Tricky subject because so many projects pivot on the client expectations - which often don't match up to the written specification.
You could deliver the core product early and ask the client which areas they would like improving.
You could guess and produce something that is extended in areas you think are important.
You could just deliver early!
You could show the client what you have done and then the client may well point out that they had a different expectations of sections 2,3,6 and 7 - and could you re-work them?
I remember reading a case study on the iPhone UI project, where they decided to exceed expectations in areas that would make the phone seem special like the touch screen and zooming in/out - but almost ignore humdrum elements like cut-and-paste - and that was what lead to the huge approval ratings.
So if you are going to exceed expectations the iPhone lead is worth considering.