Liang Rubo: How An Organization of Over 100,000 Employees Maintain its Vitality? | 2021 Code Class Entrepreneurs Annual Meeting

Liang Rubo, Co-Founder and CEO of ByteDance

“Managers are usually very much inclined to clarify company policies. This force, like gravity, tends to push any organization towards the establishment of even more rules. When it is the case, we must resist the pressure of the said force, or it will keep companies locked in a permanent rigid state.”

Recently, Liang Rubo, Co-founder and CEO of ByteDance, was invited to Source Code Capital’s “2021 Code Class Entrepreneurs Annual Meeting.” He delivered a keynote speech entitled “Keep Extending and Avoid Rigidity,” in which he shared some observations regarding how Bytedance maintains its organizational vitality and enhance management efficiency.

ByteDance currently has over 100,000 employees in more than 200 cities around the world. How to implement scalable and effective management in an extensive, diversified, and fast-growing global organization is without a doubt a challenge.

According to Liang Rubo, ByteDance’s practices include: 1. Establishing a basic management system to ensure administration efficiency; 2. Attaching importance to the cultural construction, and the alignment of ideas while increasing consensus and creating flexible rules; 3. Ensuring the implementation of management actions through a system and making good use of data to better assist the decision-making process; 4. Ultimately, it is necessary to make managers responsible for management decisions to ensure management effectiveness.

The following is the transcription of the full speech:

Hello everyone, this is Liang Rubo.

Zhang Yiming previously shared his management idea of “Context, not Control” at a past Code Class gathering. As an organization size increases, it also gains in complexity, which can lead to confusion. This can be remedied by the recruitment of new talents to the team, as well as providing more context to people already in the firm, instead of adding more rules and processes which might in return create more disarray. These notions are derived from Netflix’s culture of “Freedom and Responsibility.” Today, I want to share with you some specific practices we adopt at ByteDance.

Rules Overload May Lead to Organizational Rigidity

ByteDance is now a very complex organization. We have over 100,000 employees in more than 200 cities around the world. Our colleagues work in different fields such as content platforms like Toutiao, Douyin or TikTok, business verticals that span both online and offline, such as Dongchedi, as well as corporate customer-oriented services such as Feishu or Huoshan and education market with Dalijiaoyu and so on. These businesses, by nature very complex and with different requirements, require the support of large teams. In addition, our organization is still growing at a very rapid rate, almost effectively doubling every year. ByteDance is a large-scale, very diversified business and a fast-growing global organization.

Zhang Yiming, during his prior Code Class sharing, said that if an organization does not choose to be small, it must then be ready to face the difficulties that are likely to arise with its increase in size, and therefore in the complexity of its organization. A conventional approach is usually to add rules and processes to maintain order, but this can lead to organizational rigidity. Why? Because the department in charge of making those rules needs them to be as detailed as possible in order to avoid problems. This dependence on rules diminishes the momentum of a company by preventing it from looking for more optimal solutions.

We are in an innovative industry, and the situation we face is often malleable and requires agility. If impeded by too many restrictions, employees will be unwilling to look for a better, or more optimal solution. When confronted by rules, people are naturally more inclined to follow them, rather than breaking them. Rules will also deter them from looking for new opportunities. Rules do actually inflate the cost of innovation by increasing the need for trials and errors, the fear of doing something against the set policies, and ultimately decreasing employees’ enthusiasm for innovation. Especially when there are major changes in the industry, when the company needs to move forward and inertia is not an option, the accumulated sum of rules and processes become a severe hindrance that will create major issues.

For us, on one hand, the large scale of our organization means that we need to pay attention to management efficiency to ensure that it can be scaled without creating excessive confusion. On the other hand, we must also pay attention to its effectiveness to ensure that the organization does not become too rigid, and allow ourselves to maintain a certain flexibility in order to find optimal solutions.

Management Solutions are Not Uniform

So, what was our approach?

First of all, you need to pay attention to cultural construction. We believe that through a company’s cultural construction, we can achieve more consensus in the company. With a common culture and philosophy to guide everyone, we do not have to rely on a heavy set of rules, and people are able to work in an equitable environment with similar standards.

I remember a meeting where someone suggested that our decision-making process was unclear and recommended that we established more rules. For example, if no decision was made or no agreement could be reached within two days, then a solution should be proposed by a higher level of management. After discussion, we believed that such a rule should not be set. If this was implemented, many similar decisions would have followed, reducing the overall efficiency of policies. We should emphasize on responsibility and pragmatism, and promote initiative to keep the company moving forward. It is far more efficient to highlight fundamental points than to establish more regulations.

In ByteDance, we have basic management mechanisms, such as budget, MBO, sequence levels, compensation specifications, performance appraisal, etc. However, all these processes are not uniform. They are informational, directional, and interrelated. Everyone has the liberty and the responsibility to make reasonable decisions based on the actual situations.

Due to the lack of homogeneity, many of the company management tasks are difficult, and few people are actually able to handle it. The decision-making process often requires a lot of discussion and adjustment, which is by nature not very efficient. It requires managers to make a lot of decisions by themselves, and it is impossible to hide behind a set of rules. We believe that if we were to address management responsibility through rules and processes, it would be sending the wrong signal.

For example, in the case of a performance distribution, a manager communicating with an employee about his or her lackluster performance might say something like, “Actually, your performance is good but according to our guidelines, you did not perform well enough.” In that particular scenario, the manager does not engage his responsibility, but instead chooses to hide behind regulations. This is a resolution stripped of all meanings, it is “doing something,” but not achieving one’s managerial goals. So instead, our approach is to hold managers accountable for their own decisions, regardless of the fact that it can sometimes put them in an uncomfortable situation for a little while.

Such management mechanisms are actually very hard to implement in a short period of time, especially in a very large organization. That’s why we look for enterprise tools to assist in it. Not only do those tools ensure consistency for management mechanisms, they also help us identify exceptions and re-align the processes efficiently. Moreover, they provide us with valuable data.

For example, we use open data to make managers more aware of their management responsibilities, and help them become more responsible. Because of the transparency of the data, the manager’s behavior is easily comprehensible, whether it is the rapid promotion of a team, or whether a timely termination of an underperforming employee was done, or whether performance reviews are reasonable. By looking at the data, we can continuously improve and iterate the manager’s processes, and help them become more efficient. We are also able to look in the accumulated data for trends, in order to gather more information points that will help managers in their judgement.

Management Practices at ByteDance

How does this philosophy translate into practice? Let me start with a few common management actions.

The first one I want to talk about is the sequence level. We have sequences and levels. A sequence refers to a type of role, for example, R&D or a product role. Each sequence has its own level. But we do not have a written level standard, nor a special promotion committee. We just roughly align the reference conditions and procedures for promotion, and above a certain level, we will do company-level cross-evaluation.

We focus more on people and their abilities than models. We will use some sample comparisons to determine levels and align standards through discussions. When we look at what level an employee should be, we often compare employees, trying to find similarities, and trying to find out why they should be on the same level. At the same time, our team allows for “cross-level promotions” and “rapid promotions.” In this process, there is no hard specifications card nor standard, resulting in an overall larger flexibility.

In addition to this, when it comes to compensation questions, we, first off, determine a salary range based on the market for a specific sequence level. But this range is mainly for references purpose and can be increased or adjusted. Only when compensations need to be adjusted extraordinarily, do we require a list of reasons of why it should be the case. At the same time, we also track premium management situations, scrutinize the follow-up performance of premium employees, which in return allow us to inspect the relevance of the decision made.

Our salary adjustment and year-end bonuses have corresponding models. For a specific employee and in a specific case, the system will recommend a range of salary and a year-end bonus according to a certain rationale. Once again, these ranges are more of a reference and can be topped. We check reports to better understand the overall situation, compare it horizontally and vertically with relevant accounts, and look for more details when the premium was awarded.

Our performance appraisal is divided into two categories. The first one is the employee performance appraisal, and the second one is a team-oriented performance appraisal. During this process, there is no mandatory distribution, and there is no termination, but we do have an appraisal meeting. Using a certain set of rules to check whether the performance is reasonable, we try to reach a consensus. We will look at the performance distribution and compare it with our own history as well as with other organizations. We will also focus on cases that need more attention, such as employees delivering consistently high or low performance, as well as the ones with important performance fluctuations, and the premium mentioned above. We will also look at the performance distribution of different people in the same sequence at the same level. With this data, we can actually visually see the distribution of performance; this is to avoid errors and to try ensure that no one is wronged. Through evaluation and consensus, the overall performance appraisal tends to be quite reasonable and fair.

Regarding organizational performance, we have established some consensus, and we do not determine performance based only on the absolute output. If the business data is good, is the performance of an organization necessarily good as well? Not necessarily. We look at performance by whether or not we transcend inertia. Even solid business numbers may be hiding some problems, and may not accurately depict yearly performance because of a strong foundation model. We pay special attention to transforming inertia into energy, a driving force. The performance review of the entire organization is basically based on this consensus. As for how the entire process is carried out, it is mainly through discussion and common views, such as, “what might be reasonable regarding organizational performance?” “why do we say that additional potential was created,” or “are those numbers the results of passivity?” Through consultations and debates on these particular questions, we, then, draw some conclusions.

Then, how do we apply the results of organizational performance? We determine if the organization’s performance is good or average. What impact will it have? In fact, we do not currently have clear regulations in place concerning the application of organizational performance results, only directional guidance. If it is good, then an emphasis can be put on bonuses. As for how those are attributed, whether it is to send a strong incentive to key personnel, or to increase high performances employee’s motivation, this is determined by managers.

The system can look at the final bonus situation of employees in different organizations. If there are some anomalies, we will discuss those. As long as there is a reasonable explanation, it is all right. As I have said before, overly focusing on numbers or absolute output may mislead and shift everyone’s focus to getting the best out of one’s team, while possibly ignoring the bigger picture or overarching company goal. This will make it difficult to be true to the facts and reality, and may make it hard to discuss the business potential objectively.

Resisting Gravity and Maintaining Flexibility

Of course, our current management method is not without its challenges.

A little while ago, I was chatting with an R&D colleague that mentioned that he had enrolled on his team a very good graduate, with very good performance and who was showing rapid growth. But after working with us for two years, he was poached away by another company for almost double the salary. He felt it was a pity. I asked him, “You had such praises for this employee and you were so confident in his value, why didn’t you adjust his incentive scheme; you just let him go?” He told me that the incentive range recommended by the system was limited, and even by giving him the maximum every time, he was not able to keep up, because it just was not increasing fast enough.

This surprised me, because this manager had been with us for several years and was very familiar with the company, and yet he did not know that our policies were flexible. This also made me realize that in a large-scale organization, it is very difficult to achieve unity of knowledge and practices in management. Although we have ideas and tools, we still encounter various problems in specific cases. As far as I understand it, this is a challenge between an ideal and reality. How to improve the reality of a concept is a major problem.

Another challenge when fewer rules are implemented (especially clear ones), is that everyone has to consider a lot of variables when making decisions, and those decisions are made on a case-by-case basis. Furthermore, with everyone being responsible for his or her decisions, it is often necessary to align them on a larger scale through discussions and meetings. This not only does put greater pressure on everyone’s process, but it may also lower the general efficiency. Managers are usually very much inclined to clarify company policies. This force, like gravity, tends to push any organization towards the establishment of even more rules. When it is the case, we must resist the pressure of the said force, or it will keep companies in a permanent rigid state.

On the other hand, the management problems we encounter in the real world are very complex. We cannot control, characterize or deal with the complexity of those situations through a set of rules. We need to be flexible so that managers can make sound decisions. The extent of certainty is very hard to grasp, especially when an organization is growing rapidly. It is necessary to ensure the effectiveness of management, and to ensure that the company is not paralyzed due to inefficient management. I think this is also a challenge we have been facing.

Finally, to sum up, the problem we face is actually how to efficiently scale and manage a large, pluralistic, and fast-growing global organization.

Our current approach is to increase consensus, and through culture, to reduce to a minimum of policies and rules, to achieve management efficiency through basic management mechanisms, to ensure management effectiveness by letting managers assume their responsibilities, and to ensure the implementation of management actions through a system, and making good use of data to better assist the overall decision-making process. During this process, we have to resist gravity and we must stay flexible all the time. There are still many challenges that lie ahead, and we are still exploring solutions to solve them.

Thank you.