We rebuild the page closest to your sale and run it against the one you have now, on your own live traffic. The scoreboard is revenue per visitor, not a dashboard metric.
Usually the product page, because that is where the money is decided. We rebuild the highest leverage page rather than spreading changes thin across the store.
Old page and new page run at the same time on your own traffic, so seasonality and ad spend hit both equally. The result decides, not an opinion.
Two or three tests a month after the rebuild, each one measured the same way, so the page keeps improving against evidence instead of preference.
Fewer opinions, more tested changes: every rebuild has to beat the page you already have before it stays live.