4 ms·

What look like complex failures in organizations are often rooted in simple causes. In Microsoft, it's stack-ranking and Enron-style (that is, visible in transf

What look like complex failures in organizations are often rooted in simple causes. In Microsoft, it's stack-ranking and Enron-style (that is, visible in transfer packet) performance reviews. When you have stack-ranking, you never get harmonious teams and "all-star" sub-organizations are out of the question. When performance reviews are part of the transfer process, people become largely immobile and teams become permanent camps and you get warring departments. Closed allocation, at that scale, is cultural toxic sludge in general. Simple causes. Simple fixes. Wordy say-nothing emails that inject political complexity, when simple solutions are available, are really not helpful.

hga 13y ago
BTW, Wikipedians considered Microsoft's stack ranking notorious enough to provide the date they started doing it (http://en.wikipedia.org/wiki/Vitality_curve#Microsoft http://en.wikipedia.org/wiki/Vitality_curve#Microsoft): "Starting in 2006 Microsoft has used a Vitality Curve despite intense internal criticism." Which I've noted elsewhere is the year in which Bill Gates stopped his day to day roles at Microsoft. I could talk about how Ballmer himself is a significant problem, but then again these "simple fixes" would surely require removing him from an position of direct power in the company.
- mikestew 13y ago
  Microsoft has been doing stack ranking long before 2006. I left in 2005'and attended stack rank meetings, where managers indirectly decide your review score before you even get the form from HR. it goes back to at least the late 90s, IIRC.
  - hga 13y ago
    Were those in the bottom rank shown the door in due course?
    - mikestew 13y ago
      There was certainly mention of "up or out", and a bad review or two would bring about the PIP stick (which seems to be only legal CYA before the firing, not for any hopes of real improvement). And, frankly, it was a long time ago and I haven't gone to any effort to remember details. I will say that in general it was as screwy and demotivating then as it is described to be now. It was based on a model proposed by Ballmer's hero Jack Welch. What Ballmer forgot is that Welch himself said the "trim the bottom 10%" model didn't work once a company got larger than a certain size.
      - hga 13y ago
        Thanks! The only detail I was interested in was the timing; the Wikipedia citation of 2006, which I will stop using, did strike me a bit recent.
      - michaelochurch 13y ago
        What Ballmer forgot is that Welch himself said the "trim the bottom 10%" model didn't work once a company got larger than a certain size. I'm not sure that size is the issue. I think that the percentage of deadwood a company has being static is the problem. A typical 10,000-person company probably has enough deadwood to make a 10% cut safe: a mature company that's accrued enough slackers and incompetents can endure a cut of the obvious nonperformers and few will be upset about it. Even many of the nonperformers will agree, to some extent, that it's time. Besides, the company will have a cut anyway if it underperforms. But if you cut 10% every year, after two or three iterations you are cutting good people, and the cultural effects of that make the process not worth it.
kvb 13y ago
This seems simplistic to me. Doesn't Valve, your favorite poster boy for open allocation, also stack rank? There seem to be several sub-organizations at Microsoft that are harmonious, at least to an outsider's view (e.g. Azure since Scott Guthrie's involvement). As I understand it there's not too much pressure to hit the stack ranking curve until you get up to the several-hundred-report levels.
- michaelochurch 13y ago
  Valve stack-ranking is for compensation only-- not a "fire the unluckiest X percent" system. That can still be damaging, if low rank results in permanent underclass status, but that's a human problem that I have no insight into as it pertains to Valve. I don't know enough about Valve to know how it works, or even if the culture lives up to the press. So I'm not going to try to comment on that. In general, though, stack-ranking is destructive even if no one gets fired. In companies that use it, it becomes impossible to transfer without a top-10% political success review (sorry, I mean "performance" review) history. But if you're doing that well politically, then you don't want to transfer because it entails rolling the dice again. The result is that people become pretty much immobile. Closed allocation actually forces engineering ladders (with their attendant negatives) into existence because unsuccessful/persecuted people can't move and successful people won't move at all unless they can get a permanent credibility bump (promotion). But that leads to an arrangement where managers avoid promoting their best people because they know it might cause them to transfer.
bornhuetter 13y ago
Stack ranking seems, to me, to be an incredibly bad idea for Microsoft as a whole. Do you know if they are considering moving away from it? I think Stack Ranking can work in some situations, like mechanical (ie. non-heuristic) work, but for Engineers? I've never understood why they even considered that it would work.
- kvb 13y ago
  How do you ensure that performance is rewarded fairly and consistently across an organization with tens of thousands of engineers? I'm not wild about the concept of stack ranking but I don't see any alternatives that don't have serious drawbacks of their own.
  - bcoates 13y ago
    Is that a goal? I'd rather own a company where the pay was unfair and inconsistent but it delivered on products and made money. I'd rather work for one, too. It's not like Jack Welch's worst idea isn't a driver of politics and strategic[1] moves in the real world. [1] as in, bad for the company but the most rational move for the employee, like attempting to make sure every person added to your team is worse than you.
  - kyllo 13y ago
    How about evaluating each engineer's job performance (based on sane productivity and code quality metrics) without forcing them to a curve, where even if your whole team is high-performing, someone will get a bad review and be marked for firing just because it's required that someone gets a bad review? Jack Welch-style rank-and-yank is cruel and wildly inappropriate for software developers (or any human workers, I'd argue). It encourages political blame-shifting/scapegoating games that distract from the actual work. Just having the practice is also an implicit acknowledgement that your hiring process blows. Performance curves are just a demented and sadistic way to manage people. They're also the epitome of b-school cargo culting.
    - kvb 13y ago
      The problem with this is that it enables weak-willed managers to claim that everyone on their team is great, serving as a kind of grade inflation that makes true high-performers seem less impressive. As I understand it, the curve is only enforced at something like the VP level, so it's not like a team of 10 great engineers is going to be forced to have 2 people scored as underperforming. Are there 200-person groups at Microsoft where everyone's better than average?

hga 13y ago

BTW, Wikipedians considered Microsoft's stack ranking notorious enough to provide the date they started doing it (http://en.wikipedia.org/wiki/Vitality_curve#Microsoft http://en.wikipedia.org/wiki/Vitality_curve#Microsoft): "Starting in 2006 Microsoft has used a Vitality Curve despite intense internal criticism." Which I've noted elsewhere is the year in which Bill Gates stopped his day to day roles at Microsoft. I could talk about how Ballmer himself is a significant problem, but then again these "simple fixes" would surely require removing him from an position of direct power in the company.

mikestew 13y ago
Microsoft has been doing stack ranking long before 2006. I left in 2005'and attended stack rank meetings, where managers indirectly decide your review score before you even get the form from HR. it goes back to at least the late 90s, IIRC.
- hga 13y ago
  Were those in the bottom rank shown the door in due course?
  - mikestew 13y ago
    There was certainly mention of "up or out", and a bad review or two would bring about the PIP stick (which seems to be only legal CYA before the firing, not for any hopes of real improvement). And, frankly, it was a long time ago and I haven't gone to any effort to remember details. I will say that in general it was as screwy and demotivating then as it is described to be now. It was based on a model proposed by Ballmer's hero Jack Welch. What Ballmer forgot is that Welch himself said the "trim the bottom 10%" model didn't work once a company got larger than a certain size.
    - hga 13y ago
      Thanks! The only detail I was interested in was the timing; the Wikipedia citation of 2006, which I will stop using, did strike me a bit recent.
    - michaelochurch 13y ago
      What Ballmer forgot is that Welch himself said the "trim the bottom 10%" model didn't work once a company got larger than a certain size. I'm not sure that size is the issue. I think that the percentage of deadwood a company has being static is the problem. A typical 10,000-person company probably has enough deadwood to make a 10% cut safe: a mature company that's accrued enough slackers and incompetents can endure a cut of the obvious nonperformers and few will be upset about it. Even many of the nonperformers will agree, to some extent, that it's time. Besides, the company will have a cut anyway if it underperforms. But if you cut 10% every year, after two or three iterations you are cutting good people, and the cultural effects of that make the process not worth it.

kvb 13y ago

This seems simplistic to me. Doesn't Valve, your favorite poster boy for open allocation, also stack rank? There seem to be several sub-organizations at Microsoft that are harmonious, at least to an outsider's view (e.g. Azure since Scott Guthrie's involvement). As I understand it there's not too much pressure to hit the stack ranking curve until you get up to the several-hundred-report levels.

michaelochurch 13y ago
Valve stack-ranking is for compensation only-- not a "fire the unluckiest X percent" system. That can still be damaging, if low rank results in permanent underclass status, but that's a human problem that I have no insight into as it pertains to Valve. I don't know enough about Valve to know how it works, or even if the culture lives up to the press. So I'm not going to try to comment on that. In general, though, stack-ranking is destructive even if no one gets fired. In companies that use it, it becomes impossible to transfer without a top-10% political success review (sorry, I mean "performance" review) history. But if you're doing that well politically, then you don't want to transfer because it entails rolling the dice again. The result is that people become pretty much immobile. Closed allocation actually forces engineering ladders (with their attendant negatives) into existence because unsuccessful/persecuted people can't move and successful people won't move at all unless they can get a permanent credibility bump (promotion). But that leads to an arrangement where managers avoid promoting their best people because they know it might cause them to transfer.

bornhuetter 13y ago

Stack ranking seems, to me, to be an incredibly bad idea for Microsoft as a whole. Do you know if they are considering moving away from it? I think Stack Ranking can work in some situations, like mechanical (ie. non-heuristic) work, but for Engineers? I've never understood why they even considered that it would work.

kvb 13y ago
How do you ensure that performance is rewarded fairly and consistently across an organization with tens of thousands of engineers? I'm not wild about the concept of stack ranking but I don't see any alternatives that don't have serious drawbacks of their own.
- bcoates 13y ago
  Is that a goal? I'd rather own a company where the pay was unfair and inconsistent but it delivered on products and made money. I'd rather work for one, too. It's not like Jack Welch's worst idea isn't a driver of politics and strategic[1] moves in the real world. [1] as in, bad for the company but the most rational move for the employee, like attempting to make sure every person added to your team is worse than you.
- kyllo 13y ago
  How about evaluating each engineer's job performance (based on sane productivity and code quality metrics) without forcing them to a curve, where even if your whole team is high-performing, someone will get a bad review and be marked for firing just because it's required that someone gets a bad review? Jack Welch-style rank-and-yank is cruel and wildly inappropriate for software developers (or any human workers, I'd argue). It encourages political blame-shifting/scapegoating games that distract from the actual work. Just having the practice is also an implicit acknowledgement that your hiring process blows. Performance curves are just a demented and sadistic way to manage people. They're also the epitome of b-school cargo culting.
  - kvb 13y ago
    The problem with this is that it enables weak-willed managers to claim that everyone on their team is great, serving as a kind of grade inflation that makes true high-performers seem less impressive. As I understand it, the curve is only enforced at something like the VP level, so it's not like a team of 10 great engineers is going to be forced to have 2 people scored as underperforming. Are there 200-person groups at Microsoft where everyone's better than average?