Towards sporadic demand stock management based on simulation with single reorder point estimation

The goal of this paper is to decide whether bootstrapping and/or linear regression are suitable to estimate an initial reorder point in sporadic demand stock management based on past stock movement simulation (PSMS) in combination with neighborhood search-oriented optimization. Thus, we randomly generate demand data including 20–80% zero demand periods and simulate continuous review, fixed order quantity inventory control policy (R, Q) for 4 different arrangements of PSMS combined with local search (LS) with a number of bootstrapping sampling runs ranging from 5 to 500. The original idea of LS is to underestimate order lead time demand using linear regression (LR), overestimate lead time demand with the help of bootstrapping (B) and search the generated interval using PSMS to return R, Q with the optimal trade-off between inventory costs and service level. The outputs gained from simulation experiments show that avoiding the generation of overestimated reorder point with B seems to be a more sensible choice as the consumption of computational time is significantly higher than in case of LR. On the other hand, using an LR based initial reorder point may require the exploration of neighborhood in both directions, while B rather enables a more efficient one-way search as it suffers from significantly less blindness caused by PSMS compensating underestimated order lead time demand with increased replenishment orders. Furthermore, estimating just one initial reorder point brings a better opportunity to control the consumption of computational time through assigning a certain amount of computational time to every change of the initial reorder point, as a time to evaluate a single R, Q combination via PSMS is relatively stable.