Revenue management—the process of maximizing revenues from perishable and fixed capacities—is important to the airline industry because of intense competition and narrow profit margins and dates back to 1978 since the deregulation. Basically, revenue management systems can be classified into two streams: Quantity based RM (deciding on quantity decisions while assuming prices are known and given) and Price based RM (deciding on price while assuming quantities are known and given). Whenever quantity decisions are considered, substitution is a valuable option that can mitigate the uncertainty by aggregating resources. We focus on facilitating quantity decisions while accounting for the inherent flexibility associated with the product.



First, in the passenger segment, we consider an overbooking problem faced by an airline when the airline operates a single-leg, two-cabin class flight. While considering the variability in demand, ticket prices, denied boarding costs and no-show rates for both the classes, we formulate two models: Grid search based simulation and decision tree-based analytical model. Numerical results show that the results from both models are in tandem with each other. On analyzing, we found that upgrades result in increased denied boarding, which calls for upgrade-aware overbooking policy, i.e., imposing service-level constraints to maintain airlines' reputation.


Though RM systems become prevalent from 1978, more sophisticated models emerged only on passenger revenue management. Research on air cargo revenue management (ACRM) started in 1992 with the works of Oakley et al. (1992). Yet, research on ACRM is still in its infancy stage and it is so because of the higher complexities: Multi-dimensional, uncertain, and continuous capacities, variable tendering, and allotments. Besides these complexities, ACRM has one significant advantage, which is routing flexibility, and it can be exploited when we consider a collection of flights. We consider an airline operating sequential flights between two locations and formulate the problem as dynamic programming to model the request arrivals where we consider both the volume and weight dimensions, and a linear programming formulation to make the allocation decisions during departure.