The impact of unknown demand parameters on (R,S)-inventory control performance

For a given inventory control system and a known stationary demand pattern, it is relatively easy to calculate the safety factors needed to satisfy predetermined criteria. It is more or less customary to use these standard safety factors in practical situations as well. In practice, however, demand parameters are unknown, causing additional variation. Consequently, the performance of this standard approach generally stays below the desired level. Hence, the safety factors should be increased. This general phenomenon is studied here in some detail for an (R,S)-inventory control system with normal demand, using the two best-known service criteria. Simple exponential smoothing is used to estimate the unknown demand parameters. Using large simulation runs, new (constant) safety factors are found that do satisfy the given service levels. They are not suitable for practical use, however: long series of past observations usually are not available, while stationary demand is rare. Therefore, time-varying safety factors are presented that seem to perform well for normal demand with unknown parameters.