Estimating the long-term cost to serve new customers in joint distribution

One of the most important concerns for logistics service providers is to identify the distribution cost to serve each new customer for pricing. Compared to the analysis through cost allocation on delivery routes, cost estimation possesses the advantage of robust costing rules but is a very challenging problem due to the complex collaborative mechanisms of distribution. Based on the activities leading to a distribution cost, we analyze the relationship between multiple geographic factors and cost, and then construct appropriate attributes for estimation. Combining a data selection approach and regression or artificial neural network techniques, a prediction scheme is proposed to build models, and an explicit continuous approximation model is suggested for efficient implementation. Computational experiments demonstrate the importance of the constructed attributes and the accuracy of the proposed cost estimation method. The impacts from cost stability and delivery frequency are examined to provide further explanation and support for practical implementation.