AIR CARGO REVENUE AND CAPACITY MANAGEMENT

The traditional air cargo supply chain is composed by the shippers, the freight forwarders and the airlines. The freight forwarders secure capacity with airlines in order to accommodate shippers' demand. They bid for capacity six to twelve months before the actual departure date of the aircraft, and confirm the needed capacity a few days before departure.\n\nWe address the freight forwarders' problem of confirming needed capacity based on balancing the costs of ordering too much capacity versus ordering too little. We use a Markov decision process to model the problem. We show the value function is convex in the state variables for lead times of one and two periods. We present the structure of the optimal policy and show it is stationary. In addition we present solutions to the case with subcontracting options and order due dates.\n\nWe also address the airlines' revenue management problem with respect to its cargo capacity available for free sale (after honoring committed capacity to freight forwarders), in particular the problems of (1) accepting/rejecting incoming bookings based on bid prices, and of (2) estimating the show-up rate (ratio of bookings handed in at departure over bookings on hand) with impact on overbooking.\n\nTo address the lumpiness of demand, we split the cargo into two categories: small cargo, composed of mail and small packages, and large cargo, composed of the bulk of commercial cargo. The small cargo is approximated with the passenger arrival, and we propose a new algorithm to solve the traditional probabilistic nonlinear problem from the passenger side. The large cargo is solved using a dynamic program, which is decomposed\nat the leg level using a fare-prorating scheme. The solution from our new approach is shown via simulation to be superior to two approaches currently used: the first come first serve, and the deterministic linear program.\n\nThe show-up rate is estimated using wavelets and we show that a discrete show-up rate is more suitable than the traditional Normal estimator used in practice. The new estimator results in considerable more potential revenue.