Regression to the Mean

Introduction: Regression to the Mean

Regression to the Mean is a statistical phenomenon in which extreme observations are likely to be followed by values closer to the average. It occurs when a variable that is unusually high or low on its first measurement tends to move nearer to the mean on subsequent measurements, purely due to chance variation. Understanding this effect is crucial in Lean Six Sigma and other analytical fields to avoid misinterpreting natural fluctuations as genuine process improvements or declines.

Background

The concept was first identified by Sir Francis Galton in the 19th century during his studies on the relationship between parents’ and children’s heights. He observed that very tall parents tended to have shorter children, and very short parents tended to have taller ones—both closer to the population average. Galton called this effect “regression toward mediocrity,” which later became known as regression to the mean. Since then, it has been recognised as a common pattern in fields such as finance, psychology, healthcare, and manufacturing.

Key Elements / Features

• Statistical Nature: Regression to the Mean occurs when random variation causes extreme values to shift closer to the mean upon re-measurement.
• Not a Real Change: The shift does not imply true improvement or decline but reflects natural variability.
• Relevance in Process Data: Appears when performance metrics fluctuate due to random noise rather than actual process changes.
• Common Misinterpretation: Managers may incorrectly attribute the return to average performance to new policies or corrective actions.
• Analytical Awareness: Recognising this phenomenon prevents false conclusions in experiments or performance reviews.

Applications / Examples

• Healthcare: Patients with very high blood pressure readings often show lower values later without treatment, simply due to natural variation.
• Business Performance: Sales teams performing exceptionally well in one quarter often show lower results the next, even without any strategy change.
• Manufacturing: A machine showing unusually high defect rates one day may produce normal results the next, without intervention.
  

Example: In a Six Sigma project, a spike in defects one week followed by a drop the next may appear as an improvement, but it could simply be regression to the mean.

Relevance / Impact

Recognising regression to the mean is vital for accurate data interpretation. It prevents teams from making unnecessary changes based on random fluctuations and helps focus on genuine process improvements. By accounting for this effect, organisations can make more reliable, evidence-based decisions and avoid overreacting to temporary highs or lows in performance data.

See also

• Variation
• Control Charts
• Control Limits
• Statistical Process Control (SPC)