One way to prove a theorem in analysis is to initially recognize what can go wrong. In this talk we will discuss a few recent results that follows a detailed version of this approach to establish existence of maximizers of some functionals.