Purpose
– Ensuring the sufficient service level is essential for critical materials in industrial maintenance. This study aims to evaluate the use of statistically imperfect data in a stochastic simulation-based inventory optimization where items' failure characteristics are derived from historical consumption data, which represents a real-life situation in the implementation of such an optimization model.
Design/methodology/approach
– The risks of undesired shortages were evaluated through a service-level sensitivity analysis. The service levels were simulated within the error of margin of the key input variables by using StockOptim optimization software and real data from a Finnish steel mill. A random sample of 100 inventory items was selected.
Findings
– Service-level sensitivity is item specific, but, for many items, statistical imprecision in the input data causes significant uncertainty in the service level. On the other hand, some items seem to be more resistant to variations in the input data than others.
Research limitations/implications
– The case approach, with one simulation model, limits the generalization of the results. The possibility that the simulation model is not totally realistic exists, due to the model's normality assumptions.
Practical implications
– Margin of error in input data estimation causes a significant risk of not achieving the required service level. It is proposed that managers work to improve the preciseness of the data, while the sensitivity analysis against statistical uncertainty, and a correction mechanism if necessary, should be integrated into optimization models.
Originality/value
– The output limitations in the optimization, i.e. service level, are typically stated precisely, but the capabilities of the input data have not been addressed adequately. This study provides valuable insights into ensuring the availability of critical materials.