It is possible that overuse of A/B testing results in a local maximum, with sweeping changes needed to reach the global maximum.
That's not overuse - that's just bad use :-) There is nothing to stop people split testing radical redesigns.
That's not overuse - that's just bad use :-) There is nothing to stop people split testing radical redesigns.
What I was trying to say was that the problem isn't with A/B testing - it's with overly simplistic usage and interpretation of A/B testing.