I'd dispute that many design changes can be proven in any meaningful way. There are too many variables, and optimising for clicks on one button might affect all

I'd dispute that many design changes can be proven in any meaningful way. There are too many variables, and optimising for clicks on one button might affect all others on a page. It's very hard to measure everything at once, and it doesn't sound like Google tried in this instance.