AGIFORS Reservations & Revenue Management 2004
Technical Program

Presentations ( Based on submitted abstracts to date* )

Technical Presentations

Revenue Management Performance In A Low-Fare Airline Environment

Thomas Gorin and Peter Belobaba - MIT International Center for Air Transportation

The growth of low-fare airlines has changed the competitive environment in many ways, but it has not diminished the importance of effective revenue management, for traditional and low-fare airlines alike.  This presentation summarizes results of Passenger Origin-Destination Simulation (PODS) research on the impacts of different RM techniques under the less restricted and more compressed fare structures typical of low-fare airline markets.  We also present findings as to the revenue performance of O-D control methods in large airline networks that include a mix low-fare markets and markets with a more traditional fare structure.

Algorithms For Revenue Management In Unrestricted Fare Markets

Craig Hopperstad - Hopperstad Consulting

Peter Belobaba - MIT International Center for Air Transportation

While the assumption of independent fare class demands was questionable even under traditional fare structures, it is completely invalid under the increasingly less restricted fare structures typical of low-fare airline markets.  We present some alternative approaches to RM seat inventory control of unrestricted fare classes, including simple load factor threshold methods, sell-up models, and dynamic programming techniques.  We use the Passenger Origin-Destination Simulation (PODS) to assess the performance of these RM methods in a competitive airline market environment.  

Changes in Passenger Purchasing Behavior and Their Impact on Revenue Management Models

E. Andrew Boyd, Royce Kallesen, and Ed Kambour - PROS Revenue Management

The changing business environment is making ticket price an ever more important consideration in passenger purchasing behavior. We discuss how traditional revenue management models change as the ability to segment passengers deteriorates and passengers focus on price.

Models of the Spiral Down Phenomenon

Anton Kleywegt – Georgia Tech

Tito Homem de Mello - Northwestern University

William Cooper - University of Minnesota

It has been observed in simulations of revenue management processes that if the revenue manager thinks that each customer demands a particular fare class, but customers are actually more flexible and tend to prefer cheaper fares, and the revenue manager uses observed sales to forecast demand, then the observed demand for cheaper fares will tend to increase over time, and the protection levels and average prices will tend to decrease over time.  We study a number of models to gain a better understanding of when and how this phenomenon occurs, and show that unconstraining and the use of good forecasting methods may not prevent spiral-down if the demand model is not correct.

Delta's experience on O&D RM

Brian Wishlinski - Delta

Installation of an origin-destination revenue management system is a daunting task for most airlines.  In addition to the hardware and software components, the business teams are asked to transform their processes to support the new paradigm.  Over two years ago, Delta Air Lines began the OMNI project to move the airline's revenue management department from leg-bucket control to origin-destination forecasting and optimization.  The system is scheduled for deployment during the first quarter of 2004.  In this presentation, Delta will discuss the installation of its new revenue management system, including the functional components as well as the transformation of the associated business processes.

A Dynamic Model for Airline Network Revenue Management

Eylem Tekin and Tao Huang – University of North Carolina

We consider dynamic policies for the seat allocation problem in a multiple-fare class, network environment with the objective of maximizing total expected revenues. We first derive the structural properties of the optimal policy. Then, we propose an algorithm that approximately computes the dynamic programming value functions. The algorithm uses the corresponding linear programming relaxation and linear regression. We present computational results, and compare the performance of this algorithm with the existing ones in terms of the quality of solution and speed.

On the Choice-Based LP Model for Network RM

Garrett van Ryzin - Columbia University

Recently, Gallego, Iyunger and Phillips proposed a choice-based LP model for network revenue management which parallels the deterministic linear programming model widely used in current network RM. We build on their work in two ways: First, we characterize the offer sets (sets of available fare products) produced by this model using a notion of "efficiency" and show that the resulting offer sets are asymptotically optimal. Second, we propose a practical decomposition heuristic that significantly improves on the performance of the static LP solution. We illustrate the method on several numerical examples and discuss various extensions.

Flexible and Callable Revenue Management

Guillermo Gallego, Garudy Iyengar, and Steven Kou - Columbia University

Robert Phillips – Stanford University

We show how to enhance revenues and improve capacity utilization by selling flexible and callable products. A flexible product, e.g., a morning flight from New York to San Francisco or a four star hotel in the theater district, consists of two or more specific products servicing the same market. Customers know the set of specific products that comprise a flexible but are not assigned a specific product until a later time. A callable product is a specific or flexible product that can be called (repurchased) by the capacity provider at a pre-agreed strike (repurchase) price from self-selected customers. Callable products represent an attractive alternative over the more cumbersome alternative of calling and bribing low fare customers or bumping them at the gate. We present an analysis of the two-fare class problem with flexible and/or callable products as well as a multi-fare network demand choice model.

Real-Time Revenue Management

Scott Chandler, Shau-Shiang Ja, and Tim Jacobs - American Airlines

Current revenue management strategies utilize batch processes to set inventory control levels for aircraft seat inventory.  These batch processes typically run once every 24 hours to once every 14 days.  As a result, the inventory controls become stale as new bookings (reservations) occur.  To account for the change in booking levels, most revenue management systems utilize gradient approximations to represent the change in the inventory controls as bookings occur. The gradients tend to underestimate or overestimate the optimal inventory control, resulting in lost revenue from either dilution or unnecessary spill.

Real-time revenue management re-optimizes the O&D controls after every booking or at short intervals.  This allows the airline to more accurately estimate the inventory controls and prevent unnecessary loss of revenue due to stale inventory controls.  Benchmark results illustrate revenue gains due to the implementation of a real-time revenue management process at American Airlines.

Using Expected Maximum Utility to Calculate the Network Value Index for Airline Network Enhancements

Roger A Parker and Richard Lonsdale - Boeing Commercial Airplanes, Marketing

It has been recently proved that a logit choice model yields a particularly simple and attractive functional form for the computation of the maximum expected utility.  This is the basis proposing a generalized “Network Value Index” for the comparison and evaluation of proposed enhancements to an existing airline network, such as timing, gauge, or routing changes.  This measure is important since yield management systems attempt to optimize the capture of any consumer surplus that arises from such network enhancements, and this approach can measure the theoretical upper limit of such available consumer surplus.  The presentation will describe the underlying theory and the derivation of the Network Value Index measure, and display the empirical data used to establish the validity of the measure.  We will illustrate its use in fleet, schedule and pricing planning contexts.

A New Approach to O&D Revenue Management Based on Scenario Trees

Andris Moeller+, Werner Roemisch+ and K. Weber* (+ Humboldt-University Berlin, * Lufthansa Systems)

O&D revenue management (RM) – either leg-based or PNR-based – has become a standard in the airline industry. Despite its holistic view on the airline network, O&D RM has adopted techniques from traditional leg-based RM: Forecast and optimization systems are separated, optimization assumes certain probability distributions, and changes of the demand and supply situation in the network are responded by re-calculations of forecasts and booking controls. Whereas distribution assumptions has always been arguable, ad hoc re-optimizations of big hub-and-spokes networks has become a new challenge. In our contribution, we present a new approach which does not make any distribution assumptions and computes control parameters for a variety of demand patterns over the complete booking period. Protection levels are determined for all origin destination itineraries, fare classes, points of sale and data collection points (DCPs). This approach to the seat inventory problem is modeled as a multistage stochastic program where its stages correspond to the DCPs of the booking horizon. The stochastic program represents a specially structured large scale LP. In order to keep the problem size feasible, the number of scenarios is reduced according to a given goodness-of-fit based on distances of probability measures. Finally, the model is solved by standard linear programming software (e.g. CPLEX). Preliminary numerical experience will be reported.

The Theory and Practice of Dynamic-Programming Based Bid Prices

Larry Weatherford and Aleksey Khokhlov- University of Wyoming

Dynamic programming (DP) has been touted as the theoretically optimal way to control O&D requests.  But what are the practical concerns for implementation, if any?  This presentation will compare EMSRb leg controls versus LP-based and DP-based bid price control.  We will present results on both the time required to make the relevant calculations as well as expected revenue results.

Online Low Price Guarantees - A Real Options Analysis

Chris K. Anderson - University of Western Ontario

A common practice among large retailers is the low price guarantee, rebating consumers if they find an identical product cheaper elsewhere, or if the retailer discounts the good within a specified timeframe. This provides consumers with some level of comfort in their purchase decision.  A similar low price guarantee is provided by numerous service industries that allow reservation of capacity yet don't penalize the consumer for failure to keep that reservation; examples include hotels and car rental.  Given that a consumer is not required to keep the reservation, they may make another reservation either at a competing firm or the same firm if future prices decline (a similar dilution of reservation occurs with corporate clients whom are allowed to also pay regular rates).  The increasing availability of pricing information on the Internet affords consumers the opportunity to be more strategic in their behaviors. As a consumer we are able to quickly and easily check prices from numerous service or goods providers. The ease of price information potentially makes these guarantees very costly to the service or good provider.  We analyze the implied costs associated with these guarantees by making analogies to financial options.  Motivation for this research comes from a large car rental firm, Dollar Thrifty Automotive Group Inc., which is currently considering offering a low price guarantee to all consumers whom book a reservation though their website.

Optimisation Issues in Low Cost Revenue Management

Klaus Weber, Ruediger Thiel - Lufthansa Systems

Low Cost Carriers has not only uttered confusion to the airline business world, but also upset system vendors by introducing concepts that does not suit into the traditional revenue management systems. Alike the situation in Middle-earth before the return of the King, everyday life has become exhausting: Time series are no longer suitable for demand prediction. Customer segmentation seems to be out of date. Even those funny nesting structures, we liked to brood about are vanishing. Not to speak of the nonuniform ways of Low Cost Carrier’s inventory control. In this talk we like to present an approach for Low Cost revenue management and address particularly some related optimisation problems. Like Elves and Men, we remembered the Old Alliances and gratefully added the powerful arrows of price elasticity to our tools.

Implementing RM in firms with a diffuse ownership structure: the case of ACCOR.

Jean Michel CHAPUIS, Mathieu PAQUEROT - University of La Rochelle

Abstract: This presentation proposes to study the inherent organizational problems with the implementation of a Area Revenue Management in a company managing a mix network, including various forms of establishments such as owned establishments, infrastructures managed by a contract of management and some franchised. First, this paper points out the interests of Revenue Management strategy that focuses on maximising hotels’ revenue, located in a same area. Second, the organisational problems can be analysed through the agency theory. Some solutions can also be considered from this point of view.

Predicting air travelers' no show and standby behavior using passenger and directional itinerary information

Laurie Garrow and Frank S. Koppelman - Northwestern University

This is the first study of airline travelers' no show and standby behavior based on passenger and directional outbound/inbound itinerary data.  The presentation describes passengers' behavior using a multinomial logit methodology for domestic U.S. itineraries departing in March 2001 or March 2002.  This enables us to explore behavioral differences based on passenger and itinerary characteristics as well as identify differences in rescheduling behavior occurring after September 11, 2001.  Benefits of using passenger data to improve forecasting accuracy and support a broad range of managerial decisions are described.

To Buy or Not to Buy: Mining Airfare Data to Minimize Ticket Purchase Price

Rattapoom Tuchinda - University of Southern California

As product prices become increasingly available on the World Wide Web, consumers attempt to understand how corporations vary these prices over time.  However, corporations change prices based on proprietary algorithms and hiddle variables (e.g., the number of unsold seats on a flight). Is it possible to develop data mining techniques that will enable consumers to predict price changes under these conditions?

This talk reports on a pilot study in the domain of airline ticket prices where we recorded over 12,000 price observations over a 41-day period.  When trained on this data, Hamlet -- our multi-strategy data mining algorithm -- generated a predictive model that saved 341 simulated passengers $198,074 by advising them when to buy and when to postpone ticket purchases. Remarkably, a clairvoyant algorithm with complete knowledge of future prices could save at most $320,572 in our simulation, thus HAMLET's saving were 61.8% of optimal.  The algorithm's saving of $198,074 represents an average saving of 23.8% for the 341 passengers for whom savings are possible.  Overall, HAMLET saved 4.4% of the ticket price averaged over the entire set of 4,488 simulated passengers.  Our pilot study suggests that mining of price data available over the web has the potential to save consumers substantial sums of money per annum.

Revenue Integrity

Mary Kay Caschetta - Northwest Airlines

NWA, Inc., is implementing a revenue integrity program.  This presentation will describe the process being used, and highlight key projects, implementation issues, etc.  

Forecasting Demand Using Competitive Fare Information

Dr. John Salch, Ed Kambour, and Royce Kallesen - PROS Revenue Management

Internet distribution channels have made real-time fare data available to competitors as well as to consumers. We discuss how competitive fare information can be used to improve demand forecasts. Results are shown using real data for two carriers operating in a competitive market.

Competitive Fare Shopping for YM

Richard Ratliff - Sabre Holdings

A problem with today's airline revenue management systems is that they are "closed systems"; little or no information is included about the carrier's current marketplace competitiveness in GDS and online website retail channels.  Changes in OA schedules, fares, and availability make this marketplace competitive landscape highly dynamic.  Although MIDT data provides some information along these lines, it is limited to bookings (not shopping activity).  Most airlines recognize these limitations and are supplementing their revenue management systems with data from other sources (e.g. FareChase, SideStep and/or by checking online retail sites such as Expedia, Orbitz or Travelocity).

The Sabre Labs and Research teams have developed an initial prototype of "Competitive Fare Shopping" using Travelocity fare search results.  Fare shopping queries and associated results have been captured and aggregated into numerous different summary views.  Such shopping activity at retail "storefronts" (either GDS's or websites) can help explain "why" a carrier's O&D fare class demand is moving in a particular direction.  Such information is useful to both pricing and yield management analysts for better insights into the current competitive environment (by market and future date).   Ultimately, it may also become an important data source for choice-based ODYM forecasting.

Competitive Intelligence based Revenue Management in a Restriction-Free Pricing environment

Anand Srinivasan, Sachin Bhatevara & Shankar Mishra – Sabre Airline Solutions

It is widely accepted that competitive intelligence is a vital factor in airline revenue management. Yet, most revenue management systems ignore this factor or at best approximate it heuristically. This area however, has been drawing significant attention in recent times, mostly due to the advent of “Restriction-Free” pricing environment. We present a methodology that uses competitive schedule and fare intelligence as part of inventory control decision process. This solution uses a discrete choice model to determine optimal inventory control for an airline at a fixed point in time, given multiple itineraries and fare product offerings within a market.  We will present some preliminary results and potential areas for enhancing the model.

Is there value in Fare mapping

Jean-Francois Page – Air Canada

Over the last few years, Air Canada has implemented several enhancements to its control mechanism (bid prices, interpolation to bid prices, etc.). To enhance revenue generation, Air Canada is currently considering many initiatives one of which is fare mapping. The objective is to find simple and easy to implement fare mapping scenarios to provide better control. We will present our most recent results of different simulation exercise and compare them with previous series of simulations.

Tour Operator vs. Airline: Impact on Revenue Management Techniques

Ulrich Oppitz - Thomas Cook

The business model of an integrated tour operator will be compared with those of traditional airlines. Attendees will discover the differences in complexity and constraints. The main focus will be on the impact to Revenue Management methods, with particular regard to forecasting and optimization.

Optimal Airline Seat Allocation in a Fenceless Environment using Expected Revenue Maximization (ERM)

Hossein Tavana - Continental Airlines

Recently the notion of "yieldable" and "priceable" demand has been presented by researchers and practitioners. In this presentation a probabilistic linear programming method, namely expected Revenue Maximization (ERM), is introduced to obtain the optimal seat allocation. The method particularly prevents the spiral‑down of priceable demand in a fenceless environment, that is, where consumers would buy the lowest available fare.

Managing Revenue in Broken Fare Fence Environment

Krishnan Saranathan and Burak Ozdaryal – United Airlines

The aggressive pricing strategies pursued by low cost competition have had significant impact on unit revenue and yield for all carriers.  We will discuss the strategies pursued by United Airlines to improve unit revenue by preserving yield. We will share the results from the live tests conducted to measure the impact of these strategies and challenges faced in implementation.

Vendor Presentations

Balancing long-term passenger value and operational costs in airline operations

Bo Vaaben - Carmen Systems

Decisions taken on the day of operation to recover from disruptions are often taken with little consideration of passenger impact. Even if the ambition is to put the passenger in focus, no systems are available for customer driven decisions. Thereby, operational decisions may work against the company policies regarding prioritized passengers. This presentation describes how optimization is used at Aeromexico to reduce impact of disruptions on passengers.

*Program is subject to change

Call for papers

Come and share with us your ideas, thoughts, current trends, philosophies, and latest technological advances on topics ranging from revenue optimization, O&D control, demand forecasting and more. 

If you are interested in presenting at the study group meeting, please complete the on-line registration form and submit it directly via the www.  As always, talks are subject to approval, and time slots are available on a first come, first serve basis - so if you are interested, act now!

The AGIFORS Reservations & Revenue Management 2004 conference technical program is currently being put together, for more information please contact mailto:gina.morello@aa.com.

• 2003 Meeting - Hawaii, USA

• 2002 Meeting - Berlin, Germany

• 2001 Meeting - Bangkok, Thailand

• 2000 Meeting - New York City, USA

Welcome | Agenda | Registration | General Information | Hotel Information | Host City