Pretty doesn’t necessarily mean effective at converting. I think Amazon’s team is perfectly competent at testing whether one design makes more money than another — and I think they’re competent enough to go with the one that makes more money.
I have a suspicion they A/B test the little things, like is it better to show the price in dollars or the local currency, and is it better to have a 'add to cart' or 'buy it now' button.
They miss the big changes like 'shall we show all the products in a 3d immersive VR gallery' or 'make a site which downloads all data and works entirely offline'.
Those big changes might usually fail, but if you never try any of them, you'll end up with a site that looks and behaves like it's from the 1990's, while the rest of the world has moved on.