Satya Nadella even refers this in his book "Hit Refresh", and it seems like he has change the company structure to avoid this problem better than before.
I wonder what are the effects of Conway's Law on each of these structures.
"Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization's communication structure." — Melvin E. Conway
I have come across this chart multiple times, and I find it quite amusing. However, it got me thinking about what the ideal org-chart should look like. In my opinion, for larger companies, the Apple approach might be the best. This involves coming up with a clear and compelling high-level vision and repeating it consistently. Pheromones in a beehive essentially.
The other thought that I have had is management is kind of like inverse tSNE. At the extreme, one is mapping a single dimension (NPV of the company) to high dimensional space (people doing stuff). No wonder it is so hard.
I don't think the shape of the org chart actually matters all that much. It's the channels of communication, which way information flows, and how much autonomy the leaf nodes have that actually matter.
Case in point, Amazon vs Netflix. At Amazon, every decision requires approval of someone higher. Engineers can come up with ideas, but they have to write it up, get it approved, and then get budget assigned for that particular project from management. Even Principle Engineers have to get approvals for their own ideas and projects. And if they have ideas for new company goals, they have to send that up the chain through their management to the leaders (maybe skipping a level or two).
Netflix on the other hand has the same org structure as Amazon, but operates completely differently. Management provides context, not approvals. They say, "these are the goals we are trying to achieve". Then they leave it up to the engineers to implement it however they think is best. And if they need more budget, usually the way it works is that engineers spend the money first and if it's too much someone asks them to find ways to reduce it. And if the engineers have ideas for better company goals, they send that directly to upper management. At Netflix you never have to go through a chain of management to get things done.
>> I don't think the shape of the org chart actually matters all that much. It's the channels of communication, which way information flows, and how much autonomy the leaf nodes have that actually matter.
Perhaps that is not what is indicated by a traditional org chart (who reports to who), but the "real" org chart does have vertices (people) and edges of different kinds. The most important factor is the amount of time that you control. 1.0 means you completely control your own time, 10.0 means you completely control the time of 10 people. The real org graph is what I am referring to.
Interesting that Amazon is as bureaucratic as it is. Isn't it Bezos himself that came up with the "multiple paths to yes" concept. I suppose, some things are inevitable in an organization once someone has the power press the stop button one's income.
> Interesting that Amazon is as bureaucratic as it is. Isn't it Bezos himself that came up with the "multiple paths to yes" concept.
Are you referring to the "1 > 2, but 2 > 0" idea? That's compatible with the Amazon bureaucracy, which just requires finding a leader to sponsor the project even though it might end up redundant with another org's effort.
I don’t think so but maybe. I recall watching a video once where Bezos states (accurately I would say) that in most companies any idea requires multiple people to say “yes” while startups only need one (the vc firm). His thesis is, this is why companies fail to innovate.
It is pretty obvious in some ways. Once a company has reached a certain level, there will be many status quo forces that emerge because hey, things are “good enough” for some.
Ah, gotcha, basically the proliferation of veto points as companies grow into corporations. In its idealized state, Amazon avoided that by giving autonomy to the deeper nodes (the famous Two-Pizza Teams).
> However, it got me thinking about what the ideal org-chart should look like.
I'm not sure there's was ideal org-chart.
Especially when you take Conway's Law into account:
> Any organization that designs a system (defined broadly) will produce a design whose structure is a copy of the organization's communication structure.
I suspect that depending on what technical system you want to build, different designs are good. So you better shape your org like the design you want to achieve, because that's the only one you can ship.
I've always enjoyed this chart for the humor, but there's a kernel of truth that the structures that allow for more ownership (FB, Amazon, Apple even though it was a dictatorship) do seem to win out in the long run.
It could represent a bottom's up engineering culture, but the graphic would be missing junior engineers and leadership which would add more tree like shapes into it.
Structure definitely makes a difference: I've worked in tech companies with 2000 employees that were more nimble and higher functioning than many a 100 person startup. Even among very large companies, there's clear tradeoffs that lead to very different actual outcomes.
That said, the larger the company, the bigger the problem and the harder it is to change direction when your organization is messed up. And as a company grows, an organization that was optimal can become a disaster.
A 100k employee organization is always going to have some parts of itself that are going to feel glacially slow. But there are companies that big that still deliver great products, while others seem to only exist due to inertia, as they couldn't build anything new that people wanted at any speed. But as Microsoft has sown us, a few organization and leadership changes can turn a company that at some point looked undead into relevance. Companies like Google and Meta are large enough that, even if we made the leadership completely incompetent, the companies would probably be valuable enough to survive for decade, just on inertia
