Tales of an Ex–Microsoft Manager

Here&#x27;s a commentator at Mini-Microsoft explaining why it probably became more toxic over time, and it looks like these changes happened before the author arrived:<p><a href="http://minimsft.blogspot.com/2013/08/steve-ballmer-is-going-to-frickin.html?showComment=1377372530878#c2642513338204621608" rel="nofollow">http:&#x2F;&#x2F;minimsft.blogspot.com&#x2F;2013&#x2F;08&#x2F;steve-ballmer-is-going-...</a><p>Anonymous said...<p>My memory is really good - [ current head of HR ] is the one who did away with the &quot;life-time performance average&quot; which let people take huge risks and made Microsoft great.<p>THAT, was huge BillG wisdom. You could look at your score over time and do some really simple math to see what kinds of risks you could take. And then suddenly, a 15-year running average of 4.0 turned into &quot;your current manager doesn&#x27;t like you, and he&#x27;s required to fire one of his 10 no matter what, so you&#x27;re fired&quot;.<p>She also did away with the standing rule that you could take any written standing offer - NO MATTER WHAT (unless you were actually fired for cause) - which was Bill&#x27;s way of ensuring that great engineers couldn&#x27;t be tossed out by one horrible ass-kissing manager ... which is pretty much all of them today.<p>The day that it became possible for a single crappy manager to fire someone for saying &quot;I disagree&quot;, or &quot;That&#x27;s Wrong&quot;. Microsoft died.<p>And when&#x27;s the last time you saw that lovely deck of laminated cards that started with &quot;a passion for technology&quot; ... and a really logical employee development plan? yea, right, somewhere around 1995.<p>[ People who need to leave. ]

I remember one comment at a perf review from my boss when I worked at MS. There were two of us on the team of 35, who were working on a certain tenet. My boss said &quot;next review season, I&#x27;m going to go around and ask people who&#x27;s the [tenet] expert. It should be you, not him.&quot;<p>So although we were working together, we were also competing. And yeah, when he asked, I was the expert. I think a large part of that was because whenever the two of us got together to ping people about our tenet, it was in my office, so the emails came from me, and the bug updates were under my name. I did that half-intentionally, and feel a little bad about it.

That sounds exactly like the hell I went through at Raytheon. In fact, I may have heard that visibility quote verbatim during one review.<p>Raytheon didn&#x27;t have the problems described by Microsofties, though. The long-term employees (who were a majority at all the facilities I worked in) developed a sort of mass psychosis that was part Stockholm syndrome, part misguided patriotism, and part Dunning-Kruger. If you didn&#x27;t drink the Kool-Aid yourself the stack ranking system would be the least of your worries, though it was a useful tool for managers to deal with subordinates who wouldn&#x27;t comply.

An interesting variant of this system is one where executives give or take points away from organizations based upon team performance. Assume that individual rankings go from -2 to +2. The CEO gives each VP points according to his whim and&#x2F;or how well their organization did in the quarterly objectives. A favored VP may get +20 points. An underperforming one may get -3. Now he has to assign points to his directors, who get to give points to their managers, who give points to contributors. Now the sum performance review score at each subtree has to match the allocated points.<p>So you have fixed stacks, but they&#x27;re at least relative to team performance and&#x2F;or political favor. This system tends to depend on the initial allocations, so you should really get in the favored part of the organization.<p>Throwaway so you don&#x27;t know which company I&#x27;m talking about.

All stories about how bad stack ranking goes like this: &quot;I was a manager at X and all my subordinates were above average&#x2F;great and the system forced me to give them lower grades.&quot;<p>But it stands to reason that since the system measures performance <i>relative to the average</i> exactly half of the managers should have had the opposite problem: &quot;all my subordinates were worse than average but I had to give them higher grades than they deserved.&quot; I think the former problem is much more common because people tend to overvalue the people they are directly in contact with.<p>I suspect that if the managers could set the grade freely from 1-5, the an unreasonable amount of 4&#x27;s and 5&#x27;s would be given. In a large organisation like Microsoft&#x27;s the skill of the engineers and the ratings should follow a normal distribution.

Just to be fair here, the practice of applying stack ranking to very small populations was not standard across the company. I worked in Office for about 8 years. During those years I found that:<p>- Stack ranking is communicated up front and the process is very transparent. Managers let you know how the process works, what they&#x27;re doing, how you get bucketed, what they talk about behind the door, and will discuss your &quot;calibration card&quot; with you (the 3-5 high and low points they&#x27;ll bring into stack ranking meetings).<p>- The curve is not enforced on the level of a development team. It&#x27;s not enforced in a product group. You don&#x27;t see the hammer starting to come down until you get to populations of at least hundreds of employees.<p>I disliked the system (despite generally getting strong reviews) but it was not everywhere as demonic as it is depicted here.

Lived through this as an ex MSFT manager myself. The reality was having to throw direct reports under the bus and explaining to them the outcome was &quot;relative to their peers&quot;. Then they&#x27;d look at a smaller team and see people with much less impact getting good or great reviews (less competition in those teams). Its an aweful system and firmly believe it factors into MSFTs stagnant innovation. Lifetime average, peer reviewing and open transfers are what&#x27;s needed.