This paper presents a variant of a spare parts inventory model with cold stand-by redundancy on system level. Redundancy on system level implies that not all systems need to be operational in order to have the whole system operational. The cold stand-by feature implies that only the minimum required systems are operational. In order to determine a cost effective spare parts package such that in a cold stand-by redundancy situation a sufficient number of systems is operational for a specified period we extend the METRIC methodology. To compute the probability that the number of operating systems during the operational period is sufficient, we present both an exact, but time-consuming method, and a fast approximation method based on fitting distributions on the first two moments. This approximation method shows very small differences when compared to the exact method. Finally, we compare both methods to a simulation model in order to test the validity and impact of our modelling. © 2010 Elsevier Ltd. All rights reserved.