Despite widespread use for more than two decades, the SERS phenomenon has defied accurate physical and chemical explanation. The relative contributions from electronic and chemical mechanisms are difficult to quantify and are often not reproduced under nominally similar experimental conditions. This work has used electromagnetic modelling to predict the Raman enhancement expected from three configurations: metal nanoparticles, structured metal surfaces, and sharp metal tips interacting with metal surfaces. In each case, parameters such as artefact size, artefact separation and incident radiation wavelength have been varied and the resulting electromagnetic field modelled. This has yielded an electromagnetic description of these configurations with predictions of the maximum expected Raman enhancement, and hence a prediction of the optimum substrate configuration for the SERS process. When combined with experimental observations of the dependence of Raman enhancement with changing ionic strength, the modelling results have allowed a novel estimate of the size of the chemical enhancement mechanism to be produced.