The Mean-Reversion Feedback Trap
Separate natural performance swings from the apparent effects of feedback
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 5
- Confidence
- 98%
People commonly receive praise after an unusually strong performance and criticism after an unusually weak one. Even if the feedback has no effect, the next result will often move closer to the person's average: exceptional performance tends to decline, while poor performance tends to improve. That sequence falsely teaches leaders that criticism caused improvement and praise caused deterioration. The framework counters this trap by establishing a baseline, treating extreme outcomes cautiously, and evaluating feedback over repeated comparable cases. Leaders then focus criticism on specific controllable behavior, reward sound process, and avoid drawing causal conclusions from a single before-and-after pair. The result is a fairer coaching environment and a more accurate understanding of what actually changes performance.
Origin
Extracted from Masters in Business through Michael Lewis's explanation of Daniel Kahneman's lesson to Israeli flight instructors about regression to the mean.
Core principles
- 01Extreme performance often moves back toward normal
- 02Sequence does not establish feedback effectiveness
- 03Praise and criticism are delivered at systematically different moments
- 04Evaluate feedback across many observations
- 05Protect people from punishment justified by statistical noise
How to run it
- 1
Establish the baseline
Estimate the individual's ordinary performance range using multiple observations rather than the latest result.
Pro tip Use comparable tasks and conditions whenever possible.
Watch out A weak baseline makes every later judgment vulnerable to noise.
- 2
Identify an extreme
Determine whether the praised or criticized result was unusually far from the baseline.
Pro tip Treat both exceptional success and exceptional failure as possible outliers.
Watch out Do not assume an extreme result reflects a permanent change in ability.
- 3
Expect regression
Before giving feedback, predict how much of the next movement could occur naturally as performance returns toward normal.
Pro tip Write the prediction down before seeing the next result.
Watch out Regression is a tendency, not a guarantee for every observation.
- 4
Target controllable behavior
Base coaching on specific actions or processes that the person can change, not merely on the outcome's distance from average.
Pro tip Ask what observable behavior differed from the normal routine.
Watch out Harshness can appear effective when it merely precedes a natural rebound.
- 5
Test feedback across cases
Compare outcomes after similar feedback over many observations before deciding that a coaching method works.
Pro tip Track process adherence, morale, and results together.
Watch out Do not generalize from a single dramatic turnaround.
In the wild
Instructors observed that pilots often improved after being yelled at for poor performance and worsened after receiving praise for excellent performance. Kahneman explained that unusually bad and unusually good performances would naturally tend to move back toward each pilot's average, creating a misleading lesson about feedback.
→ Mean reversion offered a statistical explanation for the apparent superiority of punishment.
After learning the mechanism, Lewis reduced criticism while coaching children and softened the criticism he still used. He could not prove that performance improved, but he reported that the environment became happier.
→ The coaching climate improved without relying on punitive feedback falsely validated by noisy results.
Common mistakes
Crediting the latest intervention
A rebound immediately after criticism may have happened without the criticism because the preceding result was unusually poor.
Treating regression as destiny
Mean reversion describes an aggregate tendency and cannot predict every next performance with certainty.
Ignoring genuine behavioral problems
Statistical caution should not prevent specific, constructive intervention when controllable conduct is clearly harmful.
Is it for you?
Best for
It is best for managers, teachers, coaches, and parents evaluating fluctuating human performance.
Not ideal for
It is not ideal when a clear mechanical intervention has an immediate and independently measurable effect.
From the transcript
“life has this horrible feedback loop. This lesson. It teaches us that we get response to our punishment but not to our praise”
“mean reversion will make will tend to make them better after you've criticized them, and tend to make them worse after you've praised them.”
“I basically stopped. I stopped with a criticism except except when it was really lathered in and when it almost sounded like praise.”
From the episode
Michael Lewis Discusses the Culture of Finance (Podcast)