Heuristics for a periodic-review policy in a two-echelon inventory problem with seasonal demand

Abstract

This paper proposes heuristic methods to determine an ordering policy with a minimum inventory cost while satisfying the expected service level constraint in a divergent two-echelon inventory system. A periodic-review base-stock policy with an echelon stock concept was applied at all locations. Seasonal demand with a short cycle was considered at the level of the retailers. A two-phase heuristic method was then proposed to reduce the computational time required to determine a proper policy compared to the mixed-integer programming models. The first-phase algorithm was used to determine the initial ordering policy for each location, while the second-phase algorithm was used to determine the safety stock. Both algorithms could find solutions that are as good as the solutions from the mixed-integer programming models in testing instances but they substantially reduced the computational time. The solutions were tested with other instances having a 10,000-period horizon, where the average loss was within the expected service level and the number of periods with an unsatisfied expected service level was less than 1% of the planning horizon.