Regarding A/B testing, you don’t need huge numbers, just a huge change…
A subtle difference won’t show up unless you have the big visitor numbers, so it’s best for more radical changes.
The problem is you start to wonder “Was it the new headling, or that we moved and changed the color of the button/s?”
That’s where multivariate testing comes in but for that you really do HAVE to have a lot of visitors, at least 100,000.
APC