Just as in personal life, where you should view yourself top-down as a machine, for your work life, you should view your organization top-down as a machine as well.
Like a foreman on a construction site, you should understand your tools and the components of the machine:
- Understand how people’s values, abilities, and skills differ, and how they should be matched with goals.
- Understand the processes and problems around you to make well-informed decisions.
Create great decision-making machines by thinking through the criteria you use to make decisions, before and while you make them.
A great manager is essentially an organizational engineer in charge of building the machine.
- They work hard to maintain and improve their machines.
- They study how well the machine is doing with metrics. They compare the current outcomes with the ideal outcomes, and find ways to close the gap.
- They worry about what can go wrong with the machine. They constantly survey the landscape to find suspicious signs of problems.
- They study what goes wrong in the machine—either it’s a problem with the machine’s design or with the people.
- They can anticipate the results from different inputs of people. They also expect that people will be imperfect, so they design a machine that produces good results even with mistakes.
- They make decisions with the primary goal of improving the machine, not to spare people’s feelings.
A master manager basically doesn’t have to do anything. They can sit back and watch the magic happen. Having to micromanage is a bad sign of improper machine design or the wrong people in the roles.
- But beware of being so distant from the details that you’re essentially not managing at all.
Every problem should be viewed as an opportunity to train and testing your machine.
- Discuss a problem at two levels: the machine level (why that outcome was produced) and the situation-at-hand level (what to do about it). Doing just the second is micromanaging instead of designing the machine.
- If you merely replace a broken part, it’ll break again. Keep looking at higher levels to see if there’s a need to redesign the machine.
Organization Structure
What’s the best structure for an organization, in terms of departments and team structure? Dalio gives these pointers:
Departmental Construction
- Imagine the ideal organization, and only then choose the right people for it. Don’t take the people you have and then fit an organization to them.
- Build the organization’s departments around goals, rather than tasks.
- Marketing and Client Services should be different departments, even though their tasks overlap. If they merged, their goals would conflict (one is to make sales and another is to take care of existing clients).
- Make departments self-sufficient so they don’t have to requisition resources through bureaucracy.
- Minimize the number of layers in the pyramid.
- Add governance and checks and balances.
- Governance is oversight that removes people and processes if they are failing.
- Even benevolent leaders tend to get more like dictators over time, since they have limited time to make difficult choices. They lose patience with arguments and idea meritocracy, and they issue commands instead.
- The company should have a board that determines if the people running the company are capable. The board can select CEOs but does not micromanage the firm.
- The board’s reporting lines must be independent of the CEO’s reporting lines, to avoid conflict of interest.
- The board must members have the courage to hold people accountable.
- Beware of fiefdoms set up by managers, where personal loyalty to the manager can conflict with loyalty to the organization.
- Dalio believes a single CEO is not as good as a great group of leaders.
- Thus Bridgewater has co-CEOs and co-CIOs.
Team Building
- Build from the top down by hiring managers first, who can help build the machine.
- Managers should also be good at what they’re managing, so they can intervene in case their reports aren’t doing well.
- Managers should have high standards so they can recognize problems.
- Different people are good at different parts of the 5-Step Process. Build a team of people who complement each other.
- Some people are visionaries who set the big goals, some are taste testers who perceive the problems, some are detectives who investigate problems, some are designers who create the plan, some are taskmasters who make sure the plan gets implemented.
- Some can do multiple roles well, but no one can do them all well.
- Make each person’s responsibilities clear. People should not jump to responsibilities they don’t own. There shouldn’t be confusion about who’s doing what.
- Have straight single reporting lines that don’t cross.
- Reporting to two people, especially when across different departments, causes confusion, except when reporting to co-heads of the same department.
- Forbid grabbing people from other departments unless first approved by the manager.
- Limit the ratio of manager to direct reports from 1:5 to 1:10.
- Allow escalation when people can’t handle their responsibilities. Don’t treat this as a failure, but rather a responsibility. (Shortform note: This is related to the idea of admitting to your mistakes rather than hiding them.)
- Prevent key man risk by training the next generation of leaders.
- Build a perpetual motion machine that can work well without you. Conversely, no one can be so important that they are irreplaceable. You should always be trying to hire someone who is better than you at your job.
- Every key person should have at least one person who can replace her, just as you have backup parts for machines.
- Expose the next generation of leaders to the thinking and decision making of current leaders. Have the successors do aspects of their managers’ jobs for a while to get vetted.
- Beware of people who don’t cultivate their successors, since they want to hold onto power and want their subordinates to stay dependent.
The 5-Step Process, Applied to Organizations
Just like individuals, organizations employ the 5-Step Process to evolve and continuously improve. Since an organization deals with far more people, it can be even more complicated to step through the entire 5 steps.
You’ve seen many of the ideas here in previous chapters; as you’ll see, they fit neatly into the 5-Step Process.
Set Clear Goals
Make a clear “contract” on what the expectation of good work is. This means a combination of goals, tasks, and responsibilities. Without this, you can’t hold people accountable for it not being fulfilled.
Be clear on how your lower-level goals achieve your higher-level goals and align with your values.
Review goals, tasks, and responsibilities at least once a quarter.
Choose great metrics. Imagine the most important questions you need answered to know how things are going. Metrics are indications of this. You should almost be able to manage via metrics alone.
- Any single metric can mislead. Use multiple metrics, and get enough evidence to see patterns. Look into the underlying data that produce the metrics to make sure they’re accurate.
When you set rules, explain the principles behind them. This way, people don’t just pay lip service to the rules—their code of ethics will compel them to abide by the rules.
Identify Problems and Don’t Tolerate Them
It must be OK to make mistakes, but unacceptable not to identify them or learn from them.
- People need to internalize the Life Principles around embracing reality. People need to embrace their pain as a growth opportunity, and reflect on what they have to improve.
- Celebrate finding mistakes since they’re opportunities to improve the machine..
- Don’t fire someone for making an honest mistake—this incentivizes hiding mistakes.
- Do fire someone for hiding a mistake.
- People must escalate responsibilities they cannot handle.
Have as many people looking for problems as possible.
- Specifically assign people the job of finding problems. These people should report problems to people other than their direct manager, so they can report freely without fear of being punished.
- Laws are weak unless you have auditors to enforce them.
- Welcome other people to probe you for problems.
Probe deep and hard for problems.
- Managers should “taste the soup” like chefs do before dishes go out. They should inspect the outcomes themselves, or have someone else in the machine responsible.
- Have regular standing meetings to avoid interactions slipping through the cracks. Standardize the agendas with the same questions.
- Dalio likes daily updates, asking his direct reports to write a 10-15 minute update on what they did that day, any issues they encountered, and reflections.
- You should personally probe to the level that is two levels levels below you. This is the only way you’ll tell how well the person reporting to you functions as a manager.
- Have them feel free to escalate problems to you, instead of requiring them to stop at their direct manager.
- Make the probing transparent to increase the probability of being right (to let others evaluate) and reinforce faith in due process.
- Pull all suspicious threads. Small problems can be symptomatic of serious underlying problems.
- Imagine your small problems are pieces of trash you’re stepping over to get across the room. They may not be critical issues, but cleaning up reinforces a culture of excellence.
Be very specific when describing problems.
- Don’t generalize. Don’t use “we” and “they.” Name specific people and specific instances. Someone made a bad decision.
Use the 80/20 rule to find the 20% of causes that lead to 80% of the problems.
No problem is too difficult to fix. Some people avoid problems because they seem too difficult. This will only worsen the situation and lead to more work later.
- In this step, don’t worry about fixing the problems. Just identify them.
Beware of these barriers to finding issues:
- Boiling frog syndrome: with exposure, people get used to problems that would shock them if they saw them with fresh eyes.
- Groupthink: don’t assume that just because other people aren’t screaming about a problem means that there’s not a problem. Other people are probably thinking the same thing you are, but no one wants to speak up.
Keep an error log to track mistakes and solutions. Then observe patterns of mistakes to see if they are signals of actual weaknesses.
Diagnose Problems to Find Root Causes
When solving problems, a common mistake is to deal with problems as one-time events, rather than diagnosing problems with the machine. If you don’t fix the root cause, the problem will keep occurring again and again.
Get to the level of what it is about the people involved that led to bad outcomes. Root causes are described in adjectives, not verbs. Someone could be “careless,” “not well trained,” or “has low bandwidth.” In contrast, “he forgot to write a reminder” is not a root cause.
Here’s a checklist of questions to ask to diagnose the problem:
- Was the outcome positive or negative?
- Who was the responsible party in charge of the outcome?
- Did the responsible party lack ability or skill, and/or is the design of the machine bad?
- How should the machine have worked?
- Stay at the level of the machine (who should be doing what), rather than getting bogged down in procedural details.
- Go back before going forward. Reflect on how the machine was working up until now. Tell the story of how we got here.
- If the machine didn’t work, what broke? This is the proximate cause.
- Why didn’t things work as planned?
- Ask Five Why’s to get at deeper causes.
- Pinpoint a specific key attribute (for example, Thomas does not execute well, or Thomas does not perceive problems quickly.)
- If this attribute is fixed next time, will the bad outcome still occur?
- In other words, if we fixed this part, would that prevent the problem?
- If the problem is a faulty design with the machine, ask who was responsible and whether they’re capable of designing the machine well.
- Does the root cause a have pattern? Does it show repeatedly?
- This is the difference between “Thomas was careless” to “Thomas is often careless.”
- If it is a pattern, is this due to bad training or lack of ability?
- How should the people/machines evolve as a result?
- Who should do what differently?
- Do responsibilities need to be clarified?
- Do machine designs need to be reworked?
- Do people’s fit need to be reevaluated?
- Were principles violated? Are new principles needed?
Remember that this type of questioning will upset people, and they may protect themselves in a number of ways, like fixating on details instead of lack of ability.
Remember that diagnoses should lead to improvements and outcomes. Otherwise it’s just a waste of time.
Stay calm and rational. Don’t overreact to a mistake or problem.
- Avoid Monday morning quarterbacking—don’t evaluate a decision with limited information.
- Understand the context the person was in. Was this decision a quality person would have made? Did they do the right things, but the outcome was out of their control?
Here are common root causes of problems with managers:
- They are too distant from their reports.
- They have problems detecting bad quality.
- They have gotten accustomed to how bad things are (the boiling frog syndrome).
- They can’t admit they’re unable to solve their own problems, and they don’t escalate.
- They fear being punished for admitting failure.
- They fail to delegate their responsibility. They focus more on doing the tasks than on operating the machine.
Example of Root Cause Analysis
Problem: The team deals with emergencies constantly and has to work overtime.
Why? Because we don’t have enough bandwidth to fix the issues that cause the emergencies.
Why? Because the manager did not anticipate the problems and request more headcount to fix it.
Why? Because the manager is bad at planning for bad cases. (Root cause)
Design Improvements to Your Machine
Take the time to come up with a good game plan. The time it takes to come up with a plan is far less than the time that people will spend executing the plan.
Visualize the plan working. Picture, like a movie script, who will do what when, and the result.
- Put yourself in the workflow to understand what you’re dealing with.
Visualize alternative machines and outcomes. Don’t settle on the first obvious solution.
- If something is hard or frustrating, triangulate with others.
Consider second- and third-order consequences of the change to the machine.
Consider that the problem might be happening at other departments, and they may need attention to.
Make changes to the Organization Structure if needed.
Continue articulating the principles behind your decisions in the situation, and check them with other people to improve them. These principles will build habits for solving similar situations in the future.
Do the Tasks Required to Completion
Execute the plan and track its progress. Report regularly on the planned and actual progress, and forecast the progress to come. Hold people accountable for their responsibilities.
Remember that fixing the problem is iterative. It will be imperfect for some time as it moves toward the good state.
Most of success comes from doing the mundane stuff, like finding problems and fixing them.
To push through failure, visualize the success and crave the feeling of achievement. Also visualize the tragedy of not pushing through.
Expect things to take 1.5x as long and cost 1.5x as much as you expect.
Use checklists to capture what must be done and to confirm that they are done.
People intellectually know what’s good for them to do, but it won’t happen without good habits. Develop good habits through tools and protocols. Automate decision making if possible.
When you reach your goals and solve problems, celebrate!