Regression to the mean
Регрессия к среднему
Extreme observations are usually followed by less extreme ones, regardless of any intervention. Mistake regression for causation and you'll see patterns that aren't there.
If a sales team has an extraordinary quarter, the next quarter will almost certainly be worse - not because anyone failed, but because extraordinary requires luck plus skill, and luck doesn't repeat. Same for an unusually bad quarter: the next will usually be better. The «cause» is statistical, not managerial.
Managers reward what worked after a slump and punish what didn't after a peak. They're rewarding regression, not behaviour. They learn the wrong lessons.
A pilot is praised after a great landing and criticised after a poor one. Statistically, the next landing in each case will be closer to average. The instructor concludes praise hurts and criticism helps. Wrong cause.
- 1Before crediting any intervention, ask: «what would have happened with no intervention, given regression?»
- 2Use multiple data points, not one peak or one valley.
This part of the knowledge base is inspired by the book. Go to the Ukrainian edition to explore the concept in depth.
- · Методичка по Канеману - регрессия к среднему