The cynical side of me thinks this is the problem with the "landing page optimization" market. Measure, tweak, measure, tweak (repeat 20x) and before you know it, you're back to where you began!

A couple years back, I read another article about landing pages that warned against taking A/B testing logic to the extreme, as it may lead you down the wrong path.

My advice: Use your instincts/experience/feedback along with raw data.