Paper short abstract:

Prices for flights are highly dynamic. This is the outcome of a technologically complex marketplace, with the so-called airlines’ revenue management systems at its core. The paper studies the work of the engineers who built revenue management systems for airlines.

Paper long abstract:

The paper studies how the engineers who built revenue management systems for airlines work and think, based on 44 in-depth interviews, participation in industry conferences, and an analysis of the technical literature they produced. Revenue management systems emerged in the 1980s United States, as a way for airlines to price discriminate (i.e. to offer the same seat in the same class of service of the same flight at different prices) at industrial scale, and this required the automation of pricing: large datasets for systematic demand forecasting, algorithms for revenue maximization, electronic commerce infrastructure for real-time management of availability. This paper shows: (1) Market design - the field in economics - is nowhere to be found in that story. Ways of thinking, models and tools to build revenue management systems came from operations research, statistics and probability, and computer science. (2) Econometrics took center stage since the 2000s, for the estimation of customers' willingness-to-pay. The paper details the two approaches implemented by airlines: consumer choice modeling using the multinomial logit model, or the estimation of elasticity curves with machine learning techniques. (3) The accuracy of the econometric estimation matters, yet for the marketplace engineers in the airline industry two other dimensions are crucial: data collection and the techno-economics of the marketplace itself. To estimate willingness-to-pay, airlines can use their own historical booking data, or they can use online shopping data gathered on their own website. The later requires moving around powerful middlemen: the global distribution systems.