Service Systems with Postponable Acceptance : Multi - project Selection and Assignment Problem

We study the dynamic order acceptance and resource assignment problem of a service system which provides heterogeneous services using heterogeneous servers. Unlike traditional acceptance and/or assignment problems, decision makers in our settings can strategically postpone acceptance decisions anticipating more profitable orders to arrive, with the risk of lost orders and low resource utilization. We formulate this problem using a dynamic and stochastic programming approach, find the structural properties of the optimal policy, propose approximate policies, and conduct extensive computational experiments to show the effectiveness of proposed policies. This problem is motivated by the multiproject selection and assignment problem of a software organization; however, it can be applied to general service systems where postponable acceptance decisions are allowed.