<b>Targeting your worst-converting page guarantees an apparent win — even from a placebo.</b>

<b>Targeting your worst-converting page guarantees an apparent win — even from a placebo.</b>
Regression to the mean: extreme low performers drift back upward on their own. Any change you ship gets credited with the rebound.
— Always keep a control on the same poor-performing cohort. The lift is the gap vs control, not vs last month.
— "We fixed the page and it recovered" is the most common un-controlled illusion in CRO.
No control, no causation.
Read the number, not the story. [concurrent control · same cohort]