Is the Stacked Ranking of Employees Actually Bad?
Employees of Microsoft–actually, it’s usually former employees of Microsoft–hate the stacked ranking system which grades all employees on a curve. It runs something like this: In any given year, on any ten person team, only two are rated exceptional, and one is fired. Though I did not mention it in my Ballmer piece, some detractors blame Microsoft’s fall from grace on rank-and-yank.
We can take these claims at their surface meaning:
- Stacked ranking actually is bad.
But there are other explanations that would explain the presence of detractors:
- Microsoft had difficulties, and stacked ranking, were there to be blamed while the true causes (e.g., mean reversion) are hard to identify and psychologically unsatisfying.
- Only employees who felt shafted by the stacked ranking system are motivated to speak about it.
The typical way people choose among hypotheses like this is to attempt to reason it out. You will find plenty of posts on the Internet saying something like
Stacked ranking is awful because it fails to consider the possibility that some teams of ten people have either multiple or no bad performers. Similarly, mandating a specific number of commendations is silly. Stacked ranking is a straitjacket that forces managers to bend to it. Additionally, it leads to Darwinian politicking among employees as they jockey among one another.
Without some sort of concrete guidelines, managers (being human) are unable to really judge what constitutes extraordinary, ordinary, and unsatisfactory performance. This becomes all the more important when judging employees across teams. And any ratings scheme in which employees are evaluated can be subject to politicking.
These narratives are both plausibly convincing, but they are not above testing, which is what researchers attempt to do. What has been found (paywall) through experimentation is that forced distribution rating systems are perceived as more difficult and less fair by ratees, and raters are not fans either, feeling that the systems impede their ability to give proper feedback.
When comparing different forced distribution systems, ratees are most concerned (paywall) with how low-rated performers are treated. This raises the possibility that the firing of employees based on the system was the real problem at Microsoft, not the ranking itself.
After adopting a competitive, forced distribution system, employees may have a hard time adjusting back to a system that stresses cooperation. In general (paywall), “[i]t is more difficult for teams to change from competitive to cooperative reward conditions than it is for them to change in the opposite direction…”
On the other hand, it seems like people with higher cognitive abilities are more likely (paywall!) to be attracted to companies that use a forced distribution system. The system then, could be a way to attract better employees in the first place. And you can probably save money at the corporate library by getting bulk discounts on Ayn Rand novels.
That any management decision has mixed results is unsurprising. All difficult decisions are about trade-offs. But this doesn’t mean that there are multiple correct answers all equally defensible.
I think a wise CEO might split the difference. Managers do in fact need concrete guidelines for evaluating employees to ensure consistency. But that doesn’t mean those guidelines have no flexibility. One method of doing this would be point allocation. A manager of ten employees would have 100 points to allocate among them. This would allow flexibility in identifying both good and bad employees in a way that reflects real differences.
Another method might be to offer the manager three tiers into which to categorize his employees: A,B, and C with the knowledge that if he gives half the teams As, then each of those employees will not be rewarded in the same way as the A employee of a manger who only hands out one A.
Additionally, I’d make sure one of these two happen:
- The evaluation criteria include cooperative behaviors.
- Peer evaluations are collected to identify good teammates.
This would hopefully help address issues of politicking.
I would not make firing an explicit part of the system. Firings should happen perhaps every few years or perhaps just once per business cycle when it makes the most economic sense. Letting go of inevitable bad hires is difficult, but necessary. But one does not need to pre-commit to firing 10% of the company each year to do that.
Photo credit: Flickr user gageskidmore