But well, stack ranking is not only about sorting (the name is a lie), it's about placing people in fixed sized cohorts. The real question is if there is a limitation on the number of raises or a minimum on the number of people treated as "low performant".
You do get to see the average size of cohorts, and you will need to explain why your team doesn’t meet them - for example if the entire team project is going exceedingly well, or the team is really made out if a lot of extraordinarily talented people - then you can argue why you have more than the average rate of excellent rankings.
Anyway, even if it honestly for "extraordinarily talented" people, that's still stupid stack-ranking.
The same you’d rank a person’s work, you would rank a team or org’s work.
If that team is doing really well, it’s likely they will have a distribution of higher rankings than the average. If they’re doing really bad, it can be the opposite.
The point is that if you finished ranking some team and your rankings are higher than average, you need to have a reason why.
What makes a ranking into a stack ranking?
The curve determines rating and compensation, influences promotion decisions, projects people get to work on and what opportunities come their way. The seems like a sort to me given that budget and project opportunities are fixed in many places.
EDIT: It sounds like at least some people are claiming that Meta has an internal policy that X% of your engineers under you are supposed to be given "Meets Most Expectations" ratings, which would also be pretty close to stack ranking (if at least allowing for some management leniency) https://news.ycombinator.com/item?id=32930370
With large enough populations you see fairly consistent statistical information across time. Little fluctuations even out. So at org sizes of, say, 500 people you'll see pretty consistent ratings distributions between cycles. This is where you get the "we have a roughly expected ratings distribution." If you've deviated far from this then it means either that your org is training people on the evaluation process incorrectly or there is some other cause of a weird distribution. When something doesn't match, people simply investigate for an alternative solution.
What people seem to be frightened of is when somebody says to a manager of 10 people "hey you've got too many high ratings, drop some." Google and Meta are massive companies so I suspect that this happens somewhere but I've never even got the scent of this happening from my position at Google. I've had cycles where 100% of my team was in the top 20% of ratings without anybody raising an eyebrow.
On stack ranking & the penalty for being in a great team - yes, also happened, being in a great team but you still had to rank someone lower, better to be in a worse team to get higher ratings. It had obviously negative impact, but there was some combination or organizational pride that had been cultivated, plus we thought we'd all be millionaires anyway so what did it hurt. But I bet some people had to leave who were the unfortunate victim of these affects.
A team of 20 people may have no low performers. An org of 400 probably has a couple, at some point the law of large numbers does really start to apply, and you're no longer in competition with the people you're working with on the daily.
If one member of your two-pizza team is going to get fired each year, it makes sense to throw someone under the bus on the reg. If one member of your 100 person org is going to get fired each year, that strategy doesn't make sense.
I tried for you, remember me when your next job is hiring managers.
Ideally you'd do this before calibration so the rating doesn't mess it up.
They are still super atheletes, but they are judged against other NBA teammates, not average professional basketballer in the world.
Graded in the context of peers’ performance does not imply that they sort employees by performance.