I'm assuming that BEOL variations aren't generally correlated. And that this kind of analysis would be useful.