I think statistical significance is the more important thing in a/b testing, but they only included links to external sources. It's hard to tell the difference between coincidence and real difference between two versions when changing simple things like button colors and I still feel ignorant about the subject.