How A/B testing helped a price comparison app increase revenues by 28%

So one of the tests' result was that the control version returned more hits. Do you think too many variations can spoil the broth?