Because you are able to model the behaviour of the humans you are dealing with, you understand the system you are modifying, and you understand how to mould culture, none of these issues will come up.
These are just frameworks that sound intelligent but are not useful.
Conversation
"Wait, but Cedric, Goodwin's law is a thing! Other intelligent people talk about it!"
Yes, but the way you defeat Goodwin's law is not by talking about Goodwin's law. Bloom talks about Amazon in his tweet thread. Amazon actually deals with Goodwin's law quite well.
2
1
2
In Working Backwards, Colin Bryar and Bill Carr dedicated an entire chapter to how they think about metrics.
The number of times they mention Goodwin's Law: 0.
Amazon has a process they call DMAIC. The book tells the story of the step-by-step unfolding that led them to it.
2
7
Bryar and Carr are extremely believable, by the way. They were in the room when the 6-pager was designed, when the decentralised org in Amazon was built, and when Amazon's approach to metrics was still being built out.
DMAIC enables them to sidestep Goodwin's law.
1
1
2
Most people, with no org design background, would read Working Backwards as a 'manual of techniques to apply'.
Org designers read Working Backwards as an accounting of the step-by-step unfolding that LED Amazon to design those systems.
My summary:
1
7
Novice org designers would read stories of misaligned incentives and go "Ha! What a bad idea! Use these frameworks!"
Experienced org designers would read stories of misaligned incentives and ask: "How did they get there and what processes did they try after that incident?"
1
1
4
I spend a lot of time talking about believability. I said that you should not pay attention to ideas from people who:
- Have not had at least 3 successes in the domain, and
- Have a coherent explanation when probed.
This is not my idea, it's Ray Dalio's:
2
8
But I think this thread is a good example of why.
The instant I read Bloom's thread, I was like "ok, this seems ... off."
Every idea was correct.
Every idea sounded intelligent.
Every idea was also quite useless.
1
3
And the reason for that is because the WAY you use the ideas are not the way you might think they should be applied.
The shape of the expertise of org-design is a step-by-step unfolding. Not taking that into account is a novice mistake.
1
4
Why this is the case is a different thread for another day.
(The basic idea is that nobody can perfectly predict org response to incentives because orgs + org culture are somewhat complex and dynamic and adaptive. So you need to iterate to see how the system adapts).
1
7
Anyway: all of this is to say, be careful who you read. Make sure they are believable in the domain. Or you might be led astray by correct ideas that are just not useful because they've never been put to practice.
The end.
Replying to
Correction: I meant Goodhart's Law, not Goodwin's law.
I should've known better! And I say this as someone who summarised Mainheim and Garrabrant's "Categorizing Variants of Goodhart's Law" paper in the past!
Quote Tweet
"Wait, but Cedric, Goodwin's law is a thing! Other intelligent people talk about it!"
Yes, but the way you defeat Goodwin's law is not by talking about Goodwin's law. Bloom talks about Amazon in his tweet thread. Amazon actually deals with Goodwin's law quite well.
Show this thread
1
4
