Speech | tan
Finishing | xin-long zhao, tail end
Tan, Baidu chief architect, Baidu search company technology committee co-chairman. His main research field is distributed system and search engine. He is the main designer of Baidu BVC proxy computing and Matrix private cloud, and has won baidu’s top award twice. Presided over the design of baidu’s new-generation search architecture, which greatly improved timeliness and computing scale; At the same time, he led a series of major innovations in Baidu search, such as speed search and whole site HTTPS, and was also the overall leader of Baidu MIP project.
This article is compiled from tan’s speech delivered at the 2nd GTLC Global Leadership Summit hosted by EGO.
How can a programmer lead a major project without a team?
I am very happy to discuss the topic of technical leadership with you. First, I would like to introduce myself briefly: I work as an architect in Baidu Search, and I am a programmer in every sense of the word. This is something that I think is very special among all the participants. For more than a decade of my career, I have maintained a purely technical professional serial identity, and speak the language of “I don’t bring teams.”
Although I do not lead a team, I am directly responsible for and leading many of the major technological innovations and product upgrades in search. It’s not because I’m a genius, it’s because the Baidu search team has a unique technology culture, which is very distinctive and very attractive. This unique technical culture enables engineers who do not have technical management functions to take on technical management responsibilities when the organization needs them and make unique valuable contributions.
And that culture, or mechanism, is what I want to share with you today — the non-authority technology management mechanism.
What is the non-authority technology management mechanism?
What is a “non-authority technology management” mechanism?
First of all, let’s talk about non-authority technical management. I cobbled together a phrase that translates simply as: You’re in charge of something, you’re in charge of a technology, you’re in charge of a project, and none of the project members report to you.
This sounds particularly strange, because management is usually based on authority, otherwise many direct and effective measures of adjustment can not be used, such as pay and performance, for example, when the project team is required to work overtime, it is directly stipulated that whoever does not show up will be deducted. I did not do this, I could have talked to each other.
Why can’t the technology field avoid non-functional management mechanisms?
Whether we like it or not, it is a very sad fact that non-executive management is almost inevitable in the field of technology. The reason is simple: authority must be strongly related to organizational structure, which is set and optimized by companies and departments for business purposes.
As we all know, the only constant on the Internet is change. When new opportunities and new times come, there will be great conflicts with the existing organizational structure.
Many of you have experienced the migration of the Internet from the PC era to the mobile era. At that time, companies typically had a division called Mobile Products, which was dedicated to mobile. But most of the company’s business was definitely in PCS, and the main technology assets were still in the PC division. There’s going to be a lot of cross-team technology integration between PC and mobile in order to get the most out of this asset and grow the business quickly, and there’s going to be a lot of pain. Late mobile goes mainstream and a radical architectural overhaul solves the problem. We are now saying that the age of artificial intelligence may be just around the corner, so this process will go through again.
Why are non-ex officio mechanisms important?
Since it can’t be avoided, the reverse, and more optimistically, is also important to organizations and businesses for three main reasons.
The first reason is that it can well break through the limitations of the current organizational structure and solve the problems just mentioned.
The most straightforward solution, though, is to restructure once and for all. However, those who have experienced it know that it is more painful to adjust the structure, and by the time you are done, you will be gone. And there’s a degree to it: when do you start adjusting? If it’s too small, it’s not enough. Often wait until feel this matter is very big time, has missed the opportunity. If, at this point, everyone is more used to working without authority, it will be easier to break through the limitations of the existing organizational structure. So, I say it is a more flexible and soft way.
The second reason is that it helps keep the technology advanced. Modern society or companies are strategic organizations with a high degree of division of labor. If you invest a lot of time in people management, you must be backward in the technical field. So sometimes it’s good to have cutting-edge engineers do these things.
The final reason is that it brings the value of autonomy. If you ask an engineer to do management, he will feel like a farmer has turned over to be the master, “I can do a lot of things,” and their motivation, satisfaction, and loyalty will increase dramatically.
How to build non-authority management mechanism?
Since the situation of non-authority management is inevitable, and the construction of corresponding mechanism is very important, it is better to take the initiative to build such mechanism and make it the new normal, rather than passively avoiding it.
So how do you build it?
To build this mechanism, we must first lay a good framework.
The first step is to make a name for yourself.
A brief History of Mankind, a popular book recently, tells a funny story: The ancestors of modern humans are only one species in the genus Homo, but there were many other species at that time. All of them could use tools, make fires, talk and even be stronger than our ancestors. What caused all these people to get crushed? And then one day, our ancestors developed a very distinctive skill called gossiping, much like we do today. Gossip skills lead to more complex relationships. Because of their common likes and dislikes, more and more people get together, and then they develop a second skill called storytelling, especially the ability to make up something that doesn’t exist in the natural world, and get everyone to identify with it, believe in it, and spread it. The tribe is a fictional concept, but because of the imaginary community, more people can get together and work for the community. Now we want to create a new mechanism, a new system, and we must first create a hypothetical community through which we can do things. At Baidu, this hypothetical community is the technical committee. That’s the first step.
With a hypothetical community, the second step is to select the core team.
The core team is very critical, because in the early stage, they will be the main force to implement management, with a strong demonstration role, so we must pay attention to personnel selection. Three things are very important. First, technical excellence, second, understanding the business, and third, resourcefulness. The first two points are easy to understand, focusing on resourcefulness. Resourcefulness means the ability to use whatever resources are available to reach your end. What does that mean? Because without authority is without authority, but you may have other resources, you can convince big bosses to support you, you can create buzz — there are people who can see those points, who can use those resources, and that’s a very big difference.
After the core team is established, the last two steps need to be taken: delegation of authority and delegation of responsibility.
The first is empowerment. Although we’ve been talking about non-empowerment, to be honest, it’s really hard to be in tech management if you really have no power at all. Although we cannot grant him the actual rights in management, we can grant him the right to control technology, such as the right to Review the professional promotion of engineers, or Launch Review on the Launch of major projects. On the one hand, these things are very suitable for professional organizations to undertake, on the other hand, they can accumulate enough authority to carry out other things, people will listen to him. There’s a catch: we have to prioritize, technology has to serve the business, and the people who are really responsible for the business are the management, the managers. It’s got to be understood, it’s got to be balanced, it’s not going against it.
Finally, give responsibility. After granting rights, let TC give responsibility so that it can work. This is also the most critical of the four steps, and one that often goes wrong. There are two problems in this step: the first one is that the responsibility we give is too vague and too general. For example, your responsibility is to control the technology, but what is the control technology? The second problem is the credulity that engineers can be highly disciplined, so they do not check the results, and finally TC becomes a purely temporary organization responsible for reviewing projects. My personal experience is that it is necessary to have a very clear definition of what he or she is going to do, and check it regularly. For example, it is clearly defined that what TC is going to do is to periodically arrange all technical indicators and make technical plans, and this should be done every six months. As you do so, identify any problems, whether they are opportunities or challenges, and turn them into specific projects to follow up on, and finally review your project results. Only through very strict regulations and strict inspection, can we really put responsibility into action.
Actually land and practice
With this framework in place, the next step is repetition. No matter engineers or managers will be very uncomfortable, so we must do it repeatedly, deliberately, to make a regular.
Everyone has heard of the 10,000-hour rule, and many of the best people in the industry are great not because they are geniuses, but because they put in enough effort. The same is true for organizations. If you want to build new mechanics, you have to do it over and over again — at least a hundred projects. A lot of people have successfully moved from passive acceptance to active drive, where you tell them and they do it, and then they discover that the system works so well that they take the initiative to find problems and solve them.
The actual case
Here are a few specific cases to illustrate.
Case 1: Search architecture
As shown in the figure below, this is the search architecture. In terms of the process, the search business is divided into three steps: offline, online and front-end, each with a big manager. TC plays a very good complementary role in it.
From a TC perspective, the figure on the left looks at the technical directions from a business perspective, and the red one on the right is the business metrics system, which metrics are important from a business perspective. So with this diagram, you can do a couple of things, the first is to merge things like, for example, you have storage systems online, you have storage systems offline. Underlying technologies can converge in one direction. The second and more important thing is to drive business development and technology innovation in this way. Looking at the right half from the business point of view, is the current indicator system perfect? Should we add a new indicator? Which existing indicators need to be improved? Depending on the target on the index, look to the left. Are existing directional facilities reasonable? If it doesn’t make sense, we break the limit.
Case 2: Search for security
Many users will be hijacked when they visit Baidu search, some will jump to competing products, some will add a lot of advertising. In the original management division, there was no organization directly corresponding to this problem, so TC took the lead in the first place. After system sorting, it was found that most of the problems occurred in the network stage, and the most direct way to solve the problem was to change the search to HTTPS.
It sounds simple, but it’s really complicated. The most complicated part is that you must change all the resources on the page to HTTPS.
Search itself as an open platform, there are many third-party departments, and even third-party companies have resources displayed on search, so it is necessary to pull all other departments and even other companies to transform together. The organization expanded further to include both operations staff and products. It took a year, but we finally did it. For Baidu, there are many important products, such as Tieba, map, users will encounter these problems, so it should be to promote HTTPS transformation of these products together. We then put together a company-wide promotion team to remodel all of these products. That’s when the situation changed again, and we wanted all the websites on the Internet to switch to HTTPS to make the whole web secure. This process is relatively long, and the things to do are relatively certain, this stage becomes a physical team to promote. When we look back at the process, sometimes it’s ex officio, sometimes it’s ex officio, it’s very flexible. The point is, do it the right way at the right time.
Case 3: Search speed
A similar example is search speed. Speed is a fundamental metric for all products, and for search, speed is critical.
In the past, the search speed was divided into offline, online and front-end, including offline optimization, online optimization, online retrieval speed, and front-end optimization, front-end rendering speed. The problem here is that although people do a lot of things, they don’t work well. When we recognize the problem, we leapfrog beyond our respective functions and look at it from a global perspective. Once the problem is viewed from a global perspective, a solution can be proposed from a global perspective: when a user searches for key words, predict what he will search for and complete the search results in advance. For example, when a user searches for Andy Lau, I know he will search for Andy Lau when he types in Andy Lau. This plan is still quite clever, more important is the change of management mechanism behind.
The auxiliary mechanism
As you all know, management is very hard, and this non-authority technology management is even harder. Therefore, in order to achieve non-authority technology management, some auxiliary mechanisms can also be made.
For example, we can help design delivery manager positions and let them do the corresponding things; Another example is to help engineers without too much experience in project management, communication process, control and promotion, etc. It’s less painful for the engineers to start out with, and there’s a lot of assurance along the way. You learn and grow faster as you do things.
In addition, the cultivation of ability, after all, management needs management skills, generally speaking, this kind of training is only open to the management, not engineers, this is also a problem, so we should provide some mechanisms.
Summary and reflection
Finally, to summarize, this kind of non-authority technology management may seem strange, but it is unavoidable. At the same time, he is very important to the development of the company. If you want to build such a mechanism, you should first establish a framework; With a framework, it takes practice to make it the norm and provide as many supporting mechanisms as possible to speed it up.
Technology is overrated in the short run and underrated in the long run — as is the case with much of the world. If you can stay rational enough and patient enough, it will surprise you at the right time.
Q&A
Q: How to track the implementation and implementation of technical specifications?
Tan: We are technical people, so do things must be technology-driven. We are developing a tool to automate the process setting and ensure that it can only be released after executing the process. Naturally, you can make sure the plan is on the ground.
Q: This time, it is mainly about management in the technical field. You are already a senior engineer, and you manage through TC or promotion channels. How about other situations, such as designers or project managers, rather than technical specialists?
Tan: Similarly, you can set up a designer committee. I think technology is a very broad word, design is a technology, product development is a technology, it’s very similar. Technology may be easier to evaluate, but product and design are more difficult to evaluate. These are all professional. I think, as long as there is expertise, we can use similar professional committees to solve the problem.
Q: I have two questions. The first one is about whether TC employees are full-time or part-time. Part-time employees will have input and output, and whether full-time TC work will affect their career development. Second question, we are a fast-developing start-up company, and we also want to establish a technical committee. If we let young engineers into TC, will they be questioned in communication or authority?
Tan: On the first question, first of all, it is clear that it is a part-time job. It is impossible to have such a full-time job. In fact, he is not that full of work, and full-time would lead to disconnection and ineffective.
How to solve the problem of part-time job? Some people have certain technical ideals and hope to promote big things. Although it is not their job, their goals coincide with this, so they will spend a lot of energy on this. Find these people and do what they do best. On the other hand, some people have excellent skills, but without such ideas, there is no need to cram them in, otherwise it may backfire.
Second question, because organizations are still very concerned about technical influence, seniority is a very important factor to consider, but seniority has nothing to do with age. Some people want to do this very much, but the seniority is very young, how to do? We could have a working group down there, working as a group, with the most influential people in charge. This can play a better role.
Brilliant Open Web
The BOW (Brillant Open Web) team is a dedicated Web technology building group dedicated to promoting the development of Open Web technology and making the Web the first choice for developers again.
BOW focuses on the front end, on the Web; Analyze technology and share practice; Talk about learning. Talk about management.
Technology connects the world, openness wins the future, pay attention to OpenWeb developers, reply “Jiaqun”, let’s promote the development of OpenWeb technology together!
OpenWeb developers
ID: BrilliantOpenWeb
Technology connects the world, openness wins the future