Galley checks whether your A/B test really won

Field notes

Galley accepts experiment exports rather than asking users to re-enter counts, with manual column mapping available when an export does not match a recognized shape. Its report combines a plain-language verdict with confidence and corrected p-value details, then checks revenue per visitor and average order value alongside conversion rate. The page also describes a sample ratio mismatch check, intended to flag uneven traffic allocation before the experiment’s apparent result is interpreted.