Near-optimal heuristics to set base stock levels in a two-echelon distribution network

We consider a continous review two-echelon distribution network with one central warehouse and multiple local stock points, each facing independent Poisson demand for one item. Demands are fulfilled from stock if possible and backordered otherwise. We assume base stock control with one-for-one replenishments and the goal is to minimize the inventory holding and backordering costs. Although this problem is widely studied, only enumerative precedures are known for the exact optimization. A number of heuristics exist, but they find solutions that are far from optimal in some cases (over 20% error on relistic problem instances). We propose a heuristic that is computationally efficient and finds solutions that are close to optimal: 0,2% error on average and less than 5.0% error at maximum on realistic problem instances in our computational experiment.