Safety stock allocation problems including uncertainty in the supply-side have forced to assume a speci?c distribution of the lead time yet this is usually not reliable. From a guaranteed-service approach, a new model based on robust optimization in which no distribution is assumed is proposed. The robust model is a mixed-integer linear program. Then, the model is validated and applied to several complex real-case supply chains. Preliminary results are promising and show how the model could be a powerful tool for practitioners to manage risks from the supply-side in a robust yet efficient way.