Efficient heuristics for inventory placement in acyclic networks

The strategic safety stock placement problem is cast as a constrained separable concave minimization problem. Some network-specific algorithms do exist in the literature, but their utility is limited to small, sparse, and special supply chain network structures. In this paper, we present two efficient, easy-to-implement heuristic algorithms for placing strategic safety stock in general acyclic supply chain networks. The computational study demonstrates that the algorithms are able to obtain near-optimal (within 4% and 7% in average) solutions efficiently by solving a finite series of LPs (7%) or fixed-sized MIPs (4%). More importantly, their performance in terms of solution quality is nearly independent of the network size (for simulated instances with up to 100 stages). For general acyclic supply chain networks with 8000 nodes and 32,000 arcs, the LP-based algorithm typically finds solutions in under 5 minutes.