Optimal policies under risk for changing software systems based on customer satisfaction

Abstract The critical nature and impact of decisions regarding software-system changes demand a comprehensive framework for policy analysis. In this paper, we develop a Markov decision process model for the determination of jointly optimal warranty, maintenance and upgrade policies for software systems. The decision maker is the supplier, who needs to establish policies regarding the timing and nature of changes in the sytem. The state of the system is the customer satisfaction index, which is subject to declines of random magnitude, as a consequence of forgone warranty, maintenance and upgrade opportunities that arise randomly in any period. We consider a class of control-limit policies, defined by warranty, maintenance and upgrade thresholds. These thresholds divide the state space into four regions in which the supplier would, given the appropriate opportunity, do nothing, warrant only, warrant and maintain, or warrant, maintain and upgrade. We determine the optimal policies under different criterion functions, develop solution algorithms, provide computational examples, and identify a class of optimal or near-optimal policies of practical interest.