Experiments reach statistical significance faster when split evenly between control and treatment groups due to higher statistical power from variance reduction. Unequal splits prolong reaching significance and can negatively impact users. The recommendat