Moving airlines towards a Competitor-Aware Revenue Management System

Thomas Fiig

Chief Scientist, Amadeus

This content is only available in this language.

Despite four decades of scientific advancement behind Revenue Management Systems (RMS) for airlines, these systems still rely on several simplifying assumptions in order to make the optimization problem manageable.

One such simplification, that is particularly unsatisfactory, is that these systems do not explicitly account for competition. Rather than modelling the true customer choice including the full set of offers available to the customer, they simplify the world view as monopolistic, meaning that customers are assumed to have only one offer (the airline’s offer).

However, this simplification is inadequate – given that airlines operate in a fiercely competitive environment, where customer demand will immediately be shifted based on competitors’ pricing decisions. This is not new knowledge. Airlines have over the years made several attempts to incorporate competitor price information into RMS: initially by overwriting RMS recommendations using rules-based systems, such as a maximum price gap vis-à-vis competitors, and more recently by applying Dynamic Pricing.

Yet, the fundamental scientific challenge remains of how to account explicitly for competition in RMS. For many years, the holy grail of Revenue Management has therefore been to construct a truly Competitor-Aware Revenue Management System (CARMS).

This was believed to be impossible, because it requires the airline using CARMS not only to collect competitors’ historical prices for all markets, departure dates, and days to departure; but also predict the future of competitors’ prices as well as price responses (reactions) at the same detailed granularity level. 

At this year’s AGIFORS conference, I presented, in collaboration with Mike Wittman and Clement Trescases, the concepts and feasibility of CARMS. We demonstrated that although this vision is futuristic – it is practically attainable given advancements in technology and available data. We were able to accurately predict future competitor prices for the selected test markets with a Mean Absolute Percentage Error (MAPE) of about 10%. This is similar to the accuracy at which airlines can predict their own prices. 

We find this quite remarkable. In addition, we implemented in a simulation environment a first version of CARMS achieving revenue benefits of 2% - 5%. Finally, we showed that Dynamic Pricing – that was originally invented as a corrective action – naturally integrates within the CARMS framework.

For this work our presentation was awarded as one of the best technical papers. We are excited to push the frontiers of RMS and about what the future holds for the airline industry in this regard. For more, please keep an eye open for our forthcoming paper on CARMS in the Journal of Revenue and Pricing Management.


Airline IT, Machine Learning, Artificial Intelligence

Do you want us to email you the latest insights?

Subscribe to updates