The effect of the thickness of ion exchange membranes has been investigated for electrodialysis (ED) and reverse electrodialysis (RED), both experimentally and through theoretical modeling. By developing a two-dimensional model based on Nernst-Planck theory, we theoretically find that reducing the membrane thickness benefits process performance only until a certain value, below which performance drops. For ED, an optimum thickness can be identified in the range of 10-20 mu m, while for RED the maximum power density is found for membranes that are three times as thick. Model calculations compare well with experimental data collected with a series of homogeneous membranes with the same chemical composition and a thickness in the range of 10-100 mu m. Our results show that the classical picture that membranes should be as thin as possible (as long as they remain pinhole-free and structurally stable) is insufficient, and must be replaced by a more accurate theoretical framework.