These might be evident in retrospect, but it's usually tricky to predict what will matter and what won't during design time. (Someone smarter than me said something about premature optimization once.)

This is also usually where experience can make a big difference.