Likewise, I wouldn't give it much credit if MS (or Google, or whoever) does well. With a few notable counterexamples, most very large companies end up adopting processes to manage the size of the company, but how large is the impact of the particular process chosen? Is it really controlling success or failure?
Bad management/vision can happen in any system.
"Vision" is one thing but a lot of people have vision (or think they do). Very few people have that rare combination of interpersonal empathy and competency around complexity that enable them to reliably make good decisions as executives of large companies. (Worse yet, over 90 percent of the people given those positions get them as favors and utterly lack both skill sets.)
Um, no. The back-story you're missing is that, up until about 1980, it wasn't uncommon for people to stay at one company for 20+ years, even after they'd essentially ruined their reputations and were regarded as failures, because it was so disastrous and rare for people to actually get fired. So for companies to have 10 or 20% that was obvious deadwood was not rare. By "deadwood" I don't mean people who tried hard but took a long time to onboard or who worked on unlucky projects and could be rehabilitated, but I mean people who showed up at 11, left at 4, and might do no real work other than showing up for a couple meetings per week and making it obvious through inane comments that they didn't understand what was going on.
The first year of rank-and-yank, few objected to most of the firings because the people let go were the obvious nonparticipants, and it was better to get rid of them than lose the whole company. By the second year, it was more tense. By the third and fourth years, it was an outright morale-killer. It turns out that the percentage of people in a company who deserve to be fired isn't a constant.
In his book, Jack Welch says stack ranking put managers in a very difficult position around 3-4 years after it was implemented, simply because now they had choose worst among the very best. Of course not all of it is good, because even if you plan to hire and replace them by somebody else. What is the guarantee that they will be good?
But he later also says, the functions of the HR in GE were very different than in other companies. They were treated on par with the office of the CFO.
Since the HR itself is stack ranked, hiring wrong people means they will be getting ranked low and fired sooner or later. So they will be forced to hire the very best.
People work towards the goals you set them to achieve.
Why is this so surprising?
I often wonder if Microsoft's inability to get great products to market in a timely fashion is due to stack ranking. They clearly have great people who work there, who can make great products. Even the Zune was a great product. They arrive on the market about 3 years too late, and then require iteration on top of it. Is this because people there optimize their careers for "their number", rather than what's truly best for the company? Only time will tell.