I mean, advertising is a bit infamous for this. Outside of immediate-action advertising ("Want to buy this?" and "Click here"), it's very hard to prove the effects of a given program.

Sales boosters like coupons are already tricky - did they create new buyers, reengage existing buyers, or do customers just hold off on buying until there's a discount? And brand awareness ads are even harder - did your Toyota add two years ago drive a sale today? You can't even survey for that, since in theory it works on people who don't remember it consciously.

So I have a dark suspicion that advertising is more full of stupid metrics than most roles, simply because real analytics are harder to produce.