This is a more problem than networks vs hierarchies since both face this problem. All companies that succeed by definition start with a successful social structure, but as a companies increases the number of employees the social structure has trouble scaling. Clausewitz discusses a simple example of this with an army, as the army grows the number of links in the chain from a commander to a unit increases. He proposes a number of solutions (increase the unit size, increase the branching degree, split the army, etc) but in typical Clausewitzian fashion admits there is an inherent trade off between size and effectiveness.
Valve maybe choosing to stay small to avoid scaling beyond their effective structure. This is a counter-intuitive strategy in the world of bigger is better, but I'm interested in seeing how it plays out.
Valve shouldn't be a surprise. Despite all the ink spilled over management literature, even normal companies are typically full of dysfunction. There is no royal road to getting rid of defacto bosses.
> Valve maybe choosing to stay small to avoid scaling beyond their effective structure.
I wouldn't call 300 people a small number for the size of a company.
In fact, Jeri Ellsworth addresses this point (I'm quoting the transcript by develop-online [0]):
"Their structure probably works really well with about 20 people, but breaks down terribly when you get to a company of 300 people. Communication was a problem. That's where management… Well if I had anything to do differently, would be to make sure a layer of management could do communication correctly."
So it looks like Valve has already scaled beyond their effective structure.
>So it looks like Valve has already scaled beyond their effective structure.
Such scaling problems are not binary. They are probably less effective than they were at 20 people but they would be even less effective at 30,000. One way we might look at this is to assume (incorrectly) that a flat hierarchy is a complete graph which is the way most small informal groups are run.
A complete graph of 30 people has 435 edges.
A complete graph of 300 people has 44,850 edges.
A complete graph of 3000 people has 4,498,500 edges.
This is a more problem than networks vs hierarchies since both face this problem. All companies that succeed by definition start with a successful social structure, but as a companies increases the number of employees the social structure has trouble scaling. Clausewitz discusses a simple example of this with an army, as the army grows the number of links in the chain from a commander to a unit increases. He proposes a number of solutions (increase the unit size, increase the branching degree, split the army, etc) but in typical Clausewitzian fashion admits there is an inherent trade off between size and effectiveness.
Valve maybe choosing to stay small to avoid scaling beyond their effective structure. This is a counter-intuitive strategy in the world of bigger is better, but I'm interested in seeing how it plays out.