Supply Chain Management with Online Customer Selection

In regular supply chain management models, it is often the case that all the customers that have to be served are known in advance. We will consider a more realistic case by looking at models with online customer selection, where customers arrive sequentially, and after each arrival one has to immediately decide whether or not to accept the customer. This means that these selection decisions, which are done by what we call an online algorithm, have to be made with no information available about any future arrivals. We will consider two online algorithms for three different supply chain mod-els with online customer selection: the joint replenishment problem, the economic lot sizing problem and the traveling salesman problem. In addition to that, we perform computational experiments to evaluate the performance of these algorithms in several scenarios. These experiments show that the cost incurred by the online algorithm is very close to the optimal cost incurred when one would have full knowledge of the customer arrivals in advance.