Index

Mental Models

Regression to the Mean

When a variable is extreme on one measurement, it will tend to be closer to the average on the next.

Regression to the mean prevents overreaction to outlier events by reminding you that variance naturally pulls results back toward the average.

Core question

Is this extreme result likely to persist, or is it partly random and due to revert?

Example

A sales rep has a record quarter. Before restructuring the entire team's strategy around her approach, consider that some of that performance is random variance that will naturally regress.

How to apply

  1. 1.Identify whether the outcome contains a significant random component.
  2. 2.Compare to historical averages and distributions.
  3. 3.Avoid drastic policy changes based on a single extreme data point.
  4. 4.Wait for repeated observations before concluding a true shift has occurred.

Pitfalls

  • ·Attributing regression to whatever intervention happened to coincide with it.
  • ·Ignoring genuine trend breaks because you assume everything reverts.
  • ·Applying regression logic to processes that are skill-dominated, not luck-dominated.

Questions

What is a common regression to the mean mistake?

Praising a manager for improvement after a bad quarter, when performance would have bounced back regardless of any intervention.

Does regression to the mean apply to skill-based activities?

Only partially. The larger the luck component in an outcome, the stronger the regression effect. Pure skill activities regress less.

Related