A third option is that they are so successful for reasons unrelated to their UI, and that the A/B testing is actually just not having the desired effect of crea

A third option is that they are so successful for reasons unrelated to their UI, and that the A/B testing is actually just not having the desired effect of creating a good UI. For example, it could easily be the case that A/B testing is causing them to introduce changes that have a short term positive effect (such as more people buying a product when they visit the site, or even just changes that have a temporary "novelness" factor which wears off after they've been in use for awhile), but a long term negative effect (such as a gradually built dislike of the site causing them to be more likely to shop elsewhere a year or two down the line), and only the former shows up.