I disagree, this may sound a bit heartless, but it's the engineer's salary that should be measured here.
The hard part is translating frustrated users into financial cost (e.g. lost revenue when users are leaving for competitors and such) which in turn can be translated into hourly wage which then can be used to calculate if optimisation is worth it. And then compare to the lost opportunity of making other optimisations or improvements.
Obviously this is near impossible so the best you can do is make an educated guess.
The hard part is translating frustrated users into financial cost (e.g. lost revenue when users are leaving for competitors and such) which in turn can be translated into hourly wage which then can be used to calculate if optimisation is worth it. And then compare to the lost opportunity of making other optimisations or improvements.
Obviously this is near impossible so the best you can do is make an educated guess.