Non-stationary stocking policies in service level agreements

Service levels agreements are commonly used in supply chains to commit a supplier to a certain performance target. A popular measure of performance is the item fill rate. For finite review horizons, the distribution of the item fill rate is studied by Thomas (2005), whose results are replicated in this thesis. In the extension, a similar replenishment model is studied with i.i.d demand, a single customer, no lead time and negligible ordering costs. Under these assumptions, the performance of a simple, practical and realistic non-stationary stocking policy is evaluated by Monte Carlo simulation and compared to that of a stationary base-stock policy. This policy turns out to outperform a base-stock policy based on meeting fill rate targets in the long run. However, when a high target needs to be met with high probability over a finite period of time, the base-stock policy does better and the original non-stationary policy requires adjustment. The main contribution of this work is to show that adopting a non-stationary stocking policy may save costs incurred by the supplier.