Coffee, tea, or …?: Markov decision process model for airline meal provisioning

This paper develops and analyzes a finite horizon Markov decision process model for the airline meal provisioning activity focusing explicitly on developing policies for determining and revising the number of meals to upload. Using one year of daily data from over 40 flights, the paper shows that the optimal policies can result in both improved customer service and significant dollar savings, especially in long haul flights. It also applies the model to derive an efficient frontier and investigate tradeoffs between having too few and too many meals on a flight.