Tuesday, August 04, 2009


Reed Hastings and I had a conversation in April to catch up. I was reminded of it again when his presentation on corporate culture came to circulate around various social networks. One of the questions Reed asked me at the time we had the discussion was whether I had any thoughts about management in the valley, having worked at more startups than he had (and also being in the less enviable position of not being the CEO at any of the startups).

One thing that conversation did was to get me thinking about the academic model of management that Google espouses. When I say management, I'm really talking about promotions, because that's how company culture reinforces its values (slide 7 in Reed's presentation). In 2004, I was an enthusiastic supporter of the academic peer-driven model. It felt much better than having a manager evaluate an individual contributor --- that model, I felt, always risked the problem of having a manager that was susceptible to brown-nosing. By the end of last year, however, I was seeing weaknesses in the academic model, many of which are articulated better than I ever could in The Trouble with Physics. In particular, what I saw was that certain types of problem-solvers and approaches were systematically under-valued --- and in many ways it's better to be someone who puts out fires than someone who prevents them from happening in the first place (note that it's not just the academic model that has this problem --- the top down model also does this --- in many large organizations, it's far better to quietly prepare a fire-fighting scheme if you see a fire happening than to try to prevent fires). The problem was that I couldn't see a better model, despite all the weaknesses of the peer-driven model. Like Democracy, I thought, the peer-driven model was the worst one in the world, except for all the others.

Reed told me he was a big fan of the traditional top-down model well done. The problem was that at startups (or even most big companies), I had yet to see the traditional top-down model well done. I frequently saw yes-men type middle management who couldn't say no to senior management, and all too frequently talent ignored in favor of hiring yet another manager from outside (something that startups do all too frequently). So I started asking around. I was struck by an insider's explanation of how Silos were broken down at Microsoft: the top 50 or so managers were called "partners", and their compensation was not at all related to their areas of influence, control, expertise, or title. Their compensation was tied completely to how the company performed, which basically made it so that all partners would help each other out if it was important to the company. That's a fascinating approach to solving the Silo problem, and to me, anyway, it provides an alternative to the academic peer-driven model that I saw as being imperfect. The problem is that the top down model still depends very much on having good managers. Even having had the luck to have exceptionally good managers at Google and elsewhere, I still run into enough poor managers at otherwise high quality places to believe that management hiring is anything but a crap-shoot at best --- no amount of interviewing will tell you that this guy who's a super-star on his resume is actually going to lead all your good people to leave over the next 5 years because of the way he plays favorites amongst his reports. This is why I believe that startups should grow managers from inside if at all possible. As Andy Grove once wrote: "People often complain that when you turn a great engineer into a manager, you get a mediocre manager and lose a great engineer. But think about the alternative? What message do you send if you pass over the smartest folks on your team in favor of someone from outside?"

In any case, I see the Netflix culture document as Reed's expression of his ideal approach to management --- it's another data-point in how to approach the fundamental problem of how a company organizes itself. There're also many ways of doing it wrong --- the worst of which is to pick a model that doesn't fit your management style and then coming across as hypocritical to all employees.

Ultimately, however, one must bear in mind that these discussions are moot if the business is unsuccessful --- that Google, Netflix, and Microsoft have to contend with these organizational headaches is a good sign --- failed businesses never have to worry about scale.
Post a Comment