Yiming Zhang: Context, Not Control | Code Class 2017

Successful entrepreneurs are sure to witness their companies growing from small to large. As a CEO, you are both a creator and an executive, which requires that you continually “update” yourself, in order to appropriately and in a timely manner, handle the management challenges brought about by your company’s development.

At the 2017 Code Class Annual Meeting, Founder and CEO of Toutiao, Yiming Zhang, shared some of his insights with Code Class members on the topic of, “As a CEO, You Must Avoid Rational Conceitedness. It’s about Context, not Control.

Mr. Zhang’s full speech is as follows:

At last year’s Code Class, I talked about how we must do our best to recruit exceptional talent. What should you do after having gathered together a group of extraordinary talent? This will be the focus of what I will be sharing with you this year:

How do you establish an effective organization? How do you grapple with the various challenges you face on the management level as your company grows bigger and bigger? These are some issues we are discussing and deliberating continually at Toutiao throughout its growth process. Presently, we are inclined towards a “Context, not control” solution when it comes to problem-solving.

Allow me to draw an analogy in order to illustrate the difference between context and control.

There are two different methods by which computers process a task. The first type is when a supercomputer uses a calculator to process a very intense task. The second type employs distributed computing, relying on many machines to collectively handle a task simultaneously, thereby splitting up the task, as while as the resources required for the task. There are also two different enterprise management models which respectively share similarities to the aforementioned computing methods.

The first type views the CEO as a supercomputer. The CEO shoulders the responsibility of outlining strategic design and proposing strategic plans. Next, each layer of tasks is split up and distributed for execution. If any problems arise during the execution process, they are reported back to the CEO who will summarize the information and reassign the tasks. There are many processes, approval steps, and management mechanisms in this model. In the past, many enterprises have adopted this management style; its hallmarks include devising strategies and operational flow control.

The second type involves more people and more decision-making, allowing for more ideas to work their way from the bottom upwards, as opposed to strategic allocation taking place from the top and trickling downwards. This process requires more people to make collective judgments based on the context of the whole picture, rather than executing orders.

To break it down to the basics, what is context, what is control?

Context refers to the gathering of information necessary to make a decision. These include elements such as what the inherent driving principle is, how the market environment is making out, how the overall industrial landscape is looking, what the level of priority is, what the degree of development will be, as well as business and financial data. 

Control includes elements such as committees, orders, splitting up tasks, summarizing, processes and approval, among others.

So why is our preference for “Context, not Control”?

In our view, there is always an element of risk related to control. When humans judge their ability to control themselves rationally, they experience a type of hallucination, and it is more so the case with people who are both intelligent and rational. Such people are often “rationally conceited.” CEOs often have experiences of being successful, especially in the early phase of their companies. Moreover, they often do not have superiors above them to answer to, neither are they challenged by other people and therefore they are susceptible to seeing themselves as brilliant and imposing. Nonetheless, everyone has blind spots, and since the industry is constantly evolving, our wealth of knowledge becomes relatively limited.

Sometimes CEOs have the misconception that the methodology that they bring to the table is exceptionally good and that their models are uniquely stylish. They wish to implement them or promote them within the confines of their company. However, they fail to realize that there is a vast gap to be bridged between abstract knowledge and forms. Most often, rationality is only applicable to knowledge abstraction, yet when it comes to the solving of concrete problems, it might not provide any substantial benefit. Of course, we are not negating the role of rationality. Instead, we are attempting to avoid the perils associated with the over-amplification of rationality.

Grand strategies passed down from the top often have a tragic ending, and there are many real-life examples of such in the business world.

One example is Windows Vista, which was a project pushed by Bill Gates, according to his own technical concept. Bill Gates had many grand notions concerning Windows Vista and planned for its online launch in 2003. These theories sounded lofty and ground-breaking, but it was not until 2006 that the product was made officially available online. Between 2003 and 2006, they reconsidered the whole concept, lowered their target, and revamped the entire plan before its eventual debut. Not immune to blunder, Steve Jobs made the same mistake. The first time he left Apple to NeXT, he proposed an ideal computer model, including a stylish operation system, a completely object-oriented language. Nevertheless, in the end, sales did not take off. There is also a similar example with EZ Pod in China, which once held high aspirations. Their eventual failure can be attributed to a number of factors prevalent at the time, such as miscalculations with regards to the entertainment industry, Internet bandwidth or the policy landscape.

Besides bringing to strategic problems, control can also cause an enterprise to react sluggishly as a result of pursuing a sense of control. All of us seated here are CEOs. We can add together the grand sum of all of the expenses, contracts, and offers which require approval on a daily basis. Suppose you have 15 items that require approval on a given day, then that would equal a ballpark figure of some 5,000 in one year. How many of these have you seriously thought about before approving? How many items will have undergone serious deliberation by you? Alternatively, is the existence of these items solely based on a feeling of control? So as to say, if expenditure has been examined and approved, then capital will not be spent inappropriately. Comparatively speaking, could your subordinates or perhaps others do a better job of the approval process? I believe so because they are involved in frontline decision-making and have sufficient external information. Due to the fact that CEOs have a limited energy capacity, when a considerable number of approvals are postponed, it increases the time of completion for many items by a day or two days, or more.

There is a wrong solution to these problems that arise as a company scales – premature BU (business unitization). This type of scheme will lead to a number of problems:

First, a lack of cooperation between departments. For example, when BUs personally attend to a public relations crisis and recruit their own engineers, they do not need to involve colleagues in the marketing or technology department. This means that these departments do not need to cooperate with each other, or you may say that their cooperation is deteriorating, because they do put any effort into learning to adapt to each other.

Secondly, there is redundancy between departments and a lower level of professionalism. For example, the recruitment standard of a single BU is inadequate when hiring an engineer and the scale of an engineering team is also too small, which means that there is not enough mutual learning taking place. Other issues include limited improvement and enhancement, reduced level of professionalism and departmental redundancy. To a CEO, this all feels more like the job description of a contractor. I will pass this assignment on to you, and you take care of it. I will not be involved in the process, all I want is results. If this persists over an extended period of time, company culture will deteriorate.

Of course, there are some exceptions. For example, suppose there is a relatively independent or very mature business, it does not necessitate internal support and cooperation within the company, and business unitization can take place. The significance of the existence of a company is to allocate work and facilitate collaboration in order to conduct business activities within the company. One must ensure that the costs of internal cooperation do not outweigh market transaction costs. In essence, large numbers of uncooperative BUs should not exist in this enterprise. Premature business unitization is a relatively prevalent erroneous solution. Many companies begin to establish many subsidiaries too early, or they split the company up into many different project teams. Sometimes they even go so far as to set up a part of the company as an independent business and also finance it independently. In my opinion, these solutions often turn out to be weak and lazy solutions, because they can sidestep issues such as cooperation and communication.

In contrast with control, what are the benefits of the management model that emphasizes context?

(1) First, distributed computing allows more people to use more CPUs to compute, allowing more people to participate in decision-making, and leverages collective intelligence. As a member of management, you need to spend a mere 30 seconds when it comes to granting approval and decision. Others, on the other hand, might spend three hours, as they might conduct further research before making the final judgment.

(2) Second, execution can be expedited. It does not require that you report to each layer of management, neither does it require that you report to a certain department. There is also no need to queue up in order to see the CEO. Therefore, you can respond to matters in a timelier manner.

(3) Third, there is ample external information input. In the control model, all information must pass by the CEO. The CEO then redistributes this information. To quite a significant degree, the CEO becomes an interface between the company and external entities. In comparison to solely relying on the CEO to singlehandedly interact with the outside world, comprehend market sectors or macroeconomics, if you were to allow more colleagues and more supervisors to engage with the industry directly, the information gleaned would be more sufficient, and offer a different perspective.

(4) Four, creativity inspired by a sense of participation. When something is done repetitively, if the staff not only knows what they are doing, but also understand the reason behind what they are doing, work will become more meaningful. It will no longer be just following random orders. This is quite beneficial in allowing free reign to employees’ creativity.

(5) Fifth of all, scalability. The forms in which the construction of context is expressed might be in the external system, or in knowledge sharing documents. These are all things that are reusable and scalable. The time and energy of CEOs and their management teams will all eventually encounter bottlenecks. When depending on physical energy, brainpower, and endurance to solve problems, you will encounter bottlenecks, and there is no scale effect.

Of course, sometimes there is also a need for control. For instance:

(1) Emergency situations and key projects. For example, a major PR crisis that requires a swift response. It is the same thing when it comes to key projects if your competitor is already encroaching on you. To conduct a distributive discussion from the bottom up at this time would leave you with insufficient time to solve the dilemma. The window of time is but a fleeting moment, therefore when dealing with emergency situations and key projects, control is a necessity.

(2) Establishing a business and the early stages of a new department. If a new department is formed or a new executive assumes office, and there is trouble meshing in with the rest of the company, then that also necessitates control. In the early stage of launching a new business, when more support, equipment, and resources are needed, a CEO is also indispensable in terms of unifying coordination and spearheading progress. 

(3) Mismatched placements. If there is a substantial disparity with regard to the company’s ideas and those held by someone in a certain senior position in the company, then this requires that a superior utilizes control to intervene.

Now, why do these types of problems tend to pop up after a company has already been developing for some time, rather than in its inception stage?

Usually, in the early stages of a company, the CEO is viewed as an expert in business. Both company affairs and industry conditions are less complicated. Therefore, it is most effective for the CEO to be in charge of decision-making personally. Nonetheless, as the company grows, the CEO’s energy and attention are preoccupied with many different things such as public relations, financing and external activities, among others. The organizational structure itself can be enough to deplete managerial energy. Moreover, the business landscape becomes complex and diversifies, so the CEO is no longer the expert, even to the point that the CEO might also no longer be the sharpest person.

We require CEOs to learn and develop quickly. Supercomputers are becoming increasingly powerful, and their scope of knowledge is ever broadening, yet the energy level of a human is limited, and in many aspect, CEOs can no longer perform the way they did in the start-up phase. 20 years ago, Bill Gates was an exceptional software engineer, now 20 years later, if we were still to employ his concepts and ideas in conducting large-scale projects, they would prove to be very limited. Of course, some companies do not have this problem, because the industry in which they find themselves in is more stable, with relatively little innovation, and all they need to do is stick to tried and true recipes, for example, Lao Gan Ma chili sauce.

In summary, we believe that a good organization should include:

(1) Exceptional people. You need a distributed processor and not merely an executor. Each distributed computer must have the ability to make judgments and must be intelligent.

(2) A management model using ample context and limited control. Each person has his or her role to play, to maintain control over all the contextual information and make business-related decisions. When needed, they must make small-scale interventions.

With these two points, you will be able to ensure that transaction costs within the organization will be kept to the bare minimum, all while making high-quality decisions.

Based on this notion, when we encounter problems in our company we usually have the habit of asking whether or not we have sufficient context, instead of increasing control. For example, if problems arise in a given development, initially, we will not consider having someone higher up in the company deal with the issue. Instead, we will weigh the issue based on whether there is enough context or not. We will consider whether or not we have brought industry conditions, business-related data and past instances of failure to the CEO’s attention.

As a manager, when you are able to make better decisions than others, is it because your ability, or is it because your context is more adequate? Is there information asymmetry? If you observe carefully, you will discover that sometimes, managers take advantage of information asymmetry to exemplify their value. In light of this, it is crucial that we first lay the foundation for the construction of context within the company. However, this is easier said than done, as it entails much tedious work regarding communication, management, and product technology.

From a specific operational level, we have adopted some practices which I would like to share with you:

(1) Reduction of rules and approvals. We do not allow departments to casually set rules, although rules are a must. Still, we hope that regulations can be as simple as possible, which is why we do not allow pages of rules which are extremely difficult to carry out. We try to reduce examination and approval procedures and hope to avoid them as much as possible.

(2) Flexible organizational structures, rejection of territorialism and flexibly adjusted reporting lines. We want everybody to realize that reporting lines are a means by which information is reported. Provided the business requires so, adjustments can be made at any time. If there is a project of great importance, we might necessitate the support of our colleagues in the marketing department to give a helping hand. During that period, the supervisor of that project will also be the supervisor of the colleagues in the marketing department.

(3) The weakening of hierarchy and titles. We encourage young people to put their ideas forward. The first time I assumed the role of CEO was when I was 26. I am sure that 26-year-olds in our company have a lot of hands-on experience under their belts, as they have received excellent educations. As long as they are provided with the proper context, they will also be empowered to make the right decisions. To avoid the base node being suppressed by a sense of formality, we weaken the hierarchy, first of all, by disallowing titles such as “boss,” “CEO”, and “guru.” As soon as different forms of address come into play, it stifles the flow of ideas. People have the propensity to give precedence to what their “mentor” has to say before they will voice their own opinions. We do not have to deal with all the differences in preferential treatment, which is commonplace when titles are used. For example, issues such as a person of such-and-such ranking should be fitted with such-and-such a computer, or a person of such-and-such ranking should be equipped with such-and-such a desk. This will only give rise to a sense of hierarchy and will also affect certain colleagues when it comes to expressing their opinions.

(4) We encourage transparency of internal information. We encourage group chats, allowing sufficient communication to take place between departments, instead of one-sided communication with the CEO. We discourage one on one communication since we believe that one on one communication yields low effectiveness. If a new colleague who has recently joined our ranks or a senior executive wants to talk with me on a one on one basis, I will often tell them that they can write me a note, but they must first send it to other members in the company whom they will need to collaborate with.

The OKR (Objectives, Key Results) of our management level is made accessible for viewing by subordinate staff, thereby allowing everybody to know what is going on, why something is being done and what people in other departments are doing.

The formulation process of OKRs is not split up into a top to bottom format. Instead, it involves everybody mutually aligning with each other. You take a look at the OKR of your superior, the OKRs of other departments or the OKRs of your peers in order to get an overview of what are the most pressing tasks are that await completion in the company, and what you can do to assist. As much as possible, quarterly meetings also allow greater participation among more, instead of senior executive meetings which exclude the participation of lower ranks. We often conduct face-to-face meetings with the CEO, which gives employees the opportunity to raise questions and gain an understanding of company developments.

(5) We believe that the support by an internal system is necessary in order to establish context fully. We have nearly 100 individual internal tools to develop teams and attempt various kinds of tool trials. For instance, we have developed our own OKR system which is integrated with the internal IM (Instant Messenger), so as to allow different counterparts to see what each other is doing. These essential tools, first, make things easier for people, and second, they are scalable. When newcomers join the company, they can quickly adapt to the OKR system and view internal data as well as obtain information internally. They can also come to the realization that not only do they have the right to acquire information, but also share in the responsibility to support related work.

In our view, in practice, this implies the development of a company as a product, to allow the context inside the company run more efficiently, and fortify the distributed processing power of the system.

This the end of my sharing, thank you.