- The title I started this page was “as centralized as possible, as decentralized as necessary”. However, I know understand that at the bottom of this there are two countervailing forces at play:
-
Centralized systems are more efficient
- Improving efficiently is a matter of deciding to do something differently than before. Decisions are always made by individuals.
- Coming up with decisions always means creating explanations. This can only be done by people. I still sense that there is a deep connection between decisions and explanations, see Every Explanation is a set of decisions.
- The highest level at which a decision can be made without losing details we should care about for reaching our goal is the most powerful one.
- At first I thought this statement is in conflict with The quality of an explanation can not be deduced by the level it is on. But in this statement “without losing details we should care about” points to “just the right level” again. So no contradiction here.
- A strategy is a design and direction imposed by leadership on an organization. Strategy began when people realized that telling warriors to “go out and fight the invaders” didn’t work. Leaders had to impose a structure, a design, on how the group would fight. - The Crux
- This requires executive power. Without executive power, heading an organization is similar to managing an apartment complex.
- You also don’t want to have people make decisions in cases they are not on the right position to do so:
-
A common example of forcing unnecessary decisions is asking your team to vote. If you ask your team something like “should we prioritize making tests faster” and then do what the majority says, everyone who answered against your path will disagree with what you did. In this case you forced them to make up their mind, and then decided against it. Alternatively, you could ask things like:
- Based on what you have information about, what would you say are the top 5 priorities? What information would you need to prioritize more effectively?
- What are the pros and cons of prioritizing a test speed acceleration project now?
-
Source: https://staysaasy.com/product/2023/07/12/changing-minds.html
-
-
Gathering data about reality requires input from many places
- Centralized data gathering is more error prone, because our sample is more likely to miss something of importance. In the case of figuring out what people value, we need to design systems that give people the freedom to act. Living in a “system that cares about what the aggregate of people care about” is something we by definition care about. Prices are the outcome of a decentralised discovery process.
- In that sense open protocols are systems that gather data of what users value in a way that closed protocols can’t. Thus We need open protocols for our infinite games.
- However, gathering data is itself expensive (so you don’t want to collect too much) and transporting it to the point of a decision maker is expensive, too. Add on that humans have only 24h per day and thus “cognitive overload” is a serious constraint on how decentralized you can make decisions. (This gives rise to platform teams as defined in Team Topologies).
- Size in an organization makes it more difficult to move information from the source (where data is gathered) to the locations where it can be best used (where decisions can be made to act on the information).
- Gathering data always means relying on a compression.
- Stories and anecdotes compress qualitative data and in my opinion good ones: Stick to qualitative insights as long as feasible. Only then introduce quantitative methods.
- Prices in a marketplace compress the information about what people value.