Impacts of revenue management on estimates of spilled passenger demand

Abstract

In the airline industry, spill refers to passenger demand turned away from a flight because demand has exceeded capacity. The accurate estimation of spill and the lost revenue it implies is an important parameter in airline fleet assignment models, where improved estimates lead to more profitable assignments. Previous models for spill estimation did not take into account the effects of passenger choice and airline revenue management. Since revenue management systems protect seats for later-arriving higher fare passengers, revenue management controls will influence the number of spilled passengers and their value because they will restrict availability to lower fare passengers even if seats on the aircraft are available. This thesis examines the effect of various revenue management systems and fare structures on spill, and, in turn, the marginal value of incremental capacity. The Passenger Origin Destination Simulator is used to simulate realistic passenger booking scenarios and to measure the value of spilled demand. A major finding of the research is that in less restricted fare structures and with traditional revenue management systems, increasing capacity on a flight leads to buy-down which can result in negative marginal revenues and therefore revenue losses. This behavior is contrary to conventional wisdom and is not considered in existing spill models. On the other hand, marginal revenues at low capacities are greater than would be predicted by first-choice-only spill models because some passengers will sell-up to higher fares to avoid spilling out. Additionally, because of passenger recapture between flights, adding capacity to one flight can lead to revenue losses on another. Therefore, the marginal value of incremental capacity is not always positive. Negative marginal revenues and associated revenue losses with increasing capacity can at least be partially mitigated by using more advanced revenue management forecasting and optimization algorithms which take into account passenger willingness to pay. The thesis also develops a heuristic analytical method for estimating spill costs which takes into account the effects of passenger sell-up, where previous models tend to underestimate the spill cost by only modeling passengers' first choices. The heuristic demonstrates improved estimates of passenger spill: in particular, in restricted fare structures and for moderate amounts of spill, the model exhibits approximate relative errors on the order of 5%, a factor of two improvement over previous models.