The medieval tournament had a function beyond entertainment: it kept the knights competing with each other for the lord's favor rather than comparing notes about the lord. Stack ranking, the practice of forcing every team's members onto a curve with a mandated bottom, does the same work inside a company. Whatever its stated purpose (rigor, meritocracy, fighting grade inflation), its structural effect is to relocate the employee's principal rival from the market outside to the desk beside her.
The design deserves to be stated plainly. Under a forced curve, my colleague's excellent quarter is bad news for me; helping her may cost me my rating, my raise, or in the harsher variants my job. The company has, deliberately or not, priced cooperation negatively. It has also guaranteed a permanent supply of internally generated fear, since someone must occupy the bottom bucket regardless of how well the team performed. Tournaments produce champions; they do not produce teams, and most modern work is done by teams.
The evidence on manufactured inequality
The clearest evidence on what differentiated rewards do to working groups comes from a field experiment run with manufacturing workers in India. Breza, Kaur and Shamdasani (2018) randomized whether co-workers in a unit were paid equally or differently. When pay differences did not track observable productivity differences, output fell by about 0.45 standard deviations and attendance dropped 18 percentage points; cooperation eroded even where cooperating served the workers' own interests. But when workers could see that the higher-paid colleague was genuinely more productive, the inequality did no damage. The finding is precise: it is not inequality that poisons a group, it is inequality the group cannot verify as just. Stack ranking, whose verdicts emerge from closed calibration meetings and relative judgments no one outside the room can audit, is almost a machine for producing exactly the unverifiable kind.
The second cost is quieter. Tournament cultures teach people what happens to those who surface problems: bad news attaches to its bearer, and the bearer is standing on a curve. The literature on employee silence, meta-analyzed across 168 samples and 63,134 employees, associates silence with worse job attitudes, lower well-being and worse performance, with fear-related antecedents doing much of the explanatory work (Hao et al., 2022). And the foundational research on psychological safety showed why this is a performance issue rather than a comfort issue: teams learn (report errors, ask for help, experiment) only where interpersonal risk feels survivable, and learning is the channel through which safety becomes results (Edmondson, 1999). A ranking system that makes every admission of difficulty a tournament entry converts the team's error stream, its most valuable diagnostic information, into a liability to be hidden.
Reversing the direction of the gaze
The standard defense of forced ranking is accountability: without the curve, who tells the truth about performance? The question smuggles in an assumption, that evaluation must flow downward, from the person with power over careers onto the people beneath. There is a documented alternative that keeps the rigor and reverses the direction. As Ricardo Semler has told it, everyone at Semco anonymously evaluated their boss twice a year. Two consecutive low scores meant losing the management role, a rule that applied to Semler himself, and returning to well-paid technical work was a legitimate, non-humiliating exit. Separately, teams could veto a manager. The first year was uncomfortable; then, in his telling, the managers who used to shout began to ask, and those who could not adapt went back to doing good technical work at the same salary.
Notice what this design accomplishes that the stack rank cannot. It applies competitive pressure where power is concentrated rather than where it is absent. It makes the evaluation verifiable to the evaluated group, the exact property Breza's experiment identifies as the difference between motivating and poisonous differentiation. And it removes the tournament from the level where cooperation happens. Colleagues at Semco had no structural reason to withhold help from each other; the only person competing for standing was the one with authority over others' days, which is where a healthy system wants its competition.
The house always won anyway
The tournament's last defect is that its prizes were smaller than advertised. The knights at the tourney were competing for favor, not for the manor; the manor was never on offer. Stack ranking distributes modest differentials at the cost of the collaborative surplus an entire team could have produced, and the surplus was the larger number. Companies that abandoned forced curves did not do so out of kindness. They noticed they had been paying a fortune, in silence, hoarded information and sabotaged cooperation, to run a contest whose main output was fear of the person one desk over. A company gets the rivalries it designs. The only question is whether it designed them on purpose.
Sources
- Breza, Kaur & Shamdasani (2018). The Morale Effects of Pay Inequality (QJE 133:2). Randomized field experiment: pay inequality not justified by observable productivity cut output ~0.45 SD and attendance by 18pp and eroded cooperation; verifiable justification eliminated the harm. https://academic.oup.com/qje/article-abstract/133/2/611/4430649
- Hao, Zhu, He, Duan, Zhao & Meng (2022). When Is Silence Golden? A Meta-analysis on Antecedents and Outcomes of Employee Silence (J. Business and Psychology). Across 168 samples and 63,134 employees, silence associates with worse attitudes, well-being and performance, with fear-linked antecedents prominent. https://link.springer.com/article/10.1007/s10869-021-09788-7
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams (ASQ 44). Team psychological safety enables learning behavior, which mediates its effect on performance. https://journals.sagepub.com/doi/10.2307/2666999