An Introduction to the Airline Industry and Air Traveler Behavior

Click to enlarge

After the U.S. airline industry was deregulated in 1978, airlines were thrown into a new environment, one in which they could compete on price and determine where, when, and with which planes they wanted to fly. Sales and marketing gained importance as new programs, such as frequent flyer programs and corporate volume agreements, were developed to attract and retain valuable customers. In the decade that followed, operations research played a critical role in an airline’s individual success as the industry raced to develop models in areas such as scheduling, pricing, revenue management, marketing, sales, and operations. Many of these first-generation decision support systems developed shortly after deregulation are still in use today.

These first-generation decision support systems were built at a time in which computing power was smaller. It was not possible to model air travel demand using individual booking records, particularly for those systems that updated forecasts on a daily basis. Computational resources simply could not handle processing millions of booking transaction records within a 24-hour period. As a result, first-generation decision support systems, such as revenue management systems, would forecast product-level demands (or even cabin-level demands, as is the case of some no-show demand forecasts). Although these demand forecasts were limited in that they could not capture how individual customers were making trade-offs among price, schedule, level of service, carriers, and other product-characteristics, these first generation systems nonetheless worked very well. Bob Crandall, former CEO of American Airlines, notes that American’s first revenue management system generated more than $500 million in annual revenues for the airline [3].

One of the reasons why these first-generation revenue management systems worked well is because many of the assumptions embedded in the underlying demand forecasting models reflected market conditions in the 1980’s and 1990’s. During this time period, the majority of customers made reservations by calling travel agents or an airline’s reservation system. It was time consuming for customers to call multiple agents and airlines to search for information on the different products and prices that were available. As a result, it was fairly straightforward for airlines to develop products that segmented leisure and business customers. The most common product attributes used to segment customers included advance purchase restrictions, the Saturday night stay requirement, and ticketing exchange and refund provisions. Also during this time period, legacy network carriers dominated the market. Although low cost carriers such as Southwest were gaining momentum, they were not yet considered as a major competitive threat.

The threat became clear during the 2000’s, as a dramatic shift in how customers made airline reservations occurred. Bill Brunger, retired senior vice president of networks at Continental Airlines and colleague Sheri Perelli report that from 1998 to 2005, domestic leisure flight tickets sold through the internet increased from 1% to 35% [1]. The internet made it much easier for customers to search for price information. The appearance of online travel agency sites Travelocity and Expedia in 1996 further facilitated the comparison of prices across multiple carriers. Today, PhoCusWright reports that more than 60% of online leisure travel customers purchase the lowest fare they can find [6].

Legacy carriers also faced pricing pressure from low cost carriers that introduced new pricing models. The majority of low cost carriers use one-way pricing. Carriers who use one-way pricing cannot segment business and leisure travelers based on a Saturday night stay. Some low cost carriers, most notably JetBlue, moved away from the concept of advance purchase restrictions. These changes resulted in the blending of business and leisure customers across different products. Business customers who were once loyal to legacy carriers were now able to purchase lower-priced products close to the day of departure from low cost carriers.

Faced with the explosive growth of online booking channels, increased market penetration of low-cost carriers, unprecedented fuel costs, continued security threats post 9/11, health outbreaks (SARS, H1N1), economic recessions, and the global financial crisis [4,7], the seven largest U.S. network carriers collectively lost $35.1 billion in the first decade of the 21st century [8]. This decade also saw customer satisfaction levels plummet, as passengers faced reduced flight schedules, higher load factors, and long security lines [2]. From 2001 to 2005, four out of the seven largest network carriers filed for bankruptcy (e.g., Delta, Northwest, United, and US Airways), and from 2005 to 2011 eight major US carriers went through a merger or acquisition (e.g., America West and US Airways in 2005; Delta and Northwest in 2008; Continental and United in 2010; Southwest and AirTran in 2011).

Some of the revenue loss seen among legacy carriers in the 2000’s is because the first-generation decision support systems did not function as well under these new market conditions. Armed with increased computational power and the ability to process millions of individual passenger-level booking records, airlines began to investigate how they could build the next-generation of decision support systems that use discrete choice models to predict demand. Unlike time-series and other forecasting methods based on historical averages, discrete choice models predict demand as the probability an individual will select a particular product; these probabilities are based on how similar or different available products are, and how much customers are willing to pay for different product attributes, such as nonstop flights, preferred carriers, preferred departure times, etc. Demand predictions based on discrete choice models more accurately incorporate air travelers’ behavior, by capturing how travelers’ make trade-offs among product attributes.  As Scott Nason, former VP of Revenue Management at American Airlines stated in a presentation at the 2011 annual INFORMS meeting, "I can’t tell you how soon, but some of you will see very good choice models adopted as the basis of RM demand forecasts during your lifetime" [5].

Dr. Garrow's research group is working on several projects that will enhance our understanding of air travel behavior. Through this improved understanding of customer behavior, they will be able to enhance revenue management, scheduling, and operational decision support systems. In turn, this will enable airlines to better match supply and demand, improve operational performance, and provide better service and products to customers.



  1. Brunger, W.G. and Perelli, S. 2008. The impact of the internet on airline fares: Customer perspectives on the transition to internet distribution. Journal of Revenue and Pricing Management, 8(2/3): 187-199.
  2. Carpenter, D. 2008. Airline Customer Satisfaction Nosedives: “Dismal”. The Huffington Post, May 20, 2008. Accessed on July 22, 2008.
  3. Cross, R. G. 1997. Revenue Management: Hard-Core Tactics for Market Domination. New York: Broadway Books.
  4. CNN. 2010. Dow Jones Industrial Average. Accessed on June 23, 2010.
  5. Nason, S. (2011). The airlines' evolving revenue models.  Presentation at the INFORMS annual conference in Charlotte, NC. (transcript forthcoming in the INFORMS Aviation Application and Revenue Management and Pricing newsletters).
  6. PhoCusWright (2004). “The PhoCusWright Consumer Travel Trends Survey."
  7. Seeking Alpha. 2009. 2008 Dow Jones Performance: Third Worst on Record. Accessed on June 23, 2010.
  8. US Department of Transportation Research Innovative Technology Administration Bureau of Transportation Statistics. 2010 Air Carrier Financial: Schedule P-12 (from 2000-2009). Downloaded on May 16, 2010.