Yeah, and then when you point out that the B version has to be built, put into production, and a tool paid for that will feed A and B alternately for hundreds of tests to have any statistical validity… oh, never mind.