Sinofski on performance reviews


Since I left Microsoft a little over a year ago, I’ve written a few rants that I’ve never published. Amongst the subjects I tackled were some attempts at “what’s wrong with Microsoft’s performance review system?”. None of these efforts were worth posting; I guess I just needed to get it out of my system.

So here is a well thought out post by Steven Sinofski (ex of Microsoft) on the theory and practice of a performance review system. It’s level-headed, insightful, and definitely worth a read.

A couple of points of my own:

Space it out. Microsoft should seriously consider ditching the annual simultaneous review of everyone in the company. MS should do something like review people on their hiring anniversary (or at least, in initially random cohorts spread over the year). Stack rankings could happen more often (2-3 times a year?) so that decisions would have enough data.

This would dramatically reduce the political turmoil around the Big Review. It might give meaning to the self review system (which should otherwise be junked, as it is merely a post-facto substrate for the stack rankings which generally happen a month or more before people write their own reviews).

Eliminate the curve. Obviously the fixed stack rank buckets need to go. This has been said before. Conflating stack ranking and firing is a narrow way to look at things (there are valid reasons you’d want to fire high performers, for instance, and equally valid reasons to retain someone who’s not doing well).

Learn how to fire the right people. Microsoft conflates stack ranking with identifying the people to get rid of. I think of ranking as purely a gauge of comparative performance so that you can figure out how much to pay. You might want to get rid of people in the low buckets, or you might want to keep them all — it depends on the group. There are circumstances where you need to fire high performers, too (especially in situations where they are toxic or doing damage).

I keep hearing rumors that MS is going to make revisions to its review system (apparently people are turning down offers because of it). It’ll be mildly interesting to see if it improves.

This entry was posted in Uncategorized. Bookmark the permalink.

3 Responses to Sinofski on performance reviews

  1. Ingo Karkat says:

    RE Space it out: Years ago, the tech company I work for (comparable to Microsoft) moved to simultaneous year-end reviews, arguing that the varying age of review results are unfair when salary increases and bonuses are decided at a single point in time.

    Since then, managers are complaining (privately) that all those reviews have to be done on top of their regular work, simple things like finding a meeting room become a challenge during that period, and in general the depth and breadth of the performance review has suffered greatly. Apparently, nobody has yet told HR that the cure is worse than the original problem…

  2. Xelous says:

    All very interesting… Just one thing… What’s a review?… I work in a UK company, owned by a large Euro conglomerate, and we have no seeming review system what so ever, even painfully obvious cases of incompetence and unrest go unattended, so the chance of being reviewed in anyway is remote in the extreme.

  3. Lee says:

    I work in a UK company, owned by a large US corporate, and we do have a review system, where even painfully obvious cases of incompetence and unrest go unattended!

    I also used to work at MSFT Redmond and left because of the company’s review system. I also wrote up a critical analysis of the review system at the time.

    Among the many faults are that:

    * a review is subject to the manager’s ignorance/knowledge of their direct effectiveness (rough if you have non-technical managers reviewing directs’ work on technical subjects)

    * a review is subject to the manager’s own fear or lack of fear for the direct, the manager’s liking for or disliking of the direct, or even the manager’s sense of what his or her’s own manager would like to see in the review.

    Crap system for the above pragmatic reasons.

    Interesting side note: this system was originally designed by Charles Bedaux with the express purpose of delivering as much human performance increase for the lowest incentivisation cost as possible. For that reason, the system – and Charles Bedaux himself – was very close to the Nazis and their supporters in the anglophone world.

    You can still see traces of evidence for the above statement if you carefully read:

    and you can find more evidence if you read:

Comments are closed.