My argument is that in many cases it's not local optimization at all. It's not optimal for anyone, their freedom is constrained by upper level management (the h

My argument is that in many cases it's not local optimization at all. It's not optimal for anyone, their freedom is constrained by upper level management (the hierarchy / the corporation / the hivemind) which means neither the managers or their reports have freedom. If the managers actually had more freedom, you'd get something closer to a bunch of small groups optimizing locally and producing (ideally, with the right incentive) collectively good results. Here's a really easy example from my career (not Google, though): A manager needs to deliver on schedule. He has a set number of people on his team. One of those people is going to quit because he has a baby on the way, and the company's salary is not good enough. If the manager had enough freedom, they'd have options available. The company is going to cut some of those off, but I'll enumerate some of the ones we potentially had at our fingertips (I was one of the team's managers in this situation): * Give him a raise so he stays * Promote someone qualified from the team, hire a junior employee to replace them * Hire a senior employee to replace him and train them up ASAP to fill the gap * Adjust the schedule to make up for the loss of his contribution * Adjust the scope of the project to same as the above In the end the studio head basically told us to go fuck ourselves, and the report ended up quitting. The project came in late, as I expected. This happened multiple times and in multiple cases it was direct interference from higher nodes in the chain - neither us in management or the labor below us were able to optimize at all. The product as a whole ended up shipping multiple years late due to lots of smaller-scale interference like that.