The gains may be very real, but it stuck with A-B test problems. I don't know if there is anywhere that might mention the time period the data was collected, but if I ran my controls in Feb-Mar and experiment in Apr-June, I think anything this side of a parachute could show improvement
|