Objective During the last century, running shoes have been subject to drastic changes with incremental however improvements as to injury prevention. This may be, among others, due to the limited insight that experimental methodologies can provide on their 3D in situ response. The objective of this study was to demonstrate the effectiveness of finite element (FE) modelling techniques, in optimizing a midsole system as to the provided cushioning capacity. Methods A commercial running shoe was scanned by means of micro computed tomography and its gel-based midsole, reverse-engineered to a 200 μm accuracy. The resulting 3D model was subjected to biorealistic loading and boundary conditions, in terms of time-varying plantar pressure distribution and shoe-ground contact constraints. The mesh grid of the FE model was verified as to its conceptual soundness and validated against velocity-driven impact tests. Nonlinear material properties were assigned to all entities and the model subjected to a dynamic FE analysis. An optimization function (based on energy absorption criteria) was employed to determine the optimum gel volume and position, as to accommodate sequential cushioning in the rear-, mid-, and forefoot, of runner during stance phase. Results The in situ developing stress fields suggest that the shock dissipating properties of the midsole could be significantly improved. Altering the position of the gel pads and varying their volume led to different midsole responses that could be tuned more efficiently to the specific strike and pronation pattern. Conclusions The results suggest that midsole design can be significantly improved through biorealistic FE modelling, thus providing a new platform for the conceptual redesign and/or optimization of modern footwear.