***You must cultivate a habit of never tolerating any kind of bad habit, regardless of its severity.
Only through intolerance, even disgust, towards these matters can you do them well; you need to make this mindset a habit and turn it into instinct.
***For example, when setting goals, just set the goals without thinking about how to achieve them or what to do if you make a mistake. When diagnosing problems, don’t think about how you will solve the problem, just diagnose. Confusing these two steps can lead to undesirable outcomes because it hinders you from discovering the real issues. This process is progressive: if you do each step carefully, you will obtain the necessary information to move on to the next step and do it well.
This is one of my problems; I always think of solutions when diagnosing problems, and then my whole thought process gets muddled. Identifying problems and thinking about solutions are two steps that are progressive and should not be confused.
***If you are too proud of the knowledge and skills you possess, you will learn less of the right things, the quality of your decisions will decrease, and you will find it difficult to fully realize your potential.
Humility is a virtue; remember, humility.
***Remember, you are looking for the best answer, not the best answer you can come up with.
Consulting others in areas where you are not proficient is a good choice; consulting others in areas where you are proficient can provide different perspectives. Humility is a virtue. We pursue the real optimal solution, regardless of where that solution comes from.
***When designing the personnel characteristics for Bridgewater baseball cards, I combined three aspects: adjectives we already use to describe people, such as "conceptual," "reliable," "creative," "determined"; behaviors of people, such as "holding others accountable" and "sticking to implementation"; and terminology used in personality tests, such as "extroverted" and "judging." After the cards were created, I designed a program: everyone scores each other, and the person who scores highest in a particular area (like "most creative") has a higher weight in scoring others in that area. Individuals with a proven track record of achievements in specific fields are considered more credible in those areas, thus receiving higher decision-making weight. After recording everyone's characteristics with baseball cards, those who have never worked with them can reasonably anticipate their behavior. If a person changes, their scores will also adjust; if they remain unchanged, I can be more confident in my expectations.
The characteristics described on the baseball cards are divided into three aspects: 1. adjectives describing people; 2. behaviors of people; 3. terminology used in personality tests. Then, through mutual scoring, we determine each person's decision-making weight in different areas. The content of the cards will change as people change. This is essentially a unified standard; the entire system uses the same set of metrics, minimizing communication costs, especially for newcomers, who can quickly understand the abilities and personalities of team members.
***Habits are essentially inertia, a strong tendency to continue doing what you have always done (or to continue not doing what you have always not done).
You need to think carefully about how to establish habits; you can refer to "The Power of Habit."
___We mainly use four types of tests: MBTI, Workplace Personality Inventory, Team Orientation Questionnaire, and Hierarchical Systems Theory.
We found that 25 to 50 characteristics can effectively describe a person.
This is something I had not tried before; psychological testing can be integrated into the recruitment process, attempting to describe a person using characteristics, which is the same principle as the baseball cards mentioned earlier.
***No matter how high your probability of making the right bet is, it is always valuable to increase your probability of making the right bet. I often see people making decisions when their probability of making the right bet is above 50%, but what they fail to see is how much more they could gain by further increasing that probability (you can almost always improve your probability of making the right bet by obtaining more information). The expected value gain from increasing the probability from 49% (where the chance of being wrong is slightly higher) to 51% (where the chance of being right is just slightly higher) is 17 times less than the expected value gain from increasing the probability from 51% to 85% (a 34 percentage point increase). Treat probability as a measure of your likelihood of making a mistake. Increasing your probability by 34 percentage points means that one-third of your bets will turn from losses into gains. Therefore, even when you are very confident, it is beneficial to stress-test your thinking.
Striving for excellence, enhancing your winning probability is always extremely valuable, especially regarding probability issues, by obtaining more information to improve your winning probability. If time is unlimited, this is an endless pursuit. However, it is essential to maximize your winning probability within the opportunity window; once you miss the time window, the winning probability becomes meaningless.
***Compare the value brought by more information with the cost of not making a decision to prioritize.
Order is important; you need to establish a sequence and not miss opportunities.
***I categorize current computer-aided decision-making technologies into three main types: expert systems, imitation, and data mining (this is my classification, not the one commonly used in the tech community).
At Bridgewater, we use expert systems, where designers articulate decision criteria based on their rational understanding of a series of causal relationships and then observe what different situations arise under different conditions. However, computers can also discover patterns and apply them to computer decision-making without needing to understand the logic behind those patterns. I refer to this decision-making technology as "imitation." This approach may be effective when the same situation occurs reliably and consistently, such as in a game with extremely strict rules. But in reality, things are constantly changing, so such systems can easily become disconnected from reality. Data mining is the fastest-growing area of machine learning in recent years, referring to powerful computers digesting vast amounts of data to find patterns. Although this approach is common, it is risky in future situations that differ from the past. Building investment systems based on machine learning without a profound understanding of reality is dangerous because when people widely believe and apply certain decision rules, prices will be affected. In other words, once a profound insight becomes widely known, its value gradually diminishes. Without a deep understanding, you cannot determine whether what happened in the past was genuinely valuable; even if it was valuable, you cannot ascertain whether that value has disappeared or even turned negative. A common situation is that some decision rules become so popular that they significantly impact prices, making contrary actions more reasonable.
Remember, computers lack common sense. For example, a computer can easily misinterpret facts; seeing a person wake up in the morning and eat breakfast, it might conclude that waking up causes hunger. Instead of making many bets I am unsure about, I would rather make fewer bets that I am sure about (preferably unrelated bets). At the same time, being unable to explain the logic behind any decision is something I cannot tolerate. Many people blindly trust machine learning because they find it much easier than forming a deep understanding. For me, deep understanding is essential, especially regarding what I do. I do not mean that imitation systems or data mining systems are useless. In fact, I believe that under conditions where the future's development scope and patterns are similar to the past, these systems can be very helpful for decision-making. With sufficiently powerful computing capabilities, all possible variables can be considered. For example, by analyzing data from chess masters playing under different conditions or analyzing the surgical processes of outstanding surgeons during various operations, valuable chess programs or surgical protocols can be created. As early as 1997, the computer program "Deep Blue" defeated the world's top chess player Garry Kasparov in this manner. However, when the future differs from the past, and we cannot identify all causal relationships due to a lack of understanding, this approach becomes unfeasible. With an understanding of these causal relationships, I did not make the same mistakes as others, the most obvious example being the 2008 financial crisis, when almost everyone believed the future would be like the past. By focusing all attention on logical causal relationships, we can see the truth of things.
The thoughts of big names on AI are somewhat similar to my views; machine learning lacks understandable logic, which differs from how we think. The limitation lies in the numerous and rapidly changing external rules. If the rules are strictly controllable, machine learning, or what the author describes as "imitation," is highly desirable, such as in Go; expert systems distill and summarize human internal logic, allowing machines to deduce decisions based on human logic. I share the author's viewpoint that deep understanding is essential; I believe that through logic, we can deduce the future; machines have issues understanding causal relationships, and they may confuse causality with correlation, which can lead to significant problems.
***Every principle has exceptions; no principle can replace common sense judgment.
Here, I fully understand the meaning of "common sense." Common sense is essentially a basic understanding of the principles governing the operation of things; it refers to logic, which has an essential and unchanging foundational logic in this world. Principles are superficial, while common sense is foundational.
Extreme truth-seeking and extreme transparency at different levels may be more suitable for the current situation. If there are no levels, then the quality of all employees must meet very high standards, which is difficult to achieve in my industry and living environment.
***Ensure everyone understands the difference between fairness and generosity.
This is important; benefits are generous and do not need to be fair; one cannot complain about someone not being generous simply because they did not enjoy benefits; basic guarantees should be fair. The idea of "not worrying about scarcity but worrying about inequality" does not apply here.
***Do not get caught up in "complaining" or "praising," but focus on "accurate" or "inaccurate."
Complaining and praising are both emotional responses towards people; here, the focus should be on describing the situation; "accurate" or "inaccurate" is a choice that better describes the facts.
***First assume that either you did not communicate well or did not hear clearly, rather than blaming the other party.
This reminds me of my communication with the property management; when I heard the other party's statement, I felt a surge of anger, but it might just be a misunderstanding. Therefore, assuming you are wrong when encountering communication issues is a good way to avoid misunderstandings.
***Meetings should include the following elements: agenda, conclusions, action items, responsible personnel, and deadlines.
Do not let meetings be held in vain; these need to be recorded by someone specifically.
***Credibility weighting can serve as a tool, but it cannot replace the decision-making of the responsible person. The decision-making mechanism of credibility weighting can supplement and challenge the decision of the responsible person, but it is not meant to overturn the decision.
The decision-maker can overturn the conclusions reached by the credibility-weighted vote but must take responsibility for it. When decision-makers choose to replace the consensus reached by others with higher credibility with their own judgment, they are making a bold declaration and will accept the verification of the final outcome.
Credibility weighting is like GPS; it cannot replace the responsibility of the decision-maker. The decision-maker must still be accountable for their decisions, even if they choose a decision with high credibility weighting. If it is wrong, the decision-maker is still responsible for the decision-making error; the logic here is that whoever decides is responsible. If there is no decision-making power, there is no need for responsibility, and if there is no responsibility, there is no decision-making power.
***People tend to choose those who are similar to themselves, so when arranging interviewers, ensure they can identify the type of person you want to hire. If you want to hire visionary people, find someone visionary to conduct the interview; if you seek well-rounded individuals, form an interview team that possesses all those qualities. Do not select people you do not trust as interviewers (in other words, ensure they all have credibility).
I deeply resonate with this; if the entire company were filled with people like me, it would be disastrous. During recruitment, psychological personality tests can help mitigate this issue to some extent. Keep good records of employee interviews, and after employees join, carefully observe whether their performance aligns with their interview.
***If a person is not performing well at work, consider whether it is due to insufficient learning or a lack of ability.
If it is due to insufficient learning, arrange training; if it is due to a lack of ability, they must be eliminated.
***You should delegate specific tasks to employees. If you confine yourself to details, you either have management or training issues, or your subordinates lack capability. A hallmark of an excellent manager is that they do not need to do anything personally. Managers should view getting bogged down in minutiae as a bad signal.
If it is an issue of subordinate capability rather than skill, then a change of personnel must be considered. Managers should treat getting caught up in specific details as a dangerous signal.
***Confidence in employees should come from understanding, not random guessing.
Do not assume that employees' answers are all correct.
All issues must have evidence; they cannot be based on conjecture; this is not an empty statement. Employees must be able to explain the logic behind their answers.
***Do not maintain too much distance. You should be very familiar with your employees, regularly communicate feedback, and engage in high-quality discussions. Although you do not want to be disturbed by idle gossip, you must be able to quickly obtain information from the right people. Your work design must include time for these activities; otherwise, you risk mismanagement. The tools I developed allow me to have a window into what people are doing, what their preferences are, and I follow up promptly when issues arise.
Use "daily updates" to understand team members' behaviors and thoughts.
Understand what they report directly to your subordinates and observe their behavior in front of your subordinates; otherwise, you cannot know how your subordinates manage their subordinates.
Allow your subordinates' subordinates to report to you at any time. This is a very effective way to enhance accountability.
I have doubts about this part; should there be restrictions on hierarchy? I actually question this idea because I do not know if doing so will make my subordinates feel that I am interfering in their work, or if it will hinder their work. If their subordinates feel they can report directly to me, bypassing their superiors, the authority of the superior will be diminished, making their management more difficult; it is as if they are being sidelined, which may not be a good thing for the overall management of the company. I am not very clear on this and need to think more.
***Make the accountability process transparent, rather than holding people accountable privately.
This process should only involve mid-level and senior managers; should it not be open to those below? Or should it still be open to those below? I think having hierarchical openness is more reasonable; leaders should not be held accountable in front of their subordinates, as this would undermine their management authority.
***A basic fact is that if you are not affected by the results of your actions, you will not be responsible for those actions.
Ensure that the results of actions will impact the person implementing them. This is the foundation for establishing rules and decision-making responsibility.
***Acknowledge and address key personnel risks. For every key employee, at least one alternative must be arranged. Identify successors and train them to fulfill their roles.
This can eliminate many risks later on, but it is challenging to implement. If a person knows they are training their successor, the idea of teaching a disciple to starve the master will inevitably arise. I believe that mature companies should abandon the concept of key employees and instead ensure that most knowledge is retained within the company's systems, processes, and decision chains.
***Do not treat everyone equally; treat them reasonably and with distinction.
However, treat employees according to the same set of rules.
These two statements are not contradictory; the same set of rules applies, but simply put, people should be treated reasonably and with distinction based on their abilities or other factors. For example, a long-serving employee who has fought for the company for five years should be treated differently from a new intern, even if they are in the same position. Doing this well will make people feel a sense of belonging to the company; doing it poorly will leave them feeling disheartened.
***Do not easily be forced to concede.
God does not bleed.
***When a person is in distress, they must be privately comforted.
The distress here should refer to personal distress, and the emphasis should be on "privately."
***People often unconsciously tend to do what they like rather than what they need to do. If they neglect the priority of tasks, you need to redirect them. This is also one of the reasons why it is important to frequently listen to employees report their work progress.
It is not only important to know what has been done but also what should be prioritized.
***Beware of lacking focus and futile "theoretically should."
This is actually something I do the most as a consultant because I am not responsible for these matters. Therefore, many such statements arise. This is harmful, and I need to try to change this habit.
***When unable to fulfill responsibilities adequately, submit the issue to your superior for resolution, and encourage your subordinates to do the same proactively.
I believe this practice, on many levels, helps everyone understand that achieving the goal is the top priority; who completes it or other circumstances are secondary. The main thing is to achieve the goal. Therefore, even if the goal is not completed by them, it is still acceptable. But achieving the goal is very important.
***People gradually develop a strong tendency to accept unacceptable things, and these unacceptable things can be shocking to the discerning eye.
Therefore, it is essential to regularly review the entire system.
***Do not use vague terms like "we" or "they" to obscure personal responsibility.
Be specific; this is not impolite or shirking responsibility; it is being factual and a manifestation of extreme transparency.
***To make a proper diagnosis, first ask the following questions: 1. Is the result good or bad? 2. Who is responsible for the result? 3. If the result is bad, is it due to the lack of ability of the responsible person or a problem with the machine design?
This order must not be mistaken.
***Plans do not necessarily require everyone's consent, but the responsible person and other key employees must reach a consensus on the plan.
This is akin to informed consent; it is not imposed from the top down but achieved from the bottom up, thus avoiding situations of shirking responsibility.
***Build the organization from the top down. An organization is the exact opposite of a building: the foundation of an organization is at the top.
Top-level understanding; the foundation of the organization is at the top. There must be leadership before there are employees, applicable to any company and its departments.
***To ensure that key tasks are completed correctly, it is better to "do it twice" than to "double-check." The error rate of double-checking is higher than that of doing it twice. Doing it twice means having two different people independently complete the same task, resulting in two unrelated outcomes. This not only ensures better answers but also allows you to understand the differences in their behaviors and capabilities.
There is a fundamental difference between doing it twice and double-checking; the review mechanism of double-checking is not reliable. For example, the approval processes in DingTalk or Feishu are not very effective, but doing it twice wastes resources. Therefore, this approach should only be used for core or very important events that will have a significant impact.
***Continuously think about how to create a leverage effect that yields high returns from small investments. The leverage within an organization is not the same as leverage in the market; find ways to achieve greater returns with fewer resources. At Bridgewater, I generally apply a 50:1 leverage in my work, meaning that whenever I spend one hour discussing work with my subordinates, they need to spend about 50 hours advancing related projects. During our discussions, we go through the overall planning and deliverables, then they start advancing, and we review the progress, after which they continue to advance based on my feedback—this is how we work in cycles. My subordinates work similarly with their subordinates, although their time ratio generally ranges from 10:1 to 20:1. I always hope to see employees work as hard as I do (ideally better than I do), so that I can maximize the output of my one hour.
This is something I had not thought of; it is essentially a positive energy version where a superior's words lead to subordinates running around. The superior's words are not just random statements; the subordinates running around is the right approach.
***The attribution of decision-making authority must be clear. Ensure everyone knows how much weight their vote carries so that in situations where disagreements persist and decisions must be made, they can proceed without hesitation.
Thorough discussion, selecting the best ideas, but clear decision-making authority is an important tool for advancing matters within limited time. The clarity of decision-making authority also allows everyone to recognize that even in the presence of disagreements, progress can be made without hesitation. Progress cannot be stalled due to disagreements.
***At Bridgewater, we have discussed that designing effective quantitative metrics requires four steps: (1) Understand what the company's goals are; (2) Understand the processes to achieve those goals (your machine, including people and design); (3) Identify the key parts of the process that are most suitable for quantification to understand how the machine operates to achieve the goals; (4) Research how to leverage key metrics to adjust processes and change outcomes.
First, set the goals, design the pathways to achieve those goals, and identify the key parts of the pathways that are most suitable for quantification. This is the method I see for designing effective quantitative metrics; researching how to leverage key metrics should be about maximizing the time efficiency of managers.