"Revenue Management in the Travel Industry" In

Home , Revenue management, Yield management

REVENUE MANAGEMENT IN THE us straight towards bankruptcy....Wheredid TRAVEL INDUSTRY we go wrong?...we didn’t get our hands around the Yield Management—Don Burr, former CEO of People Express Airlines [5–8] WARREN LIEBERMAN If I were in charge of the antitrust division, Veritec Solutions Belmont, IwoulddamnwellseeifIcouldbuildacase California that yield management is predatory pricing—Alfred Kahn, former Chairman of the Civil Aeronautics Board under President Yield Management1 could easily turn out to be Carter and known as the father of airline the most important pricing breakthrough of deregulation [9] the 20th Century—Professor James Makens, Yield management as it applies to airlines is Wake Forest University [1] the control and management of reservations ...look to yield management at the airlines as inventory in way that increases (maximizes, if the model for future pricing in fields as diverse possible)...profitability, given the flight as health care, telecommunications, consumer schedule and fare structure.[emphasis financial services, insurance, hotels and added] [4] government services. Yield management Revenue Management is the art and science of techniques will rewrite the predicting real-time customer demand at the supply-and-demand equations that determine micromarket level and optimizing the price service availability and price [2] and availability of products [8]. [Yield Management] promises to spearhead a revolution in pricing strategies during this decade [3] INTRODUCTION Yield Management is the single most important technical development in The methods and information used by firms transportation management since we entered in the travel industry that govern the range the era of airline deregulation in 1979—Bob of products they offer for sale, the number of Crandall, former CEO and President of AMR sales they make, and the prices they charge 2 and American Airlines [4] have undergone dramatic change in the past We estimate that yield management has 30 years. As noted in the above quotes, these generated $1.4 billion inincrementalrevenue changes have revolutionized the travel indus- in the last three year years [for American try and are penetrating other industries as Airlines]...We expect yield management to generate at least $500 million annually for the well. foreseeable future—Bob Crandall [4] Whereas some of these decisions were ...our competitors used...Yield Management once based on intelligent ‘‘rules-of-thumb,’’ in every one of our markets, and they pushed they now draw on sophisticated mathematical models. In some cases, the advances in decision support tools in combination 1As the practice of yield management spread with advances in electronic distribution beyond the airline industry, the term changed to and management of inventory have enabled revenue management. Except for providing back- travel firms to offer and manage products ground on the term’s evolution, and in keeping that would not otherwise have been possible. with the title of this article, we generally refer Further, the financial gains made possible to the discipline of forecasting customer demand by these developments, initially known as and available supply in real-time and using that yield management (YM) and later as revenue information to control product availability and dynamically set prices so as to maximize profits, as management (RM), have been extraordinary. revenue management. This article provides an historical overv- 2This quote has become well known and is often iew of why and how this transformation cited; see, for example, Refs 8, 41. occurred. Much of the discussion on the

Wiley Encyclopedia of Operations Research and Management Science,editedbyJamesJ.Cochran Copyright  2010 John Wiley & Sons, Inc. 1 2REVENUEMANAGEMENTINTHETRAVELINDUSTRY early stages of yield management is oriented tickets were fully refundable.3 For flights towards the airline industry, as it provided in high demand, more tickets would be sold the initial setting for the early innovations— than there were seats on the plane. Known innovations that soon spread to other as overbooking, this practice was routinely industries. As yield management evolved carried out by the airlines but it did not into other industries and became known as attract widespread attention until Allegheny revenue management,thedisciplinebroad- Airlines did not allow Ralph Nader to board ened from optimizing capacity controls (e.g., his flight in 1972. Nader sued Allegheny the maximum number of sales that were and the lawsuit eventually went to the US allowed at previously established prices) Supreme Court in 1976 [10]. Overbooking to include pricing optimization (e.g., deter- decisions based on mathematical models that mining what prices should be). The latter seek to maximize financial returns subject portions of this article are oriented around to customer service constraints constitute the expanding practice of the discipline and acoreelementofmanyairlinerevenue its adoption by other segments of the travel management programs. Indeed, the first industry. phase of revenue management essentially Often misunderstood, yield or revenue consisted solely of overbooking. management has now attained a promi- Until 2000 or perhaps even later, most air- nent position among the strategies and line yield management departments did not tactics travel companies use to obtain a set prices, but rather focused on controlling competitive advantage over their rivals. the availability of reservation or seat inven- As might be gleaned from the quotes tory at different prices with the objective of above, it is a relatively new way of doing maximizing net revenue. The actual prices business and many believe it to be an at which tickets could be sold were deter- extraordinarily powerful technology. CEOs mined by Pricing Departments. Although of companies that invested in it have extolled Yield Management had no responsibility for its virtues as well as its importance for setting prices, it determined which fares and how many tickets at each fare could be pur- achieving better financial returns. This chased on each flight. In part, this resulted article provides an overview of what revenue from the organizational structure of most management is, how it evolved in the airlines, whereby the Pricing Department travel industry, and why it attracts strong and the Inventory Control or Yield Manage- interest. ment Department, were separate.4 Indeed, in 1985 when American Airlines launched its Ultimate Super Savers fares—discounts of FROM YIELD TO REVENUE: A HISTORICAL up to 70% with restrictions such as 30-days OVERVIEW advance purchase, 7-day minimum stay, and only partially refundable—enabling It is frequently suggested that the science, if the airlines to compete for passengers in not the practice of yield management orig- radically new ways, its yield management inated in the US airline industry in the department was not made aware of the early 1980s, shortly after it was deregulated new fares until they were announced to the in 1978. In fact, the early stages of yield public. This organizational decision resulted management date back quite a bit further, probably to the 1960s. Until 1985, however, the practice attracted limited attention, was 3Today, the frequency with which passengers fail narrow in scope, and remained within the to show up for their flights is much less, as many airline industry. of the tickets sold are nonrefundable. For many years, it would not be unusual 4During the late 1990s, some airlines began to for 15% or more of those who purchased restructure their organizations, consolidating pric- airline tickets to fail to show up for their ing and inventory functions or having these staff flights. In part, this occurred because unused work in a more coordinated manner. REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 3 in significant revenue dilution. The fares single product could be sold, simultaneously, stimulated so much demand, so quickly, at multiple prices. even for flights 6–12 months in the future, Yield had a specific meaning in the airline that American’s yield managers could not industry, being related to revenue in two react fast enough to reset fare and booking ways: controls on all flights to limit the sale of Ultimate Super Saver fares to appropriate 1. Revenue per available seat-mile or levels. Seats that could have been sold revenue per available seat-kilometer to business travelers who were willing to (RASM and RASK) provided a mea- purchase tickets at higher fares, but would sure of revenue per unit of capacity. not do so until closer to the date of their 2. Revenue per passenger-mile or rev- flight’s departure, were instead occupied by enue per passenger-kilometer (RPM those who purchased fares for up to 70% or RPK) offered a normalized measure less. Many flights generated less revenue of fares paid. and profit than they otherwise might have. Prior to the late 1980s and the adoption Because the two types of reservation con- of revenue management by travel firms other trols noted above had clear and direct impacts than airlines, yield management decisions on both measures of yield, the term yield were essentially limited to two decisions: management seemed apt. In the late 1980s and early 1990s, as •thenumberofreservationstoacceptfor the practice of yield management spread to aflight; hotels, car rental firms, passenger railroads, •thefares,previouslyestablishedby cruise lines, and other segments of the travel the pricing department, that should industry, the phrase ‘‘yield management’’ was be offered for purchase to potential often viewed as airline jargon. Other compa- customers for a flight at various times nies typically focused on revenue and profit, prior to departure (the actual setting not yield. The term became increasingly con- of fares was thus done outside of yield troversial. Even as they embraced its con- management; yield management only cepts, many sought a different label for the controlled which of the fares were practice of revenue maximization through available for purchase). inventory allocation controls. The search for a new label may well have As we will soon discuss, one of the most been accelerated by the airline industry’s innovative as well as revenue enhancing ele- financial difficulties in 1990, 1991, and 1992. ments of airline yield management programs Table 1 highlights the financial difficulties was that by introducing purchase restrictions experienced by the largest seven US airlines on airline tickets—often termed fences—a during these years. With some exceptions,

Table 1. Net Earnings of the Seven Largest US Airlines, 1990–1992 ($000,000) Airline 1990 1991 1992 Total American Airlines (76.8) (165) (735) (976.8) Continental Airlines (273.8)a (340.9) (229.4) (844.1) Delta Air Lines (259.4)b (239.5) (564.8) (1,063.7) Northwest Airlines — (310) (386.2) (696.2) TWA Airlines (162.3)c 48.2 (317.7) (431.8) United Airlines 95.8 (494.1) (956.8) (1,355.1) USAir (427.2) (284) (1,223.4) (1,934.6) Total (1,103.7) (1,785.3) (4,413.3) (7,302.3) aOperating loss. bFor the six months ending December 31, 1990. cOperating loss. 4REVENUEMANAGEMENTINTHETRAVELINDUSTRY the airlines generally lost at least several support for relabeling revenue management hundred million dollars each year and some- as revenue and pricing optimization (RPO). times much more than that [11,12]. The seven Firms in virtually every segment of the airlines combined for a loss of more than $7 travel industry now possess revenue man- billion during these three years. agement programs. And each year, these Revenue management, anamefarmore programs improve. But let us not get ahead of industry neutral, began to gain favor. Refer- ourselves. Every story has its beginning and ences to revenue enhancement and revenue this is certainly true of revenue management. management replacing yield management To tell it, we begin with the airline industry. can be found as early as 1990 and began to proliferate thereafter [13–14]. YIELD MANAGEMENT BEGINS For many, the desire to ‘‘rebrand’’ yield management dovetailed nicely with the The total number of reservations that can desire to broaden its scope to include a wider be accepted for a flight is generally greater array of reservation inventory controls than than the number of seats on a plane. It is used by airlines, explicit recognition of vari- known as the overbooking level, although able costs, and price optimization (that is, the there was a time, probably short-lived, when determination of the best prices to offer). For airline staff preferred the term ‘‘revenue coor- example, cruise ships might offer 8–12 cabin dination’’ due to the negative connotations of categories; cruise line yield management overbooking [15]. Airlines generally overbook inventory controls explicitly considered the aflightinanticipationofreservationsbeing financial impacts of alternative upgrading cancelled, passengers with reservations who (and upselling) policies when taking reser- fail to show up for their flights (no-shows), vations, such as allowing some customers and other reasons. While overbooking has to make a reservation for a less expensive been a standard airline practice for over 50 cabin category but guaranteeing them years, the science of overbooking took a dra- accommodation in a more expensive cabin matic step forward in the mid-1960s. In that category. Yield management also led cruise respect, this might well mark the beginning lines to reevaluate and modify the price dif- of yield management in the travel industry.5 ferentials between cabin categories, as supply As noted by Rothstein, ‘‘[w]hen our Amer- and demand imbalances at the cabin category ican Airlines O[perations] R[esearch] Group level were analyzed with greater precision. initiated its research in 1964, the scant lit- The wider array of variable costs associated erature on overbooking yielded no models with hotel stays led hotel yield management or methods that had been implemented.’’ A efforts to consider profit, rather than revenue. year or two later, however, a newly devel- Yield management thus expanded to include oped overbooking program was implemented practices beyond those adopted by airlines. at American, dramatically improving the air- By 1993, revenue management was well line’s ability to forecast oversales (customers on its way to replacing yield management. with tickets that show up for a flight but Presentations at multiindustry conferences for whom no seats are available) and set on this topic included sessions in which overbooking levels that allowed the airline to speakers differentiated revenue manage- achieve higher load factors on high demand ment from yield management. Perhaps flights, without increasing the number of pas- as testament to the strength of this con- sengers who were denied boarding [15]. version, the annual International Air While working in the airline industry in Transport Association (IATA) Conference the 1980s, it was the author’s experience that on Yield Management, initiated in 1988, was rebranded as the IATA Conference on Revenue Management in 1993; in 2001 it 5Our decision to use the developments in the 1960s experienced further rebranding, becoming as the starting point for yield management reflects the IATA Revenue Management and Pricing our emphasis on the science, not simply the prac- Conference. There now seems to be growing tice, of the discipline. REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 5 the overbooking models used by some airlines allowed some of them to increase their load remained relatively rudimentary while those factors on high demand flights by as much as of others had evolved considerably since the 10 percentage points (e.g., from 85%–95%), 1960s. At American Airlines, the overbook- and possibly more. An airline with 750 flights ing model developed by Rothstein and his per day, an average fare of $175 per passen- colleagues in the operations research group ger per flight, average flight capacity of 120 was replaced in 1976 by one which recog- seats and an annual load factor of 61% would nized the revenues and costs associated with have annual revenues of approximately $3.5 overbooking levels and attempted to maxi- billion. If 10–15% of its flights sold out, sell- mize profitability. This model was further ing just one additional seat on these flights enhanced in 1987 [4]. increases annual revenue by approximately The science of overbooking involves as $4.8–7.2 million. Increasing the load factor many as five forecasts: of these flights by 10 percentage points could mean an annual increase of $57–86 million. 1. the probability of reservation cancel- Putting this into context, in 1981 and 1982, lations prior to departure; on passenger revenues of approximately $3.4 2. the probability of passengers with billion, American Airlines had operating reservations failing to show up for income of $44 million in 1981 and lost $18 their flight; million in 1982 [16]. To say that an airline’s revenue management capabilities, in terms 3. incremental demand when cancella- of the accuracy of its overbooking levels could tions occur before departure; be the difference between an annual loss or 4. the cost of oversales (this may not be profit, is not much of a stretch. linear for a flight and also, it varies While the substantial financial benefits to based on a variety of flight charac- the airlines from overbooking are clear, less teristics and the ability of the airline visible is the consumer benefit, which can also to retain the passenger on a different be substantial. When done well, overbooking flight that it operates); allows more consumers the ability to pur- 5. the incremental revenue obtained chase and use their product of choice. When from additional reservations. advance notice is given that a flight, hotel, or other service is sold out, but actualized For operational reasons, it may be desir- demand turns out to be less than capacity, able to limit the number of oversales on a some consumers will have been deprived of flight to a value less than what would be their ability to use that service and will have financially optimal. Such constraints have had to choose something less desirable. They been incorporated into overbooking models. will have been inconvenienced unnecessarily. Forecasting incremental demand (addi- Without a doubt, airline overbooking tional demand up to a flight’s departure) would not yield nearly the benefits it does allows implementing better overbooking lev- had the airlines not adopted a business els prior to a flight’s departure. The overbook- process they initially heavily resisted. On ing levels for an optimally managed flight heavily booked flights, it is relatively common will generally decrease over time; that is, as for passengers to inquire whether volunteers the number of days before a flight’s depar- are needed to give up their seat on the flight ture decreases, so too will its overbooking in exchange for some form of compensation level. The more likely it is that incremental and a seat on a later flight. Many show demand will back-fill cancellations, the lower clear disappointment when the airline’s Gate the levels of pre-departure overbooking. If all Agent responds that all passengers with goes well, the number of reservations hold- reservations can be accommodated on the ing at departure will be equal to the desired flight and no such volunteers are needed. departure-time booking level. This was not always the case. Prior to The financial benefits of overbooking 1978, not only did airlines not seek volun- were substantial for airlines, as it probably teers, airline executives did not believe that 6REVENUEMANAGEMENTINTHETRAVELINDUSTRY asking for volunteers would work when more YIELD MANAGEMENT EXPANDS passengers with reservations showed up for aflightthantherewereseatsontheplane. Prior to the mid-1970s, airline fares Each airline developed its own rules and pro- were highly regulated. Restriction-based cedures for determining who would fly and discounted fares, so familiar today, were not who would not. available. The discounted fares that were In 1966 and 1967, Professor Julian Simon available tended to be oriented to specific contacted airlines with an idea for how they groups such as children, stand-by passen- could obtain volunteers in exchange for an gers, senior citizens, and other class-based agreed upon level of compensation. None of categories. The availabilities of these fares the airlines thought much of his proposal, were based solely on eligibility. some even being derisive in their replies to When the US airline industry was him. Rather than let the concept die, he per- deregulated in 1978, the proponents and the sisted in his efforts to obtain support for detractors of deregulation envisioned major it [17]. More than 10 years later, his ideas industry changes. It is safe to say, however, finally took hold when economist Alfred Kahn that of the changes that were made possible became head of the Civil Aeronautics Board. by deregulation, the fare-related impacts of The CAB mandated that airlines adopt a plan revenue management were far from anyone’s for obtaining voluntary oversales [18]. expectations. Airlines varied in the types of compensa- In 1977, just prior to deregulation, Amer- tion they offered, some giving travel vouchers ican Airlines introduced discounted fares for use on future trips while others gave known as Super Savers on transcontinental free tickets for a future flight.6 The success flights and then throughout their system in of each program varied, but overall, the 1978. These fares were developed in response volunteer program enabled airlines to try to an alarming development that began dur- more aggressively to fill each seat. Between ing the summer of 1976. Businesses had 1978 and 1991, the number of involuntary sprung up that organized group charters, oversales fell from 6.4 per 10,000 passengers buying large blocks of seats from airlines boarded to 1.1 per 10,000, while the number at significant discounts and then re-selling of voluntary oversales during that same them for prices that were lower than the period increased from approximately 1 regular fare. Discussing the situation with to 15 per 10,000 passengers boarded. As his staff, Bob Crandall, who was to become you might imagine, the frequency with president of American Airlines, but was which passengers complained about being then head of marketing, said, ‘‘We’re already denied boarding also dropped dramatically selling 40 percent of our seats at regular [18]. fares, so why can’t we figure out a way to sell Although the practice of overbooking in the other 60 percent at a fare cheaper than the airlines certainly continues, the intro- what a charter operator can get?’’ [19]. duction and proliferation of nonrefundable Super Savers were American’s response fares has significantly decreased its benefits. and other airlines followed American’s lead. The need for overbooking has also declined as Super Savers required advance purchase and cancellation and no-show rates have declined had a 14-day minimum-stay requirement (a dramatically. round-trip purchase was thus required). Rec- ognizing that business travelers were willing and able to pay higher fares, the restrictions were designed to prevent business travelers from purchasing them. In addition, on many 6Offering a credit for future travel may increase an airline’s revenue as it can stimulate trips that flights the number of Super Saver tickets that would not have been taken and the cost of the trip could be purchased was controlled (i.e., only may be for more than the voucher. Compensating alimitednumberofsaleswerepermitted)to passengers with a free trip does not allow for this. avoid displacing demand from later-booking Travel credits are now more typical. business travelers who were willing to pay REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 7 higher fares. For the most part, the airlines more than 15%. But if some of these dis- relied on various ‘‘rules-of-thumb’’ and the counted fares are sold to those who would judgment and intuition of inventory control have purchased a full fare, the gains are staff to set and manage discount allocations. quickly eroded. Even worse, revenue may Revenue management was expanding from decline. If 5 of the 20 discounted fare tickets overbooking to discount allocation control, are purchased by passengers who would have although controlling the number of discount purchased a full fare ticket, the incremental fares sold so as to maximize flight revenue revenue is only 8.6%. If the discounted fares was not carried out with the accuracy or are sold to 12 of these passengers, revenue mathematical support that was soon to come. actually declines by more than one percent, Nonetheless, the introduction of Super although the flight’s load factor, the percent- Savers was truly a watershed event in the age of seats occupied, increases by more than history of revenue management. It was a 11%. Had the discounted fare been $125, a unique concept. Not only did it allow the 67% reduction, revenue declines when only 5 same product to be offered for sale at mul- of the 20 discounted fares are sold to those tiple prices simultaneously, but consumers who would have purchased a full fare. As we of the products, not the firms that were shall discuss, this was soon to be a real sce- offering them, were empowered to trade-off nario. Profit levels erode even faster when the purchase and usage restrictions versus price variable costs of transporting and servicing and decide which product to purchase. The passengers are included. number of Super Savers that could be sold on The potential value of optimally managing any flight, however, was up to the discretion discount fare sales may be appreciated more of the airline. easily by analyzing the potential impacts of At this point, it is useful to examine two ‘‘small mistakes.’’ In the example above, the key points in more detail: incremental revenue drops by approximately 9% for each passenger that purchases a dis- 1. Large financial benefits are possible counted fare rather than the full fare (i.e., when a seemingly homogeneous prod- aflightwith69fullfarepassengersand20 uct is sold for multiple prices. discounted fare passengers earns 9% less rev- 2. The extent to which these gains are enue than a flight with 70 full fare passengers achieved depends on how well these and 20 discounted fare passengers). Taken sales are controlled. from that perspective, the financial impact of one less full fare sale on a flight, let alone sev- Suppose a flight has a seating capacity eral, is sufficiently great as to merit careful of 120, the full fare is $380 and the dis- management. counted fare is $210, a 45% discount from Now, consider those flights for which full fare. As shown in Table 2, if offering demand is sufficiently high to fill every discounted fares leads to incremental pas- seat. Further, suppose that none of the sengers without reducing the number of full passengers willing to pay full fare purchases fare passengers, the incremental revenues adiscountfare.Whatisthefinancialimpact can be significant. If, in addition to sell- of displacing a potential full fare passenger ing 70 full fare tickets, 20 discounted fares by a passenger purchasing a discounted fare? can be sold, the flight’s revenue increases by That is, how much less incremental revenue

Table 2. Potential Revenue Impacts of Offering Two Fares Full Fare Discounted Fare Load Factor (%) Revenue ($) Change in Passengers Passengers Revenue (%) 70 0 58.3 26,600 N.A. 70 20 75.0 30,800 15.8 65 20 70.8 28,900 8.6 58 20 65.0 26,240 1.4 − 8REVENUEMANAGEMENTINTHETRAVELINDUSTRY

Table 3. Potential Revenue Impact of Displacing Full Fare Passengers Full Fare Discounted Fare Load Revenue ($) Incremental Revenue Passengers Passengers Factor (%) Revenue ($) Reduction (%) 70 0 58 26,600 Baseline N.A. 70 50 100 37,100 10,500 Baseline 69 51 100 36,930 10,330 1.6 65 55 100 36,250 9,650 8.1 is earned if there are 69 full fare passengers Overseas Airways Corporation (BOAC) laid and 51 discount fare passengers on the flight, out the underlying theory and insights for rather than 70 full fare passengers and 50 using operations research techniques to set discount fare passengers? discount control allocations to maximize As shown in Table 3, the flight’s incremen- flight revenue when multiple fares are sold. tal revenue decreases by approximately 1.6% The ideas were presented by Kenneth Little- for each full fare passenger that is displaced wood, a member of the team, at a meeting of by a discount fare passenger. Perhaps, this the Airline Group of the International Feder- does not seem like much. After all, is not ation of Operations Research Societies (AGI- 98.4% of perfect still pretty good? Is it really FORS). At the end of a paper that was written worth the effort to pursue an extra 1.6%? If primarily to document BOAC’s efforts to this airline has 1000 flights per day and 15% implement mathematical and statistical of its flights sell out, this ‘‘allocation error’’ methods to set overbooking levels to max- results in $9.3 million less revenue than the imize revenue, Littlewood also described a airline could have earned. Allocation errors of method for using demand forecasts and opti- only several seats per flight on sold out flights mality conditions to set limits on the number lowers its annual revenues by tens of mil- of discounted fares that should be sold for a lions of dollars. And while these numbers are flight to maximize flight revenues. According only illustrative, they are sufficiently accu- to Littlewood, the ideas on discount allocation rate to demonstrate that the financial differ- were ‘‘still in its infancy,’’ but the introduc- ence between optimally managing discount tion of new low fares by Skytrain for regular allocations and only managing them well is London–New York service highlighted the extraordinary. This is why travel companies need to control the allocation of low yield fares have been willing to invest heavily in achiev- [20,21]7.Hecouldnothaveenvisionedthat ing marginal improvements in their revenue this method, eventually termed Littlewood’s management programs. It was a lesson that Rule, was to become the basis for discount was quickly learned and acted upon by some, allocation algorithms employed in revenue but not all, airlines in the early 1980s. management systems at many airlines. Although Super Savers provided the Littlewood’s key observation (somewhat needed innovation in the pricing structure, a simplified here) was that to maximize a few more years and an impending financial flight’s revenue when two fares are offered crisis would be required before the airlines (e.g., let Y be the full fare and D the discount fully recognized and embraced the practice of fare), the discounted fares should be sold so offering multiple prices with a variety of purchase restrictions to obtain incremental sales and profits. Indeed, it was not just the pricing structure, but also the ability to expertly 7 manage the discount allocations that was The paper was entitled ‘‘Forecasting and control of passenger bookings’’ and was published also needed; but this was not yet in place. in the AGIFORS 12th Annual Symposium Pro- Fortunately, the work to do this had ceedings, October 1972, pp. 95–117. Fortunately, already begun. Although not aware of how this ground-breaking and quite readable paper has important it would ultimately be, in 1972 been reprinted and the reference listed at the end the Operational Research Branch of British of this article is far easier to access. REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 9 long as afareclasswasavailable,allfaresthatwere mapped into it could be sold.8 When a fare D P Y or, equivalently, P D/Y class was closed, none of the fares mapped ≥ × ≤ into that fare class could be sold. Typically, where P is the probability that selling an many fares were assigned to each fare class additional discounted fare will result in one [4,22]. Over time, the number of fare classes less full fare ticket sale. has increased considerably. A flight may now This implies that discounted fares should have as many as 16 different fare classes. be sold so long as the ratio between the Consequently, yield management assu- discount fare and the full fare (D/Y)isat med an indirect responsibility for price least as great as the likelihood that the setting. While others determined the actual overbooking limit will be reached if no more fares that could be offered to the public, yield discounted fares are sold. Essentially, it is management controlled when and which a comparison between the revenue received fares were actually offered by determining from a discounted ticket and the expected when fare classes were open for sale and marginal revenue resulting from not selling when they were closed. Fare classes opened the discounted fare in anticipation of selling and closed multiple times during the booking afullfareinitsplace.Itisinterestingto period reflecting bookings, cancellations, and observe that the decision as to whether or revisions to discount allocation decisions not to accept an additional reservation for a based on changes in incremental demand discount fare does not depend on the demand forecasts. distribution for discount fares, but it does As you might imagine, demand forecasts depend on the demand distribution for full play a critical role in setting inventory allo- fares. cations. Demand was forecast at the fare In 1980, Bill Swan, a member of the class level. Time series models, such as expo- American Airlines Operations Research nential smoothing and booking profiles that Department, modified and extended Lit- provided estimates of how many reserva- tlewood’s methodology for implementation tions were typically received during various at American Airlines. Such a system was time periods prior to departure, were com- implemented in 1982 [4]. American was monly relied upon to forecast demand by fare probably the first US airline to implement class. Consequently, the forecasts of future such an advanced method for determining or incremental demand were independent discount allocations. This method was of potential yield management or pricing extended to multiple fare classes and has actions (except to the extent that analysts come to be known as the expected marginal implemented manual overrides). Even with seat revenue (EMSR) approach. The earliest such simplifications, the forecast models per- published work on this method, including formed well. extensions and modifications of the approach With a pricing structure in place that can be found in Refs 22–25. allowed airlines to target fares at multiple Discount allocation controls were designed market segments and the ability to control to work within airline reservation system the availability of these fares based on science control structures. The mechanisms by which rather than human intuition and judgment, reservation systems controlled fare availabil- the airline industry, or perhaps more specif- ity differed from that of many other indus- ically American Airlines, was prepared to tries. Each fare would be mapped into one of perhaps five or eight fare classes for a flight, 8In some cases, fare-specific restrictions prevented depending on the reservation system. Fare a fare from being sold when its fare class was classes were designated by letter codes. For available. For example, if a fare with a 21-day example, F, C, and Y were commonly used for advance purchase restriction was one of the fares first class, business class, and full coach fares, that was mapped into the V fare class, that fare respectively. Discounted coach fare classes could not be sold 18 days prior to the flight, even if included letter codes M, B, V, H, and Q. When the V fare class was still open. 10 REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY demonstrate that revenue management could discipline known as yield management would transform the competitive landscape in ways soon enter the public lexicon. that no one had previously imagined. All that The business community did not yet was needed was the catalyst for action. understand the potential benefits of the new That catalyst came in the form of People pricing structure. As reported by Time, ‘‘To Express, a low-cost airline that had begun Wall Street, American’s move looked like an operations in 1981 by offering low-fare short- attempt to get out in front of the competition haul service between cities that had lim- by jumping off a cliff. Many airlines are ited air service. For $23, you could fly from already losing money, and discount fares Newark, New Jersey to Buffalo, New York; may mean bigger losses. Stock prices of for $35 you could fly from Newark to Norfolk; all the large airlines that joined in last these fares were 50–80% lower than those week’s fare war fell sharply’’ [27]. In fact, offered by other carriers. Over time, however, however, American enjoyed great financial People Express expanded its route struc- success during the next few years; revenue ture and by 1984 was competing on the management was only one of the reasons for same routes that were flown by major car- this, but it was an important factor. riers such as American Airlines. Faced with Although it had a more gradual impact such competition, the major carriers faced a on revenue management, the Airline Dereg- dilemma: matching the fares offered by Peo- ulation Act of 1978 also transformed airline ple Express would enable them to keep their routes. Given much greater flexibility in customer base but not allow them to cover determining their routes and schedules, major airlines developed the hub-and-spoke their costs, whereas not matching the low system in the 1980s. Major airlines reduced fares would result in the loss of too many their frequency of nonstop service between passengers [26]. small and mid-sized cities (sometimes drop- American’s competitive response came in ping these routes), forcing more passengers January 1985, in the form of its Ultimate to take connecting flights. For example, Super Savers, restricted fares that were dis- whereas about 10% of American’s passengers counted as much as 70% so as to be compet- took connecting flights in 1980, by the mid- itive with those of People Express. The fares 1980s about two-thirds of the passengers provided far greater discounts than Super going into a hub airport were connecting to Savers. Also, new restrictions such as limited another flight [4]. By the late 1980s, airlines refundability levels were included. Although began to make fare allocation decisions that discount allocation controls in the reservation reflected passenger itineraries: this aspect system were now automated and benefit- of revenue management is known as traffic ted from data feeds from decision support management. systems, American’s initial ability to opti- For airlines that experienced significant mally manage the discount allocation levels increases in their volume of connecting left much to be desired simply because its passengers, the traffic management aspect historical passenger demand data did not of revenue management was significant. reflect the level of demand for the new fares. When most passengers fly point-to-point, With each passing month, however, Amer- airlines that attempt to maximize their ican’s ability to maximize its revenues by revenues/profits by setting discount allo- controlling and managing the availability of cation levels on a flight-by-flight basis these fares would improve. Other airlines perform rather well. As connecting traffic would also ramp up their investments in increases, however, this practice becomes methods and systems to better control dis- more problematic. Indeed, as we shall soon count fare availability. The stage was set see, in some industries such network-related for the introduction of a wide variety of dis- issues can dominate revenue management counted, restricted fares whose availability decisions. Traffic management, that is the would seem to appear and disappear as if by practice of setting allocation levels (and not magic and for no valid reason. With that, the necessarily only discounted fare allocations) REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 11 that reflect the incremental value of a 1980s and 1990s, approximately 40–50% of passenger’s revenue to the airline’s flight the potential benefits of revenue manage- network, became increasingly important and ment could be attributed to overbooking, eventually became incorporated into airline 30–40% to discount allocation, and 10–20% revenue management systems. to traffic management.10 While 10% of a The importance of traffic management can $500 million annual benefit is certainly be seen in the following example. Consider significant, it made sense for many airlines apersonwantingtoflyfromAustin,Texas to initially focus on overbooking and discount to London, England on a discounted fare of fare allocation, the more valuable areas, $329. As there is no nonstop service from prior to traffic management. For other Austin to London, this passenger might fly industries, however, this was not the case. from Austin, Texas to Dallas/Forth Worth, and then from Dallas/Fort Worth to London. Suppose someone else wants to fly from FROM YM TO RM: BEYOND THE AIRLINE Austin to Dallas/Fort Worth (DFW) on the INDUSTRY same flight and is willing to pay the Austin to DFW full fare of $189. Further, suppose As revenue management spread to the there is only one seat still available for sale hotel industry and then elsewhere, its on the flight from Austin to DFW. Which focus remained on inventory control. Prices reservation should the airline accept if it continued to be set independently of revenue wants to maximize its revenues? management. Marriott began investing in Excluding variable costs and the likeli- revenue management as early as 1985 [28] hood of either passenger booking a different and other hotel companies soon followed. flight on the same airline, the answer is, it According to Sheraton’s Director of Market- depends.9 Specifically, if the flight from DFW ing Geoff Ballotti, Sheraton began imple- to London becomes full and demand must menting a hotel revenue management system be turned away, it is highly likely that the offered by Control Data Corporation (CDC) airline earns greater revenue by transport- in 1987, implementing the system at 28 of its ing the passenger flying from Austin to DFW hotels [5–7]. Hilton also implemented a rev- and turning down the passenger flying from enue management system in the late 1980s. Austin to London. If, however, the flight from Some hotel chains began to experiment DFW to London departs with empty seats, with offering discount rates with restrictions the airline is likely to earn greater revenues such as advance purchase requirements, by transporting the passenger flying from but for the most part, it would take several Austin to London. more years until hotels integrated advance Compared to overbooking and discount purchase requirements and refundability allocations, traffic management is sig- restrictions into their pricing structures. nificantly more difficult to address and Rather, discounted rates were targeted to generates more modest incremental revenue specific dates (e.g., weekends, holidays) or (although certainly significant) than setting groups (e.g., children, travel agents, senior flight-independent discount allocation levels. citizens) or were unrestricted rates off the Although the benefit of overbooking has full rate. The discounted rates were tied to declined in recent years as nonrefundable the full rate (known as the rack rate) by some fares have become more prevalent, in the pre-determined percentage or value. Conse- quently, the early hotel revenue management systems served a slightly different function than airline systems, as they were used to 9Such factors should, and are likely to, be considered in an airline’s revenue management system although it was more likely than not for such factors to be ignored or only marginally considered 10Personal communication with Barry Smith, for- in the initial revenue management systems imple- mer Chief Science Officer for SABRE and also the mented by the airlines. author’s experience. 12 REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY control which of several rate tiers should be typically referred to as optimizing by length- offered, rather than controlling discount allo- of-stay (length-of-rent in the car rental indus- cations. Each rate tier included a set of rates. try). It is far more profitable to earn $149 per The revenue management systems mea- night for four nights than $229 for a one- sured demand pressure for each check-in date night stay. Recommendations to close the using four primary criteria: reservations on lower rate tiers for a peak night to all guests, the books, time until check-in date, a fore- regardless of whether they wanted to stay for cast of incremental demand for that date, one, four or seven nights, could easily discour- and a forecast of remaining rooms. As the age those guests who wanted to stay several magnitude of the demand pressure increased, nights or longer. Unfortunately, the early these systems would recommend raising the hotel revenue management systems were not rate tier offered. Consequently, many of the designed to analyze demand patterns across various rates offered by the hotel to the dates and address this issue. Nor were many general public would increase, or decrease, hotel reservation systems designed with the simultaneously, by pre-determined amounts. controls needed to implement such decisions. Some discounted rates would not be included The systems were not designed to control in a rate tier, so these rates would not be room inventory in ways that would stimu- available when that rate tier was offered. late or accept demand from guests wanting The initial application of revenue man- longer-stays, while simultaneously limiting agement in the hotel industry suffered from the number of shorter-stays on peak nights. missteps that concerned some hotel exec- Asimplifiedexampleillustratesthis utives sufficiently to delay their adoption trade-off. Consider a hotel with a rate of of revenue management systems. Borrowing $149/night, for which there is sufficient from the airline industry’s revenue manage- demand to sell-out on Wednesday night, ment ‘‘playbook,’’ perhaps a little too much, even if the rate increases to $229/night. As some of the initial hotel revenue manage- shown in Table 4, increasing the rate from ment systems incorporated a design simpli- $149 to $229 on Wednesday, an increase of fication that served the airlines well; they more than 50%, results in 1135 occupied essentially ignored traffic management, the rooms during the week and the weekly network aspect of revenue management.11 Revenue managing a flight independently of revenue exceeds $193,000. Further, every others is similar to revenue managing arrival room is occupied on Wednesday night. dates at a hotel independently of other dates. Conversely, if the hotel implements length Unfortunately, this simplification was apt to of stay (LOS) controls so that more of the yield poor rate recommendations during peak reservations it accepts that span Wednes- periods for hotels with a high proportion day night include the surrounding days, and of multiple night stays. When guest stays does not increase its average rate on Wednes- span multiple dates, managing dates inde- day, the number of rooms occupied during pendently is apt to lead to pricing mistakes. the week increases by 195. The weekly rev- As it turns out, compared to increasing enue for the hotel increases to $198,000, rates on dates that ultimately sell out, hotels about a 2.6% revenue increase. In this sce- generally earn far greater revenues and prof- nario, not all of the hotel rooms are occupied its by limiting the number of shorter stays on Wednesday night, reflecting the uncer- so that they can accept a greater number of tainty associated with holding back rooms for longer stays that span these nights. This is longer stay reservation requests. In actual implementation of length of stay controls, some hotels have claimed revenue increases of 8–10% or even more when compared to 11Nor were overbooking recommendations incorporated into these systems, although this was more increasing rates on peak nights [29]. Indeed, likely for business process reasons. Hotel execu- in the above example, if the Wednesday rate tives tend to be very cautious about systematically had only increased to $189, an increase of overbooking. approximately 25%, the hotel would earn REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 13

Table 4. Comparison of Impacts from LOS Controls versus Increasing Price Mon Tue Wed Th Fri Total Available rooms 300 300 300 300 300 1500 Occupied rooms if Wednesday rate 190 235 300 220 190 1135 increases to $229 Revenue if Wednesday rate 28.3 35.0 68.7 32.8 28.3 193.1 increases to $229 ($000) Occupied rooms with LOS controls 245 280 295 275 235 1330 and rate of $149 Revenue with LOS controls and 36.5 41.7 44.0 41.0 35.0 198.2 rate of $149 ($000)

9.4% more revenue by using length-of-stay week helped the hotel obtain a 12.3% controls rather than increasing its rate. year-over-year revenue increase, despite an In practice, combining a rate increase for 11.7% reduction in average daily rate [5–7]. shorter stays while allowing longer guest The adoption of yield management by the stays to access lower rates is generally the hotel industry in the late 1980s and early most profitable pricing strategy for hotels 1990s is well documented. Less information during peak periods. The benefits of this is available about the passenger railroad and strategy are recognized within the hotel cruise line initiatives at this time. Given industry. Hotel systems and practices have permission to provide consulting and sys- been redesigned to enable this strategy to be tem development expertise to other firms, effectively implemented: no longer can guests the American Airlines Operations Research make a reservation for a longer stay to obtain Group began developing a yield management alowerrateandthencancelsomeofthedays system for Amtrak in 1987 and a year later in the reservation or simply check-out early. was doing the same for Royal Caribbean Aggressively increasing rate tiers on Cruises Ltd. (RCCL). Focused on enhanc- peak nights also had a negative unintended ing the inventory control capabilities of these side effect. Able to provide hotels with high firms, these systems did not provide any pric- volumes of demand over the year, many ing recommendations. corporations negotiated fixed rates with Among the first areas to be addressed by hotels that were significantly lower than the the RCCL effort was improving the accuracy unrestricted rates. The corporate negotiated of pre-departure cancellation rate estimates rates were not affected by the hotel’s pricing for individual and group reservations [30]. At actions. Raising the hotel’s rates on peak the time, cruise lines typically estimated can- nights might reduce hotel profitability as cellation rates based on fixed projection rates it could result in accepting more guests that reflected how much money had been with lower corporate rates; because the rate received by the cruise line and possibly one or increase slowed down the pace of bookings two other factors. Known as a handicapping for publically available rates, it enabled formula, the methods generally provided corporations with the lower negotiated rates reasonable average estimates over the course to access more rooms on peak nights, which of the year, but were subject to ‘‘catastrophic’’ was exactly the opposite of what the hotels breakdowns on individual sailings, leading to wanted. undesirable pricing actions. Over-forecasting By 1990 these shortcomings were cancellations could lead to selling cabins at recognized and soon thereafter, addressed. highly discounted rates close to departure Marriott’s first generation yield management date. Under-forecasting cancellations could system with length-of-stay optimization was require upgrading passengers into much implemented in 1991. Prior to this, Marriott’s more expensive cabins, or even worse, fully yield management systems had gone through or partially refunding their purchases while two generations. A test of the system at also moving them to other departures. the Munich Marriott during a high demand Segmenting reservations based on a variety 14 REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY of attributes (e.g., travel agency, itinerary, than travel. Within a few years, additional and type of group), and containing updating revenue enhancement techniques were mechanisms to capture trend and seasonal- included in revenue management initiatives. ity, the new method provided more accurate Perhaps the most critical development cancellation rate estimates [31]. Having was the inclusion of price optimization more accurate estimates of the number of capabilities; no longer would revenue man- additional cabins that needed to be sold on agement simply control the availability a departure enabled RCCL to make more of previously set prices. Instead, revenue profitable pricing decisions for each cruise. management efforts would include deter- Other elements of the RCCL yield man- mining the best price levels. In some agement initiative included a group evalua- cases, dynamically re-pricing a product tor model and a variety of forward-looking to maximize profits based on current and reports directing analyst attention to those forecast supply and demand extended the departures most in need of oversight and underlying mathematical models to include action. While the scope of RCCL’s initial demand elasticity estimates: forecasts of how yield management system might be consid- demand would change in response to offering ered rather modest, it was a key factor in alternative prices. Some (i.e., published) enabling RCCL to better manage and price accounts of revenue management systems its inventory, including implementing a vari- that incorporated dynamic pricing capabil- ety of pricing innovations. In 1992, Brian ities include National Car Rental [35], The Rice, then RCCL’s director of revenue plan- Moorings [36], and Princess Cruises [37]. ning and analysis, and now CFO, conser- The revenue management initiative for vatively estimated the incremental revenue Princess Cruises, begun in 1997, was imple- from these efforts at over $20 million/year, or mented in 2000. While the system included about 2–3% of RCCL’s annual revenue [5–7]. typical inventory controls, it also included a Approximately two years after Amtrak pricing analytics engine providing insights had initiated its revenue management effort, on the extent to which promotional prices SNCF, the French national railway, did the were needed to stimulate sales on a depar- same [32,33]. By 1989, revenue management ture. Of all the capabilities of the revenue efforts were also initiated at Avis, a rental management system, Princess staff believed car company. By the early 1990s all of the this capability to be the most beneficial.12 major car rental firms had implemented rev- Price optimization continues to be of great enue management systems or were preparing interest to many travel industry firms. Only to do so. The ability to evaluate and con- arelativelyfew,however,haveinvested trol reservations based on length-of-rent was significant resources. At least for some of included in all the first generation car rental these firms, however, the returns have been revenue management systems. The revenue extraordinary, easily eclipsing the 4% to 7% management system implemented at Hertz returns typically ascribed to the inventory also included fleet planning and fleet deploy- allocation controls of revenue management. ment capabilities enabling Hertz to better Anecdotally, price optimization capabilities determine where and when it was financially appear to be capable of generating revenue advantageous to transfer cars between rental increases of 10–15%.13 car locations due to projected supply and For many years, pricing has been treated demand imbalances [34]. Doing so probably as mostly art and very little science. For the made Hertz the first company to integrate most part, the science of pricing focused on revenue management with operational deci- sionmaking. Revenue management continued to 12Personal communication with Princess Cruises spread to other segments of the travel staff. industry, including tour operators, time- 13These estimates reflect the author’s experience share exchange, theme parks, and even as well as personal communication with Barry yacht rentals as well as industries other Smith. REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 15 estimating the costs of production and build- sufficiently unreliable recommendations that ing in a reasonable profit margin. The prod- firms stopped using the systems. uct’s price was simply the result. While many More recently, however, alternative mod- firms continue to engage in that practice, it eling approaches such as disjunctive mapping occurs less often as the principles of revenue are being developed to address these issues management obtain a greater foothold. The [38,39]. The widespread use of robotic pricing systematic integration of supply and demand tools that capture competitor prices enables forecasting with product design and mar- price-sensitive demand forecasts to explicitly ket segmentation concepts has proven suc- reflect different competitive situations and cessful, although there have certainly been support what-if analyses, as historical instances of failure and setbacks along the data has become available to estimate way. As was the case when revenue man- the likelihood of alternative competitor agement was initially applied to the hotel reactions. Indeed, one of the challenges of industry, many of these failed efforts resulted price-elasticity modeling is that in practice, when techniques that proved successful in demand curves may not be static; they one industry were used in another, without depend on the directional shift of price and sufficient regard for the differences in the possibly other factors. For example, when way business was conducted. price increases from A to B,themagnitude Over the past 20 years, the financial of the change in demand level may not be benefits of value-based pricing have become equivalent to the magnitude of the change in more widely accepted and communicated. demand when price decreases from B to A. The science of revenue management pro- Such modeling and estimation complexities vides a powerful mechanism that enables are more easily addressed when competitor value-based pricing to succeed. Indeed, with prices are retained. Given the potential restriction-free pricing becoming more and benefits of dynamic pricing, and the recent more prevalent in the airline industry, the availability of data to support demand fore- traditional discount allocation controls of cast modeling, we anticipate that dynamic revenue management are now giving way to pricing will be a strong focus of travel an increasingly strong focus on maximizing companies during the next decade or two. revenue via dynamic pricing to achieve pricing optimality. At the heart of any dynamic pricing CONCLUSION optimization system is forecasting demand at alternative prices and then estimating Written for academics and experienced the prices that will maximize profit. From a practitioners of revenue management as a practical perspective, forecasting demand at ‘‘single-source reference for the major theory different prices can be both complex as well and application issues,’’ a comprehensive as data intensive, as demand is affected by treatment of revenue management was anumberoffactorsotherthanprice.While published in 2004 by Kalyan Talluri and many companies have attempted to incor- Garrett Van Ryzin. The book is intended porate such variables into regression-based for those who have advanced degrees in forecasts, such efforts often produce less mathematically-oriented disciplines such than satisfactory results. Data sparsity and as operations research, statistics, or eco- the inability to model many key relation- nomics, although portions of the book are ships among the variables are some of the accessible to those without such skills and reasons that price–demand relationships are interested in obtaining a more detailed are less robust or accurate than desired. understanding of the theory underlying Manual overrides and factors designed to revenue management [40]. prevent forecasts and recommendations From a relatively obscure business prac- from being too extreme are often integral to tice only 50 years ago, revenue manage- such systems. While some of these methods ment has come a long way. Dominated by have proven useful, others have produced practitioners until 1990 and perhaps longer, 16 REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY the number of academic researchers may 5. Scorecard. Marriott’s Millions. Atlanta (GA): now surpass the number of practitioners. Aeronomics Incorporated; Premier Issue, Sec- Indeed, the subject now boasts two dedicated ond Quarter. 1992a. pp. 3–6. journals: The Journal of Revenue and Pric- 6. Scorecard. Royal Caribbean breaks through. ing Management was initiated in 2002 and Atlanta (GA): Aeronomics Incorporated; Third was joined by the International Journal of Quarter. 1992b. pp. 5, 11. Revenue Management in 2007. In addition, 7. Scorecard. Sheraton’s technological evolu- articles on revenue management appear in a tion. Atlanta (GA): Aeronomics Incorporated; wide variety of journals and books. Fourth Quarter. 1992c. pp. 3–5. Will the term revenue management be 8. Cross RB. Revenue management: hard-core retained? Will it be replaced by pricing and tactics for market domination. New York: revenue optimization or something else? Broadway Books; 1997. p. 276. What is your forecast? 9. Uchitelle L. Airlines off course. New York: New York Times Magazine; 1991. p. 612. Acknowledgments 10. Time. The law: a big bump for bumping. 1976. Available at http://www.time.com/time/ Igreatlyappreciatethetimegiventomeby magazine/article/0,9171,918209,00.html. both Kenneth Littlewood and Barry Smith to Accessed 1976 June 21. discuss some of the ideas in this article. They 11. Gowen MJ. Airline and cargo industry update. were able to provide me with key pieces of Murray Hill (NJ): DUNS Analytical Services, information that filled in gaps in the histor- Dun & Bradstreet, Inc.; 1991. p. 33. ical record and allowed me to better under- 12. Business Travel News. Airlines. 1993. pp. 40, stand the motivation behind industry devel- 42. opments. As my mentor in revenue man- 13. Lieberman WH. A revolution is brewing in agement at American Airlines, I am deeply pricing. Los Angeles: Los Angeles Times; 1990. indebted to Barry for both, the time he has (Looking Ahead), June 6, 1990. given to me and value the friendship we have 14. Cross RG, Lieberman WH. Yield manage- developed. I also owe a large debt of grati- ment. Futurescope, management and society. tude to Tom Cook, who headed the American Cambridge (MA): Decision Resources Inc.; Airlines Operations Research Department in 1991. pp. IV–1–IV–4. 1984. Without his support and trust, I would 15. Rothstein M. OR and the airline overbooking never have had the opportunities in revenue problem. Oper Res 1985;33(2):237–248. management that have captivated me for 16. AMR Corporation. 1984 Annual Report. 1985. more than 25 years. p. 44. 17. Simon JL. An almost-practical solution to air- REFERENCES line overbooking. J Trans Econ Policy 1968;2 (2):201–202. 1. Orkin EB. Yield management. Hotel & Cater- 18. Simon JL. The airline oversales auction plan: ing Technology; 1988. (Reprinted by Hotel the results. J Trans Econ Policy 1994;28(3): Information Systems Ltd., Heathrow, Houn- 319–323. slow, England.) 19. Sterling R. Eagle: The Story of American Air- 2. Schrage M. New corporate pricing tool could lines. New York: St. Martin’s Press; 1985, p. be costly to consumers. California (CA): Los 482. Angeles Times. 1991 Dec 5: D1. 20. Littlewood K. Forecasting and control of pas- 3. Lieberman WH. Yield management: the pricing revolution that goes beyond the travel senger bookings. J Revenue Pricing Manage industry. Presentation at the 4th Annual US 2005;4(2):111–123. Pricing Conference, The Pricing Institute, a 21. Littlewood K. Personal communications. division of The Institute for International 2010; Research; 1991 April 10–12. New York: The 22. Belobaba P. Air travel demand and airline Grand Hyatt Hotel; 1991. p. 3. seat inventory management [Doctoral dis- 4. Smith BC, John L, Ross D., Yield Man- sertation]. Flight Transportation Laboratory. agement at American Airlines. Interfaces Report R87-7. Cambridge (MA): MIT; 1987; p. 1992;22(1):8–31. 214. REVENUE MANAGEMENT IN THE TRAVEL INDUSTRY 17

23. Belobaba PP. Survey paper – airline yield Prometheus: globalization, history and tech- management: an overview of seat inventory nology. Denmark: Aarhus University Press; control. Trans Sci 1987b;21(2):63–73. 2004. pp. 193–216. 24. Belobaba PP. Application of a probabilistic 34. Carroll WJ, Grimes RC. Evolutionary change decision model to airline seat inventory con- in product management: experiences in the trol. Oper Res 1989;37(2):183–197. car rental industry. Interfaces 1995;25(5): 25. Curry RE. Optimal airline seat allocation with 84–104. fare classes nesting by origins and destina- 35. Geraghty MK, Johnson E. Revenue manage- tions. Trans Sci 1990;24(2):193–204. ment saves national car rental. Interfaces 26. Phillips RL. Pricing and revenue optimization. 1997;27(1):107–127. Stanford (CA): Stanford University Press; 36. Bermudez R, Dieck T, Lieberman W. Revenue 2005. p. 355. management basics in the charter boat indus- 27. Time. Sky wars: airfares take a dive. try. Revenue management and pricing: case 1985. Available at http://www.time.com/time/ studies and applications. London: Thomson; magazine/article/0,9171,959288,00.html. 2004. pp. 1–8, 184–188. Accessed 1985 Jan 28. 37. Scorecard. Making history. Mountain View 28. Hsu CH, Powers TF. Marketing hospitality. (CA): Talus Solutions; 2000. pp. 5–8. New York: John Wiley & Sons, Inc.; 2002. p. 38. Raskin M, Lieberman W, Mullin J. Disjunctive 360. mapping: changing the way we understand 29. Aeronomics. A conversation with Don Burr. and predict customer behavior (part one). Fourth Quarter. Atlanta (GA): Scorecard; JRevenuePricingManage2010.Inpress. 1992. pp. 6–7. 39. Raskin M, Lieberman W, Mullin J. Disjunc- 30. Fisher J, Mongalo Marco. Integrating deci- tive mapping: changing the way we under- sion support. SAS Users Groups Interna- stand and predict customer behavior (part tional Conference. New York; 1993. pp. two). J Revenue Pricing Manage 2010b. 619–623. In press. 31. Lieberman WH. Pricing tactics and strategies 40. Talluri KT, Van Ryzin GJ. The theory and in the cruise industry. Handbook of pricing practice of revenue management. Boston management. Oxford University Press; 2010. (MA): Kluwer Academic Publishers; 2004. p. In press. 712. 32. Daudel S, Georges V. Le yield management: la 41. Cowan AL, Gargan EA. Mirage of discount air face encore cachee´ du marketing des services. fares is frustrating to many ﬂiers. New York: InterEditions. Paris: 1989. New York Times; 1991. pp. A1, C3. 33. Mitev NN. The globalisation of transport? 42. Simon J. Origins of the airline oversales Computerised reservation systems at Amer- auction system. Cato Rev Bus Gov. Avail- ican Airlines and French Railways. In: Lynch able at http://www.cato.org/pubs/regulation/ P, Trischler H, Lyth J, et al.editors.Wiring regv17n2/reg17n2-simon.html.