[thayne]

IME a/b tests are often run by people with little to no knowledge of statistics or experimental procedure. It is pretty easy to end up backing bad decisions with data when you don't completely understand the data.