Bruce Gyory explains how dubious was the (pretty much one-poll) evidence of movement in Trump's approval rating here:
An interesting discussion of what polls can get right--and what they can get wrong, and what they did get wrong in Iowa, and what they might get wrong going forward--from veteran NY politico Bruce Gyory.