In spare parts industries, firms are dealing with a situation which is more and more uncertain due to the supply chain structure and variable demands. This paper presents a Bayesian approach to forecast demand and subsequently determine the appropriate parameter S of an (S - 1; S) inventory system for controlling plant spare parts. We apply the Bayesian approach in an innovative way to specify the initial prior distributions of the failure rates, using the initial estimates and the failure history of similar items. According to the proposed method, a lower base stock than the one currently used is sufficient to achieve the desired service level.