AGIFORS RM Study Group Research Wrap

The Revenue Management Study Group met two weeks ago in Seattle for their annual meeting, hosted by Alaska and Hawaiian Airlines. Many of the presentations carried an updated version of the same themes discussed at AGIFORS each year. How do we train the machine to produce the best possible outcomes, how often should a human intervene, and after all is said and done, how do we know which decision was right?

A subset of papers this year discussed the fact that what a revenue management system observes is not just noisy, but incomplete. Houman Goudarzi, formerly IATA’s Head of Data, and founder of ZYTLYN Technologies, argued that a finalized booking has already filtered through price, availability, and brand choice before an airline ever sees it, so true market demand has to be reconstructed rather than derived from the booking record. Thomas Fiig of Amadeus described a related but distinct new distortion introduced by AI shopping agents, which can hold a booking decision in reserve rather than act on it immediately; this is something that the RMS has not yet been built to consider. Jakub Figura and Maciej Pawełczyk of Fetcherr pointed at a third gap in the same signal: traditionally, an RMS does not consider events, despite the fact that they can have a huge impact on total revenue over the life of the flight.

A second subset of papers took on the question of attribution: how to separate what a person (or strategy) contributed to an outcome from everything that would have happened regardless. Alex Winston of Southwest Airlines built a way to isolate RM strategy’s contribution to revenue from supply, demand, and seasonality influencers. Alex Matson of telos proposed a way to evaluate whether a single override decision was reasoned well, independent of whether it yielded positive or negative results. The topic struck a cord with the group as it was followed by significant audience questions and engagement. Hermon Kweon of ADC and Nicolás Ferrari-Ortiz of LATAM Airlines addressed a related version of the same problem from the experimental side, designing a way to get a cleaner causal read on pricing tests that can only be randomized at the route or flight level.

At the end of the meeting, one presentation is voted “Best Presentation”. This year it was Kabir Samlal, presenting “Willingness-to-Pay: A Modern Causal Inference Approach”, on behalf of the ADC contingent.

The Georgia Tech group is notable for a different reason. Jeff Newman presented the lab’s ongoing work with Laurie Garrow and Alan Walker on building calibrated research networks. Antonio Ramirez, also working with Garrow, applied that same infrastructure to a price sensitivity question. And Gabriel Cedraz Diniz, a PhD student under Garrow’s advisement and a former Ken Wang Scholar, is credited by name in Newman’s paper for helping build the very networks Newman presented, before presenting separate research of his own in the same room. What is so cool about this is that it’s the same arrangement Daniel Fry described from his own first AGIFORS meeting in Shanghai in 2015. In front of the same type of crowd, he presented his MIT master’s thesis on demand-driven dispatch under Peter Belobaba’s advisement. A graduate student’s early research being tested inside this unique space that was built by others who came before them. Perhaps they may end up working alongside some of the very same people to whom they are presenting. Fry made that point directly: AGIFORS is not only where the papers get presented, but where the field’s next generation gets absorbed into a research community that already exists.

Few of this year’s papers cite each other, and most were written independently. What repeats is not any single finding, but the common thread of curiosity and an innate drive to do something better. It is a pattern that goes back as far as the archive does. The unique and addictive field that is aviation continues to produce researchers who converge on problems independently, then year after year converge to learn from and be inspired by each other. This is why AGIFORS remains relevant and held with high regard, sixty-five years and counting.

The next stop on the calendar is the 66th AGIFORS Annual Symposium, hosted by Turkish Airlines in Istanbul, Türkiye, October 25 to 29, 2026. For registration and further detail, see AGIFORS’s website.