We show that the standard Value Function Iteration (VFI) algorithm has difficulties approximating models with jump discontinuities in policy functions. We find that VFI fails to accurately identify the location and size of jump discontinuities while other methods - such as the Endogenous Grid Method (EGM) and a Finite Element Method (FEM) - are much better at approximating this class of models. We illustrate differences across methods using a standard plant-level investment model with both variable and fixed capital adjustment costs. We find that the policy functions generated by VFI are quite different from those generated by EGM and FEM. Importantly, these differences are economically significant: for our baseline parameterization VFI generates investment spikes that are 5-8% larger in comparison to the other two methods. The choice between EGM and FEM depends on the context. While EGM is faster than FEM, it is much more difficult to implement. For larger models, the modifications necessary to apply EGM can lead to high code complexity. On the other hand, FEM can accommodate larger models with minimal implementation differences and its high scalability can reduce computation time significantly.