You have to wonder if its even possible for the same tools that you use to A/B test small changes (i.e. in UI, menus, branding language) can also be applied for these more broad systemic changes. What it sounds like a lot of the software mentioned in this article does is to maximize locally. Is it even possible to then use the same process to maximize globally? If the answer is yes, then companies should ask themselves if they're willing or able to make such large systemic changes on such little data. I wonder how many people would be able to make such game changing decisions off of the results of these A/B testing surveys. My guess is probably not a lot.