In this article the importance of network components, such as links or nodes, is analyzed through the evaluation of some ''classical'' and other more recent metrics developed in the literature. In particular, the use of global sensitivity methods is suggested for quantifying the simultaneous effects of components. These assessments allow the decision-maker to identify key components, i.e. the components that affect one or more network performance functions. The article describes each metrics, highlights pro and cons and suggests how to select the most convenient one, based on four aspects: the objective of the decision-maker, the model characteristic, computation time and the information availability. A comparison among different importance measures in two networks is presented.