This paper examines the performance of two different (s,Q) inventory models for spare parts in a production plant of confectionery in The Netherlands, a simple and an advanced model. The simple approach is more or less standard: undershoot of the reorder level is not taken into account, the normal distribution is used as the distribution of demand during lead-time. The advanced one takes undershoots into account, differentiates between zero and nonzero demands during lead-time, and utilizes the gamma distribution as the demand distribution. Both models are fed with parameters estimated by a procedure that forecasts demand sizes and time between demand occurrences separately (intermittent demand). The results show that the advanced approach yields service levels close to the desired one under many circumstances, while the simple approach is not consistent leading to much larger inventories if one wants to be sure that the service level is obtained for all spare parts.