Stack ranking is a performance cycle that orders people against each other and fixes in advance how many may land in each band, rather than rating each person against what the role is supposed to deliver. It is a design choice made before any conversation happens, and it changes the meaning of every rating that comes out of the cycle. This page is for the person deciding whether to run one, not for anyone trying to survive one, and it is about the arithmetic and the second-order costs rather than about the politics.
What a forced distribution actually does
A forced distribution takes the ratings managers would have given and pushes them onto a fixed shape, so a team where everybody met the expectation still produces a bottom band. That is the whole mechanism and it is also the whole objection. Used as a calibration prompt, where managers have to defend why their team looks different from the shape, it surfaces real inconsistency. Used as a rule that must be satisfied, it manufactures a difference that the evidence did not contain.
It breaks first at small team sizes
A ten percent bottom band is a coherent idea across four hundred people and an incoherent one across a team of six, where it means exactly one person every year regardless of what happened. Most organizations that adopt stack ranking adopt it at the company level and then apply it at the team level, which is where it stops being a distribution and starts being a rota. If your average span of control is under about ten, the shape has to be enforced above the manager or not at all.
The costs that arrive a year later
Two are predictable. The first is that people stop helping each other on anything that is scored, because a colleague's good year is now expensive. The second is that managers start hiring a weaker candidate to protect a stronger incumbent, which is the single most expensive thing this design can cause and the hardest to see in any report. Neither shows up in the cycle that produced it, which is why they are usually attributed to something else.
What to use instead if the goal is comparability
The reason most employers reach for a curve is that their managers rate inconsistently and pay depends on the rating. Calibration solves that directly: managers in one room, defending their own distributions against each other, with the evidence. It gets comparability without forcing a shape, it takes a few hours a cycle, and the record it produces is a set of reasons rather than a set of quotas.
Questions people ask about stack ranking
Is stack ranking the same as calibration?
No, and the difference is the point. Calibration compares how managers rate and asks them to justify the difference. Stack ranking fixes the answer in advance and makes the managers fit their people to it.
Does stack ranking need to be published to work?
If it decides pay, people find out, so an unpublished curve is usually just an undefended one. Employers who run a distribution tend to do better by saying so and explaining how it is applied.
What team size is too small to stack rank?
Under roughly ten reports the band sizes stop being meaningful: a bottom decile is one specific person every year. Below that, either enforce the shape at a level with enough people or use calibration instead.