Interview with technocrats or not “bandit spirit” CTO Yi Guan, great Xia Guo, pursuing the soul journey of data

Time:2020-2-18

Interview with technocrats or not

Technician: Yiguan CTO Guo Wei
Edited by: segmentfault coo Jiangbo Nadia

Looking for talents like looking for a daughter-in-law

Nadia: can you briefly introduce your previous study and career experience?

Guo Wei:I am from Beijing. I graduated from high school and studied in Peking University for seven years until I graduated from graduate school. What I learned from the graduate stage was data. At that time, it was not called data mining, information and signal processing. The title of my graduation thesis was “customer churn analysis of process neural network”. At that time (2005), there was no ai artificial intelligence. I went after graduationTereData, the largest data warehouse company in the United States at that time, later went toIBM, CICC, Wanda, Lenovo, focusing on the construction of big data platform and the implementation of project application. I still feel that the systematic management and control process of big companies in data innovation application is more complex, and also inspired by Yu Yang, the founder of Analysys,Joined Analysys in 2016.

Nadia: what is the state of Yiguan in 2016? It seems that at that time I was impressed by Yiguan or a company that made analysis reports?

Guo Wei:At that time, Yiguan started to empower customer data insight through data products, and started to build a technical team, but the whole team’s ability was not up to standard. Even at that time, the market didn’t quite understand what the technology background was to do with Analysys. So, I have a stage in the middle, I have readjusted the technical team and focused more on data architecture, algorithm and other directions.

Nadia: when you joined, was the business model of Yiguan clearly planned, or did you feel the stone to cross the river?

Guo Wei:At that time, Yu Yang, the founder of Analysys, always insisted on making products and technologies, but we didn’t know which product and how far the product line could be made, but we took the first step through Analysys Qianfan (mobile Internet product benchmarking analysis platform).

Nadia: what challenges did you encounter after joining Analysys in 2016?

Guo Wei:You may have seen the “entrepreneurial curve”. At first, it rose very fast, then fell precipitously. After a period of silence at the bottom of the valley, you have to fall again to slowly rise the curve – in fact, entrepreneurial experiences are all like this.

Interview with technocrats or not

When I first entered Yiguan, everyone was optimistic that Yiguan had brands, ideas and data sources, but lacked technology. At that time, I was also very optimistic. I thought it was just to get the technology up, which was simple. But when I went to do it myself, I found that the mountain we had to pass on the way to upgrade to a technology company was really big, and there were many problems.

I haven’t written code for many years before, but I couldn’t help it at that time, especially for some core modules. In 2016, Yiguan Qianfan was able to undertake the analysis of nearly 100 million monthly active data of users. At that time, a big challenge was the problem of concurrent data reception. When Qianfan started to serve customers, a serious accident happened. For three days, the system was down and there was no data. I was particularly impressed. In June, the CEO came to me and said, “you are very hard, but how can we solve this problem?” At that time, I didn’t know how to solve it. Later, I couldn’t. I learned Lua language again, and then wrote the code again in two days. The problem was solved. This is the state at that time. The whole technical team didn’t have enough understanding of the combination of technology and business. I also found out from that timeThe most important thing for entrepreneurial teams to recruit people is talents

Nadia: the entrepreneurial team is not easy to recruit people, is it?

Guo Wei:Dai Lidong, the core contributor of our open source project, dolphin scheduler, was convinced by how many meals I went to him to eat. He is the first big data oriented talent. I think it’s good to start with him. I’ll start with himLooking for talent like looking for a daughter-in-law, and now it’s also much longer to have a party with the team and talents than to spend time with the family. It’s like a wife with nothing,Exchange true feelings for true onesThe team has grown from less than 1, 3 and 5 to more than 100 people.

Nadia: what kind of talents do you prefer when you recruit?

Guo Wei:I think the first thing to attract people is to see the values. Do you really want to do the data? When it comes to data, it’s very important to see whether the eyes will have light and light. The second is to look at the potential. Is it OK for logicality, learning ability, reading ability and communication and expression ability? The third is to see what the ability is now.

It doesn’t matter what it will be, whether it will be Hadoop or whether it can understand big data. If you really want to do it and are willing to learn it, we are willing to give you such an opportunity. Not every enterprise has such a big six billion monthly live data for you to learn, and no customer like China Merchants Bank and Dangdang can let you practice.

Great Xia Guo’s feelings of open source and the change and invariability of Yiguan Technology

Nadia: from 2016 to 2019, you joined Yiguan for 3 years, helping Yiguan transform from an analysis and reporting company in the public perception to a big data company driven by product technology. What has changed in technology?

Guo Wei:There are so many changes. The following picture shows the change of technology architecture of Analysys from 2016 to 2019.

Interview with technocrats or not

In terms of products, from 2016 to 2017, our main products are Yiguan Qianfan and Yiguan Wanxiang, which are essentially Yiguan’s own big data platform. By 2018, we found that our big data technology can help more enterprises. We began to do technology output and started to have Yiguan ark. Today, it has become an intelligent user operation product suite including intelligent analysis of user behavior, intelligent operation and open technology.

Our original things are very large, 6.8 PB, hundreds of servers, not every enterprise can use them. So in 2018, we began to simplify such complex big data components through a new architecture, and we hope to help the enterprise to use the easy big data platform.

This new architecture is called iota (big data Iota), its core idea is actually edge computing – big data used to store all the data in the cloud for computing, but now our mobile phones are getting more and more powerful, so my idea is to do some computing in the mobile phone first, and only store and query in the cloud, so the efficiency will be greatly improved. Many big data in that year According to the cluster, it can be done by concentrating on one server.

Abstract our own technology into a product and technology that users can use through new architecture iteration (Analysys ark), which is actually what we are doing in 2018.

By 2019, we have changed again. In the past, Yiguan ark only did user behavior analysis. Today, it has realized the intelligent operation function. Later, we found that many users want to do some secondary development based on their own needs. We have turned the easy view ark PAAS from a product to a platform. For example, if you want to make recommendations, we have collected all the relevant data through the SDK, and reorganized the unstructured data into structured data. Based on these data, your own analysts can make some recommendation engines, user portraits and so on.

In 2020, we will further make this platform more stable. We hope that in addition to us, more developers can develop something on it.

Nadia: so from next year, we will also do something about developer ecology?

Guo Wei:In fact, we are also doing this year. Yiguan ark itself is a commercial product, but in fact, we have a free version. This version of privatized and stand-alone server is free of charge. It’s called Argo. Before this, there is no free product for user data analysis of privatization deployment in China. Private deployment, data is put here, users can rest assured. Ordinary developers can directly connect this product to their own background, a server, and you can start to do user behavior analysis. In the future, we hope that all kinds of new gadgets can be developed based on Argo. We encourage you to develop all kinds of new gadgets. After development, you can use them by yourself and share them. We help you sell them. We also welcome you to open source based on our products and make open source components for everyone to use, because Argo It’s free.

Nadia: I just mentioned that iota architecture can achieve more efficient data processing speed. Can you explain its value in specific application level?

Guo Wei:For example, take segmentfault. Recently, users registered in a channel of SF for more than seven days have not logged in in the last week. You want to send an email to make a recall. Originally, you need to provide a requirement for the technology. It needs to run SQL, but you may have to run ETL script for two or three days to get it to you. When the company needs more, it may need to schedule. But in the Yiguan ark based on iota architecture, you only need to check some conditions. Within three seconds, the data will come out immediately.

In a few seconds, we may have to look up 10 billion pieces of data, and get a result, which is very difficult technically.So the core problem that iota architecture solves is to make a complex query in a large amount of data, and retrieve the data in seconds.So what comes with it is to help usCivilian data capabilities, so that roles other than data analysts can also use this product to do some data analysis without big data engineers writing scripts for you every day.

Nadia: when it comes to data processing efficiency, have you mentioned the concept of “data River” before?

Guo Wei:Data river is actually an abstraction of iota. In the past, cloud vendors often referred to the “data Lake” – put all the data in it, and its advantage is to save it. But when you really want to use this data, the cost will be higher and higher. The more data you put, the more data you will slowly become a data swamp. What we do is to let the data flow. Through the iota architecture, the data flows directly from the production end to the consumer, no longer passing through the data lake, no longer requiring engineers to write ETL.

Nadia: so what has remained the same for the past three years?

Guo Wei:We have two basic points that remain the same, which I call technological valuesOpen SourceThe two isClouding

This yearOur open source project, dolphin scheduler, was selected by the Apache foundationBut it is not our first open source project. We have opened many projects before, which is the result of our slow operation and embrace of open source.

Interview with technocrats or not

Data has soul, I will use this life to pursue

Nadia: now many companies are talking about that they are data-driven, of course, most of them are fake data-driven. Among your customers, is there any company that you think has really applied data to a very good case?

Guo Wei:Among our customers, there is a first tier joint-stock City firm that has done a very good job. In fact, even Internet companies may not really be able to drop the word “data driven”.Data driven is a management thinking, not a tool to get things done.

We’ve seen this bank use the Polaris indicator (the only key indicator that all employees should pay attention to), which is its monthly work for China Merchants Bank, and its salespeople should carry it. When using, we can feel that their app is really easy to use, because there is a clear analysis path for each click of app and the process of each functional module column. Many Internet companies can’t do this. It has something to do with the management stage and maturity of a company.

Nadia: big data has been a hot word in recent years. How do you personally understand big data? Looking to the future, what is the real value behind it?

Guo Wei:I have a motto calledData has soul, I will use this life to pursue.

In terms of the whole technology and industry, I think it will become more and more popular. Because people can turn the real world into numbers, information and data through various digital means, and then make relevant analysis based on this.

When I first did Bi, the reports were all generated by internal CRM. Later, we took the click of app for analysis. Now we have face recognition, offline action recognition, voiceprint recognition In fact, all behaviors of a person online and offline can be collected. Data digitizes the world itself, so how to use it? In fact, the usage of every industry and different track is different. In the future, 5g and IOT will be added Maybe you are just like the hacker empire in this conference room. If you pose and say a word, these services will arrive. In fact, these are data applications.

For big data companies, the challenge is quite big. Many data companies have become project-based companies, so they have no way to do better growth. Around 2016, there are about a dozen companies in China that are making products of the same type. Now there are only two companies that are doing well. Yiguan ark may have many competitors now, and only a few will be left in this track in the future. In the future, Yiguan will solve users’ problems according to their actual needs,Big data is not a very easy track. Every company that can stick to it deserves respect.

The bandit spirit of CTO: management is not to balance the two benefits, but to balance the two harms

Nadia: finally, let’s talk about entrepreneurship. As a technologist who has been on a large platform, what do you think of entrepreneurship?

Guo Wei:Entrepreneurship itself is a life-long business. Even though Yiguan is still trembling now, because in fact, many times you don’t know what is right.Entrepreneurship is a very challenging and interesting thing. Its charm and risk come from the unknown. The situation you face every day is unknown.

As for being a technology partner, I suggest you don’t start a business before you have business thinking. On the way to entrepreneurship, you will findTechnology is actually the necessary and insufficient condition for Entrepreneurship。 It’s not good to start a business without technology. It’s not good to just have technology. Because in the face of business logic, no matter how powerful your technology is, if you can’t provide value to users, it’s also pseudo technology. So for technology partners, business thinking is more important than technical ability.In fact, it’s not necessarily a technical bull, but you have business thinking and can communicate with people, so you may become a good technology entrepreneur.

Nadia: everyone calls you “great Xia Guo”. Why do you say that many CTOs must have banditry?

Guo Wei:To give a specific example, many people think that management decision-making is to choose the best. In fact, most of the choices are not to balance the advantages with the disadvantages, but to take the advantages with the disadvantages. There are two Xiangs in front of you, one is bigger and the other is bigger. Do you want to eat them? If you don’t eat it, you have to double it immediately. Sooner or later, you have to eat it. Then you have to make a decision.

Banditry is that you may not know what the future is like at all. These two are not very good, but you must be bold to choose one that is not so bad, and then encourage the team to be PK at the same time。 You need to carry this one down instead of researching this one and looking at that one. I don’t have time for you to see, just do it!

Nadia: so this may be the key point for many technical people to become a leader.

Guo Wei:Yes, because technical people are generally very objective and logical, but to be a CTO means that you often have to shoot a plan that you think is not good when there is no logic. Junior managers may not know that much, but senior managers must know that:There are not so easy things in the world. They are all difficult things. It depends on which one you choose.

Interview notes

At the end of September this year, I met for the first time with Guo Wei, who is known as “great Xia Guo” and CTO of Yiguan, in the Beijing Office of Yiguan. After three months, the “segment fault think no” technical person interview column was finally launched in the new year. In the past three months, second computing engine 2.0 has been upgraded and launched. The Argo community of Yiguan ark has harvested open source code from users. Opentech’s open strategy will promote a more open technology ecology through open PAAS, open architecture and open community.

The data industry is recognized as a combination of “dirty work, hard work and hard work”. However, after 19 years of “going into business”, there is no sign of fatigue in him. Love to let him work hard and have fun, two hours of chatting communication, from technology to entrepreneurship, from personal experience to company business, his sincerity, preciseness, love and chivalry show no doubt. If it’s a snowy evening, I have the impulse to say, “great Xia Guo, can I have a drink?”

(end of this paper)


“Interview with technologists or not” is a new online interview column of segment fault. We hope to pass on their real growth experience and love for technology to more people and help them grow. This column was interviewed and written by segment fault coo Jiangbo Nadia.

Seeking report: [email protected], applying for Reprint: [email protected]