In the rail industry, the important design parameters of rubber suspension systems are currently solely based on the loading part of the loading/unloading history, e.g. the load-deflection characteristics and fatigue requirement. Different energy levels and stress values are created for an identical load value during loading and unloading cycles in rubber-like materials. Hence, the performance of a rubber suspension can be substantially different during loading and unloading, which can lead to unexpected effects. An engineering approach is proposed to account for this so-called Mullins effect. Existing elastomeric models, widely used in rail vehicle design, can be modified to account for the unloading using this methodology. A typical rubber-to-metal bonded component, which is used in rail suspension systems, is selected for a verification study. It is shown that the predictions from the new approach are consistent with the results of the whole load/deflection history obtained in a laboratory experiment. In addition, if the unloading characteristics are not considered, results obtained from stress calculations can have a 20% margin of error. The proposed approach should be further verified using other types of rubber suspension systems.