Regression to the Mean

Many things vary by luck around an average. A very high or very low reading usually includes some luck, and the next reading is likely to be less extreme. This is regression to the mean. It does not need any cause.

It fools people who act when things are worst. A shop has its worst sales week, so the owner changes the window display. Next week sales are up. The display may have helped, but sales would probably have risen anyway.

To judge a fix fairly, compare with a group or period where no fix was made. If the two improve alike, the change was probably luck. Be especially careful with results chosen because they were extreme.

The worst week

Ali’s shop has its worst week in a year, so he repaints the sign. The next week sales climb, and he tells everyone that the paint worked.

His neighbor, who did nothing, had a slow week too and also bounced back. Ali compares his figures with hers and sees that both recovered alike. The paint may help, but the bounce alone does not show it.

Try this today Before crediting a fix, ask what the result would have been if you had done nothing.