In this paper we analyze the PageRank of a complex network as a function of its personalization vector. By using this approach, a complete characterization of the existence and uniqueness of fixed points of PageRank of a graph is given in terms of the number and nature of its strongly connected components. The method presented includes the use of a feedback-PageRank in order to compute exactly the fixed points following the classic Power's Method in terms of the (left-hand) Perron vector of each strongly connected components.