Electronic correlations are often discussed in terms of effective low-energy models, with the Hubbard model as the most prominent example. To simulate real materials, the Coulomb interaction U needs to be determined, taking into account the screening by high-energy electronic states. The constrained Random Phase Approximation (cRPA) is frequently used for this purpose, but its applicability has been questioned. Here I show that cRPA approximation can be justified when the electronic states responsible for the screening are energetically far away from the Fermi level. In that case, the electronic propagation length is exponentially short and provides a way to identify leading diagrams, similar to the analysis in diagrammatic extensions of dynamical mean-field theory. Nonlocal corrections to cRPA vanish. Local (excitonic) vertex corrections to cRPA exist, but are compensated by the underestimation of the band gap in density functional theory.