that the market capacity of commercial databases would decline to about $22 billion by 2021. The rapid development of cloud technology and hardware technology, as well as the rapid popularization of open source database technology, will jointly split the field of commercial database.
At the same time, there have been some breakthroughs in distributed technology in academia. This provides a theoretical basis for Liu Qi, Huang Dongxu and Cui Qiu who started their business in 2015.
The R & D cycle of a new database product is destined to be very long. Patient capital can help pingcap through the green and yellow R & D period. After several years of enterprise service, Xiong Fei, a partner of Jingwei China, realized that entrepreneurship based on underlying architecture will be the next wave. In conference room 4 of Jingwei Beijing, Liu Qi, who didn’t even know what BP was, talked with Xiong Fei for nearly three hours with Dongxu and Cui Qiu. Xiong Fei quickly made the investment decision.
Allowing users to easily operate and use the database is the core concept of tidb product design.The user needs are clear. He wants a database that can automatically scale flexibly and solve the problem of sub database and sub table (in order to solve the problem of performance degradation caused by excessive amount of data, the original independent database is divided into several databases).
In order to control the scale and complexity, tidb layers computing and storage. The first layer is computing layer SQL, which is characterized by concise language and can be used by non computer professionals. The other layer is the stored key value layer.
The closer you are to the user, the faster you can get user feedback and verify whether the idea is feasible. Therefore, the team finally chooses the order starting from the SQL layer.
How to replace the original database is also a huge challenge. User habits have inertia, and various factors are also considered during switching. In 2016, the stand-alone version of MySQL was still the mainstream. After investigation, the team found that users did not dare to switch to new products immediately because they were worried about data loss and other problems.
In the continuous iteration, tidb also defined its entry point: integrate the distributed theory with the traditional relational database, compatible with MySQL protocol and ecology, and the application can be migrated from Mysql to tidb without or a small amount of modification.
However, pingcap is not satisfied with replacing MySQL with a small amount of data. Enterprises with large-scale growth of service data are the biggest change of new distributed data compared with stand-alone mysql.The design of tidb storage and computing separation architecture allows users to expand or shrink the capacity of computing and storage online as needed, and makes the whole process transparent to application operation and maintenance personnel.
Two years later, tidb was introduced to the market and began to commercialize. As mentioned above, the two benchmark customers in the early days of tidb are hungry and the pain points of moBay are the soaring business, but the database expansion and operation and maintenance can’t keep up. Through the flexible increase and decrease of nodes, tidb helps customers realize the performance improvement and stable upgrade of the database under the large-scale growth of business.
After challenging the “big guy”, tidb has also accumulated rich scene experience. A new generation of customers like Peiyu choose all in tidb, which can not only solve the pain points, but also save manpower. In traditional enterprises, the bottom layer of each business line needs a group of manpower to support. At present, the data bottom layer built based on tidb has only one set of support service team, which supports the products and business lines of four companies, improves the work efficiency of business R & D and DBA, and reduces the labor cost by nearly 75%.
As the answers to products, business models and markets are gradually clear, capital is more and more optimistic about pingcap:Pingcap obtained the angel round financing of Jingwei China leading investment in 2015; After that, new capital will be re injected every year; On November 17, 2020, pingcap completed round D financing of USD 270 million, which was jointly invested by GGV capital, access technology ventures, Chenxi investment, times capital and Wuyuan capital (formerly Chenxing capital), and the old shareholders Jingwei China and yunqi capital continued to increase their holdings.
In the eyes of the investors accompanying the growth of pingcap, Zhang Ying, the founding managing partner of Jingwei China, often reminds the company: pay attention to cash flow, cash flow and insist on exploring commercialization; Liu Qi himself felt that pingcap was not grounded enough, so in terms of commercialization, he must make up for his shortcomings by making up for relevant sales talents.
04 open source community: nearly 1200 contributors have actually participated
Looking at pingcap’s five years at this moment, the replenishment of products, business models, markets and capital is inseparable from a keyword – open source.
It can be said that “open source” is the source power of pingcap.
Something similar to the spirit of open source is not uncommon. Within a specific scope, the source code of open source software can be reviewed, modified and enhanced by anyone, similar to the encyclopedia Wikipedia, which can be edited by all mankind. For an open source project, engineers from Alibaba, meituan and any enterprise in the world can participate in R & D, submit code and use jointly developed products.
Another famous analogy to open source is the bazaar model. In the open source book cathedral and bazaar, Eric Raymond, a legend in the IT industry, said that there are two kinds of buildings in the world: open source software is a bazaar, which is open every day, from scratch and from small to large; The traditional software is the cathedral, which needs the painstaking efforts of several generations to be built and put into use.
The reason for this analogy is that the market model allows more engineers to participate in the software development process, improve software quality, improve their own products and support their own services.
Maintaining the vitality of the open source community is the key to making tidb better used, and also the key to realizing the positive product cycle and business cycle.At present, tidb has actually participated in nearly 1200 contributors. Leveraging talents through community leverage is the strength of open source.
And how to operate the open source community? PingCAP has a set of methodology, and Liu Qi thinks it’s like running a community like jitter, tiktok, etc.
First, we should first find the most capable contributors, form product prototypes, gradually refine the division of labor, and finally form an accelerated cycle.
TiDB was the first mock exam in the first two years. PingCAP’s own team had some ideas for a module, and then assembled the most loyal and enthusiastic experts in the TiDB community by grinding the products on the Saturday morning’s Infra Meetup to form a product “blank”.
With more and more contributors, the division of labor will be more detailed, and the construction speed of modules will be faster and faster. When the product begins to enter the enterprise application, more feedback can be collected from the enterprise, so as to iteratively improve the product, which enters a big cycle. From 2018 to 2019, after the large cycle is fully started, the iteration speed is getting faster and faster. “We really run at high speed for only two years,” Liu Qi said, which is the embodiment of the accumulation of open source community.
Second, set up feedback and incentive mechanisms so that each contributor can feel respect and recognition and spread the energy of the community.
The tidb community divides contributors into several levels: contributor – committer – maitainer. They get the achievement certification medal just like upgrading and fighting monsters. Most of the codes of many community contributors are based on the sharing of chivalrous spirit. Not everyone can get a financial return. For example: some developers pursue geek spirit and perfectionism. When using open source products, they will take the initiative to improve bugs or add new functions; Some hope to share and help others, so as to obtain a sense of achievement.
Pingcap will also design some very interesting surroundings to prove its identity. For example, the iconic red mug must be a contributor. The distribution standard is very strict. Chen Yu, the a-round lead investor of pingcap, the managing director of yunqi capital, also contributed some codes on tidb and got the red mug.
Third, regularly hold activities with technical influence and in-depth users.
Before the epidemic, offline Hackathon was a regular activity held by pingcap. During Hackathon, engineers will enter the crazy hack eat crazy hack eat mode and write code until the early morning.
Another major event in the vibration technology circle is to write a technical book on tidb in 48 hours, and a work covering tidb architecture principles, best practice cases and ecological development was born. A total of 421 commits and 199 PR were generated in two days.
These sounds crazy, in fact, it is not difficult to understand. Back to the scene at the beginning of this article, pingcap screened a group of open source enthusiasts. While giving back to the community, it also formed a talent moat of pingcap.
The most fundamental is to answer a question: why is open source popular?
The foundation of the enterprise software industry is the platform and developers. Open source software is naturally friendly to developers: tens of millions of developers around the world use the open source platform to access code, patch applications, promote projects and participate in communities.
As Silicon Valley venture capital a16z said, open source devours everything. The figure above shows the number and scale of financing in the field of open source from 1988 to 2018: 200 companies that take open source as the core technology, with an overall financing scale of more than US $10 billion, and the financing amount has become larger and larger in recent ten years.
In the future, open source projects will only increase. Open source is no longer just RedHat, elastic, databricks and cloudera. It will also be Facebook, airbnb, Google and any other commercial company with open source as the core technology stack. The market and possibility of open source software are far more than currently imagined.
Although there is hope, the problem is also very realistic: can the commercialization of the open source community be replicated in different fields? Where is the ceiling of the market? How many similar companies can this market accommodate? Snowflake is still losing money. When will such companies make a profit? Although public cloud based relational databases are developing rapidly (accounting for 46% in 2024), the cloud process of Chinese enterprises lags behind. Will this affect the growth rate of relevant enterprises?
Time will give the answer.