The research develops a series of challenging stochastic dynamic programming (DP) models that capture, for the first time, important aspects of the problem of managing retail networks and address a range of issues that managers have identified as important.
The research develops new approximate solution methods for these models using decomposition, policy improvement and Lagrangian relaxation to reduce the dimensionality of the problems.
The research achieves a number of notable firsts including the analysis of pre-emptive and reactive transshipment policies in a single model and the application of index theory to the problem of vendor managed inventory replenishment.
Analysis of the models developed reveals the considerable recurrent savings in the cost of managing multi-location inventory systems that can be obtained through the adoption of our methodology.
The methods proposed advocate decision making based on calibrations of the locations via low-dimensional indices and herald a radically new approach to multi-location inventory control.