Springer the theory and practice of revenue management 2005 ISBN1402077017

4.7 Notes and Sources 168Part II Price-based RMIntroduction and Overview 5.1.1 Price versus Quantity-Based RM 5.1.2 Industry Overview 5.1.3 Examples of Dynamic Pricing 5.1.4 Modeling Dyn

Revenue Management

INTERNATIONAL SERIES IN

OPERATIONS RESEARCH & MANAGEMENT SCIENCE

Frederick S Hillier‚ Series Editor‚Stanford University

Ramík, J & Vlach, M / GENERALIZED CONCAVITY IN FUZZY OPTIMIZATION

AND DECISION ANALYSIS

Song, J & Yao, D / SUPPLY CHAIN STRUCTURES: Coordination, Information and

Optimization

Kozan, E & Ohuchi, A /OPERATIONS RESEARCH/MANAGEMENT SCIENCE AT WORK

Bouyssou et al /AIDING DECISIONS WITH MULTIPLE CRITERIA: Essays in

Honor of Bernard Roy

Cox, Louis Anthony, Jr /RISK ANALYSIS: Foundations, Models and Methods

Dror, M., L'Ecuyer, P & Szidarovszky, F / MODELING UNCERTAINTY: An Examination

of Stochastic Theory, Methods, and Applications

Dokuchaev, N /DYNAMIC PORTFOLIO STRATEGIES: Quantitative Methods and Empirical Rules

for Incomplete Information

Sarker, R., Mohammadian, M & Yao, X /EVOLUTIONARY OPTIMIZATION

Demeulemeester, R & Herroelen, W /PROJECT SCHEDULING: A Research Handbook

Gazis, D.C /TRAFFIC THEORY

Zhu, J /QUANTITATIVE MODELS FOR PERFORMANCE EVALUATION AND BENCHMARKING

Ehrgott, M & Gandibleux, X / MULTIPLE CRITERIA OPTIMIZATION: State of the Art Annotated

Bibliographical Surveys

Bienstock, D /Potential Function Methods for Approx Solving Linear Programming Problems

Matsatsinis, N.F & Siskos, Y /INTELLIGENT SUPPORT SYSTEMS FOR MARKETING

DECISIONS

Alpern, S & Gal, S./ THE THEORY OF SEARCH GAMES AND RENDEZVOUS

Hall, R W./ HANDBOOK OF TRANSPORTATION SCIENCE - Ed.

Glover, F & Kochenberger, G.A / HANDBOOK OF METAHEURISTICS

Graves, S.B & Ringuest, J.L / MODELS AND METHODS FOR PROJECT SELECTION:

Concepts from Management Science, Finance and Information Technology

Hassin, R & Haviv, M./ TO QUEUE OR NOT TO QUEUE: Equilibrium Behavior in Queueing

Systems

Gershwin, S.B et al/ANALYSIS & MODELING OF MANUFACTURING SYSTEMS

Maros, I./ COMPUTATIONAL TECHNIQUES OF THE SIMPLEX METHOD

Harrison, T., Lee, H & Neale, J./ THE PRACTICE OF SUPPLY CHAIN MANAGEMENT: Where

Theory And Application Converge

Shanthikumar, J.G., Yao, D & Zijm, W.H./ STOCHASTIC MODELING AND OPTIMIZATION

OF MANUFACTURING SYSTEMS AND SUPPLY CHAINS

Nabrzyski, J., Schopf, J./GRID RESOURCE MANAGEMENT: State of the Art

and Future Trends

Thissen, W.A.H & Herder, P.M./CRITICAL INFRASTRUCTURES: State of the Art in Research

and Application

Carlsson, C., Fedrizzi, M., & Fullér, R./FUZZY LOGIC IN MANAGEMENT

Soyer, R., Mazzuchi, T.A., & Singpurwalla, N.D./ MATHEMATICAL RELIABILITY: An

Expository Perspective

* A list of the early publications in the series is at the end of the book *

REVENUE MANAGEMENT

KALYAN T TALLURI

Department of Economics and Business

Universitat Pompeu Fabra

Barcelona

GARRETT J VAN RYZINGraduate School of Business Columbia University New York

KLUWER ACADEMIC PUBLISHERS

NEW YORK, BOSTON, DORDRECHT, LONDON, MOSCOW

Print ISBN: 1-4020-7701-7

Print © 2004 Kluwer Academic Publishers

All rights reserved

No part of this eBook may be reproduced or transmitted in any form or by any means, electronic, mechanical, recording, or otherwise, without written consent from the Publisher

Created in the United States of America

Boston

©200 5 Springer Science + Business Media, Inc.

Visit Springer's eBookstore at: http://ebooks.kluweronline.com

and the Springer Global Website Online at: http://www.springeronline.com

for the love and joy‚

K.T.

To Mary Beth‚ Stephanie‚ Claire and Andrea with love and thanks‚ and to the memory of my father John R van Ryzin‚

G.V.

The Origins of RM

1.2.1

1.2.2

Airline HistoryConsequences of the Airline History

A Conceptual Framework for RM

Industry Adopters Beyond the Airlines

An Overview of a RM System

The State of the RM Profession

Chapter Organization and Reading Guide

1.6.1 Chapter Organization

1.6.2 Reading Guide

Notes and Sources

Notes and Sources

27272832333536414450505256575859626264758181828387888990919192939598

3.4.5 Iterative Decomposition Methods

Stochastic Gradient Methods

3.5.1 Continuous Model with Gradient Estimates

3.5.2 Discrete Model with First-Difference EstimatesAsymptotic Analysis of Network Problems

3.6.1 Asymptotic Optimality of Partitioned Controls3.6.2 Asymptotic Optimality of Bid-Price Controls

3.6.3 Comments on Asymptotic Optimality

Decentralized Network Control: Airline Alliances

Notes and Sources

Group Cancellations4.3 Dynamic Overbooking Models

4.3.1

4.3.2

Exact ApproachesHeuristic Approaches Based on Net Bookings4.4 Combined Capacity-Control and Overbooking Models

4.5.1

4.5.2

Model and FormulationJoint Optimal Overbooking Levels4.6 Network Overbooking

100101102103107108111112116118118120120121122129130131135137138139147149150152152154155156158159160161162164166

4.7 Notes and Sources 168Part II Price-based RM

Introduction and Overview

5.1.1 Price versus Quantity-Based RM

5.1.2 Industry Overview

5.1.3 Examples of Dynamic Pricing

5.1.4 Modeling Dynamic Price-Sensitive Demand

Single-Product Dynamic Pricing Without Replenishment5.2.1 Deterministic Models

5.2.2 Stochastic Models

Single-Product Dynamic Pricing with Replenishment

5.3.1 Deterministic Models

5.3.2 Stochastic Models

Multiproduct‚ Multiresource Pricing

5.4.1 Deterministic Models Without Replenishment

5.4.2 Deterministic Models with Replenishment

Introduction and Industry Overview

6.1.1 An Overview of Auctions in Practice

6.3 Optimal Dynamic Single-Resource Capacity Auctions

6.3.1

6.3.2

6.3.3

FormulationOptimal Dynamic Allocations and MechanismsComparisons with Traditional RM

6.4 Optimal Dynamic Auctions with Replenishment

6.4.1 Dynamic Programming Formulation

6.4.2 Optimal Auction and Replenishment Policy

6.5.3 Relationship to Traditional Auctions

6.5.4 Relationship to Traditional Network RM

6.5.5 Revenue Maximization and Reserve Prices

6.6 Notes and Sources

Part III Common Elements

7 CUSTOMER-BEHAVIOR AND

MARKET-RESPONSE MODELS

7.1

7.2

The Independent-Demand Model

Models of Individual Customer Choice

7.2.1

7.2.2

7.2.3

Reservation-Price ModelsRandom-Utility ModelsCustomer Heterogeneity and Segmentation7.3 Models of Aggregate Demand

7.4 Notes and Sources

254257259262266272272274278280281282284285288289290291291293294

301301303303304308310311320321327330330

8.2.1 Perfectly Competitive Markets

8.2.2 Firm-Level Decisions Under Perfect Competition8.2.3 Precommitment and Demand Uncertainty

8.2.4 Peak-Load Pricing Under Perfect Competition

8.2.5 Identifiable Peak Periods

8.2.6 Uncertainty over the Timing of Peak Loads

8.2.7 Advance Purchases in Competitive Markets

333333336336338338341341343345349350351352363369372375376388395402407407408410412415419420422425427428433434439

8.3 Monopoly Pricing

8.3.1 Single-Price Monopoly

8.3.2 Monopoly with Capacity Constraints

8.3.3 Multiple-Price Monopoly and Price Discrimination8.3.4 Strategic Customer Behavior

8.3.5 Optimal Mechanism Design for a Monopolist

8.3.6 Advance Purchases and Peak-Load Pricing der Monopoly

Un-8.4 Price and Capacity Competition in an Oligopoly

8.4.1 Static Models

8.4.2 Dynamic Models

8.4.3 Product Differentiation

8.5 Notes and Sources

9 ESTIMATION AND FORECASTING

9.1 Introduction

9.1.1 The Forecasting Module of RM Systems

9.1.2 What Forecasts Are Required?

9.2.4 Method of Moments and Quantile Estimators

9.2.5 Endogeneity‚ Heterogeneity‚ and Competition

9.3 Forecasting Methods

9.3.1 Ad-Hoc Forecasting Methods

9.3.2 Time-Series Forecasting Methods

Other MethodsCombining Forecast Methods9.4 Data Incompleteness and Unconstraining

9.4.1 Expectation-Maximization (EM) Method

9.5.2 Forecasting Errors and System Control

9.6 Industry Models of RM Estimation and Forecasting

9.6.1 Airline No-Show and Cancellations Forecasting9.6.2 Groups Demand and Utilization Forecasting

9.6.3 Sell-Up and Recapture Forecasting

9.6.4 Retail Sales Forecasting

Media and Broadcasting

10.5.1 Customers‚ Products‚ and Pricing

10.5.2 RM Practice

Natural-Gas Storage and Transmission

10.6.1 Customers‚ Products‚ and Pricing

10.8.1 Customers‚ Products‚ and Pricing

10.8.2 Capacity Management and Base-Price Setting

10.8.3 RM Practice

Casinos

10.9.1 Customers‚ Products‚ and Pricing

10.9.2 RM Practice

Cruise Ships and Ferry Lines

10.10.1 Customers‚ Products‚ and Prices

Theaters and Sporting Events

10.14.1 Customers‚ Products‚ and Pricing

10.14.2 Ticket Scalping and Distribution

10.14.3 RM Practice

Manufacturing

533534541542543545546547550551552554554555556556558559559559560560561561561563563563563564565566567567567571574

10.15.1 Customers‚ Products‚ and Pricing

11.2.4 Retail Management Systems

Revenue-Opportunity Assessment and Revenue-BenefitsMeasurement

11.3.1 Revenue-Opportunity Assessment

11.3.2 Revenue-Benefits Measurement

RM Simulation

11.4.1 Generating Aggregate Number of Customers

11.4.2 Generating the Customer-Arrival Pattern

Customer Perceptions and Reactions

11.5.1 RM Perception Problems

11.5.2 Managing Perceptions

11.5.3 Overbooking Perceptions

Cultural‚ Organizational‚ and Training Issues

11.6.1 Changes in Responsibility by Function

11.6.2 Project and Organizational Structure

11.6.3 Training

Notes and Sources

574575576579579580585594594594598605

608608610611613613614614618619620620623627628631631635643651657667

Index

671709

Optimal protection level in the static model

Monte Carlo estimates of optimal protection levels

for 50 simulated data points

Adaptive-method example 1

Adaptive-method example 2

Optimal protection level in the dynamic model

Scatter plot of Q(S) and R(S)

Network examples

Comparison of DLP‚ RLP‚ and dynamic

program-ming decomposition

A network example

Stochastic gradient calculation for nested booking

limits: gradient equal to zero

Stochastic gradient calculation for nested booking

limits: gradient nonzero

Overbooking notification statement

Overbooking limits over time

Overbooking for the multiclass and binomial models

Network overbooking

Sample price path at a discount air carrier

11192939445455616982100115127127134140165168181

A six-node‚ two-hub airline network

Optimal price-skimming solution for myopic customersPerturbing the bid in a second-price auction

Illustration of the direct-revelation mechanism

Optimal allocations in the dynamic auction model

Optimal allocations in the dynamic-auction model

with replenishment

Individual demand and the aggregate demand

Revenue and marginal-revenue curves

Some common aggregate price-response functions

Revenue from selling a product at multiple prices

The equilibrium bipartite graph

Response functions for a duopoly RM game

The two cases for the function

Best response cycle example

No-purchase probabilities causing a best-response cycleForecasting module in a RM system

Wedge-shaped bookings data

Time series components

Exponential smoothing with different smoothing

parameters

Sample ACF and PACF functions

ACF and PACF examples

The hierarchial Bayes model

Kalman filter smoothing

Neural network example

Neural network activation functions

Incremental booking data

Over-fitting example

Booking curve with cancellations and no-shows

Cancellation probabilities as a function of booking time

196198200204218225249257275284311319323355382383400400401409416435437443446454463465467471495500501

Induction tree on cancellations data

Neural network for gas-load forecasting

Pricing an air travel itinerary

Revenue sources and revenue drivers for a hotel

A rental car RM system implementation

Store type breakdown for the top 200 retailers

Growth of department store markdowns

Gas pipeline network

Electricity industry structures

Capacity planning at a tour operator

Purchase plan and price setting at a tour operator

RM process for tour-operators

Washington Opera Kennedy Center layout

Advance-purchase and max-stay restrictions for an

airline trip represented by a grid

Products and customers utility reduction modeling

Nightly batch processing

Forecasting under the independent-class model

RM process flow

Quantity-based RM user interface

Quantity-based RM user interface

Static single-resource model and protection levels

Revenue performance for Example 2.3

Static single-resource model data for Example 2.4

Revenue performance for Example 2.4

Fares, demand statistics, and protection levels for

adaptive-method numerical examples

Starting values of protection levels for

adaptive-method numerical examples

Fare-product revenues and restrictions for Example 2.5Segments and their characteristics for Example 2.5

Choice probabilities probability of purchase

Q(S), and expected revenue R(S) for Example 2.5

Illustration of nested policy for Example 2.5

The different segment choices in Example 2.5 if

all classes are open and resulting demand for a

population size of 20

Inputs to the EMSR-b model

Protections for the EMSR-b model without and

with buy-up factors

Simulation results comparison between choice

dy-namic program and EMSR-b with buy-up

Example of a bid price table for a single resource

based on remaining time and remaining capacity

Problem data for the bid price counterexample

Data for the iterative proration-method example

48494950535365656672

747575759090111

Data for the Williamson [566] network example

Initial protection levels produced by DAVN

Improved protection levels produced by the

sto-chastic gradient algorithm

U.S major airline denied-boarding rates‚ 1990-2000

Binomial and normal approximation overbooking

probabilities

Comparison of normal and Gram-Charlier (G-C)

approximations

Empirical distribution of group sizes

Allocations of capacity between periods 1 and 2

and the marginal values and total revenue

Example of the marginal-allocation algorithm

Example of discrete prices and revenues

Solution of a linear program for the discrete-price

Demand-function parameters‚ itineraries‚ and

op-timal solution for Example 5.6

Empirical generalizations on promotions

Dynamic auction revenues for different

concentra-tions of customers

DLPCC suboptimality gaps relative to a dynamic

auction for different demand to capacity ratios

Dynamic auction and replenishment profits for

dif-ferent numbers of customers

Dynamic auction and replenishment profits for

dif-ferent holding costs

Network auction simulation results: average

rev-enues as a function of reserve price

Attribute weights for attributes in

alternative

Estimated elasticities (absolute values) for

com-mon products

111115116116133143149150190192195195200206219231280280287288294311314

Common demand functions

Prices and capacities for Example 8.2 without an

advance-purchase market

Prices and capacities for Example 8.3 with an purchase market

advance-Revenue and variance calculations for Example 8.9

with a single-price policy

Revenue and variance calculations for Example 8.9

with a multiple-price policy

Assumptions of ordinary least-squares (OLS) estimationMeans and covariances of some stationary time-

series processes

Results of the AR(2) forecasting example

EM algorithm iterations on constrained data

An example of airline fare codes‚ classes‚ and their

restrictions

Features of a hotel property management system

(PMS)

World’s top 10 retailers‚ store types‚ and their

rev-enues for the year 2002

U.S Apparel sales by channel

U.S apparel sales by category

Inventory definitions in television‚ radio‚ and print

media

A sample advertising purchase plan

An example of a pipeline delivery contract

Sample pipeline tariffs

Sample natural gas transportation and storage productsAmtrak accommodation and fare types

Sample freight product differentiation

An example of ticket categories for a broadway show

(continued) An example of ticket categories for a

broadway show

Washington Opera Kennedy Center pricing (2003–

2004 season)

The Mets four-tier pricing plan (year 2002)

Customer segments and subsegments by industry

Classification of segment bases

322347349361361424445448478522528534537538544544547548549562565568569573574581582

Some common segment bases used in RM

(continued) Some common segment bases used in RMAttributes and their levels for a hotel application

Major global distribution systems (GDSs) as of 1998

An availability request message as software code

and the same request as a message

Typical data tables provided by a hotel PMS

Table BIDPRICE

RMS pricing and inventory functions

Functionality of EDI for the travel and tourism

industries

Commonly tracked RM system performance measuresTask list for a RM implementation

583584586601602602603607609612624

Revenue management (RM) has gained attention recently as one ofthe most successful application areas of operations research (OR) Thepractice has grown from its origins as a relatively obscure practice among

a handful of major airlines in the post-deregulation era in the U.S (circa1978) to its status today as a mainstream business practice with a grow-ing list of industry users from Walt Disney Resorts to National CarRental and a supporting industry of software and consulting firms Ma-jor airlines‚ hotel chains‚ and car rental companies have large staffs ofdevelopers and analysts working on RM‚ and major consulting and soft-ware firms also employ large numbers of RM professionals

There are now several major industry RM conferences each year: TheAirline Group of the International Federation of Operational ResearchSocieties (AGIFORS) sponsors an annual reservation and yield manage-ment conference that attracts has attracted up to 200 professionals‚ andThe International Air Travel Association (IATA) hosts an annual RMconference that has drawn up to 800 attendees in recent years TheProfessional Pricing Society also hosts professional conferences that ad-dress science-based pricing methods and technologies and general pricingstrategy

Over this same period‚ academic and industry research on the ology of RM has also grown rapidly The number of published papers

method-on RM has increased dramatically in the last ten years INFORMS‚ theleading professional society of OR‚ has started a Pricing and RM Section‚which has now hosted several annual conferences on RM‚ each drawing

in excess of 100 researchers and professionals And several universitiesnow offer specialized RM courses‚ at both the M.B.A and Ph.D levels.Despite this explosion of both professional and scholarly activity‚ nobook has comprehensively covered the field of RM For any area in such

a mature state of development and with such widespread industry usage‚

such a reference is desirable However‚ for RM the need is particularlyacute for several reasons:

RM is very much a professional practice and as such there is a siderable amount of “institutional” knowledge surrounding it that isrelatively inaccessible to those outside the profession

con-Many of the early and even some more recent seminal ideas do notappear in published journals Even those that have been publishedsometimes appear in relatively obscure sources such as AGIFORSproceedings‚ industry newsletters‚ and standard industry practice.The terminology‚ concepts‚ and notation have not been standardized

to date‚ so it is often confusing for an outsider to reconcile the variouscontributions of the extant literature

There is often a considerable gap between practitioners and emics in the field Academics are often not aware of the real worldcomplexities faced by practitioners of RM‚ and industry practition-ers are often not aware of the more recent advances in the academicliterature

acad-Our aim in writing this book is to meet this need The book seeks—

as its title indicates—to cover both the theory and the practice of RM.

Fundamentally‚ RM is an applied discipline; its value and significanceultimately derive from the business results it achieves At the sametime it has strong elements of an applied science‚ and the technicalelements of the subject deserve rigorous treatment Both these practicaland theoretical elements of the field reinforce each other‚ and to a largeextent this is what makes the topic exciting It is this constructiveinterplay of theory and practice that we have strived to capture in thisbook

Audience

We have two primary audiences in mind for this book—(1) analyticallytrained (or at least “analytically tolerant”) practitioners in industry and(2) academic researchers and teachers We view our core reader as some-one who has the equivalent of a master’s degree or higher in a technicalsubject such as engineering‚ operations research‚ statistics‚ or economics.However‚ significant portions of the text are accessible to general orbusiness readers‚ particularly the introduction‚ Chapter 10 on indus-try profiles‚ and Chapter 11 on implementation issues In addition‚ theintroductions to the technical chapters provide high-level overviews of

each chapter‚ which are designed to provide a qualitative ing of the main topics covered and their business context‚ and give thereader a sense of the essence—if not the details—of the material.For experienced practitioners this book serves as a single-source ref-erence for the major theory and application issues involved in RM Thekey technical results in the field are organized and presented preciselyand in consistent notation‚ so that practitioners can easily refer to rele-vant models‚ formulas‚ and algorithms as needed For new employees inthe RM industry our book also serves as a useful primer on the subject‚allowing them to “get up to speed” on the details of the field quicklythrough a consistent presentation of the material For the technicallyoriented user it serves as an unbiased‚ noncommercial source for un-derstanding the competing methodologies available for RM and theirrelative strengths and weaknesses.

understand-We view the academic audience for the book as consisting of the manyresearchers now working on various RM-related topics‚ as well as thosewho work in related areas (such as supply-chain management)‚ who maywant a single-source‚ accessible overview of the main theory and practicecomponents of the field Academics who teach management science oroperations management courses may also find the book useful‚ eitherdirectly as a supplementary text or simply for the instructor’s personaluse as a reference on the subject Our experiences with colleagues outsidethe field has suggested that most are curious about RM but perhaps notconfident enough about the theory and practice to introduce the subject

in their classes This book should help “demystify” the subject for them.Finally‚ a growing number of courses have specifically focused on RM.Though not designed particularly as a textbook‚ the book should serve as

a useful reading and reference in such courses While we have not put inhomework exercises‚ we did include many small‚ technically unclutteredexamples throughout the book that illustrate the core concepts beingdiscussed

We forewarn the reader that the material in some places in the bookhas an airline bias This is as it should be in our opinion; airline RMpractice remains an important topic in its own right In addition‚ a largenumber—indeed the vast majority—of RM practitioners and researchersworking in the field today are involved directly in airline RM practices

So airline RM is deserving of rigorous and careful coverage‚ which is one

of our goals in writing this book

At the same time‚ not every industry is like the airline industry and

“airlinelike” conditions are not‚ in our view‚ that necessary to apply

RM ideas Therefore‚ we have attempted to present RM in as genericterms as possible and included several topics and chapters that generalize

beyond the airline industry We have tried to be somewhat forwardlooking in this regard‚ while at the same time not venturing too far intothe realm of pure speculation.

Content and Style

As for the choices of material‚ we have aimed for an applied technical(engineering) level in our treatment of the subject For example‚ we havechosen to present all problems in discrete time This eliminates severaltechnical complications‚ while still allowing us to address a wide range

of problems in a simple‚ yet rigorous way Moreover‚ continuous-timemodels and methods are not frequently used in practice‚ so the focus ondiscrete-time methods is well justified from a practical standpoint.Similarly‚ we have not included a large number of proofs This isboth consistent with the applied orientation of the field and reflects ourview that RM models and theory do not share enough in common tojustify a highly formalistic‚ deductive approach to the subject In afew cases we provide proofs of the theoretical results‚ but even these arerelegated to appendices When proofs are omitted‚ we provide references

to the original sources and if possible give either informal arguments orintuition about the results

In addition‚ the bodies of each chapter do not contain a large number

of literature references This is because we want the reader to “see thematerial for what it is” and not be sidetracked by a lot of discussion ofthe literature Where ideas are strongly associated with specific papersand people‚ we‚ of course‚ point this out Detailed references to theliterature and a discussion of sources are collected in a Notes and Sourcessection provided at the end of each chapter To further assist the reader‚appendices containing basic results on probability theory‚ continuousoptimization‚ dynamic programming‚ and game theory are provided tomake the technical material in the book as self-contained as possible

We tried to be comprehensive in our coverage of RM‚ covering bothquantity- and price-based RM as well as the supporting topics of fore-casting and economics While we might have risked over-extending our-selves in this regard‚ we believe such a comprehensive approach is nec-essary to fully understand the subject Indeed‚ a key contribution of thebook is to unify the various forms of RM and to link them closely toeach other and to the supporting fields of statistics and economics Thetopics and coverage do‚ however‚ reflect our own personal choices aboutwhat is and is not important to understand RM While we have tried to

be as comprehensive‚ fair‚ and balanced as possible in arriving at thesechoices‚ undoubtedly our choices have resulted in some biases However‚

the benefit to the reader is that the text has a point of view and is notmerely an uncritical inventory of all research results to date in the field.Finally‚ we have also tried to come up with a notation that is genericand consistent across all the chapters Much of this notation will not co-incide with the notation found in the original papers in the field‚ which

is by and large quite inconsistent anyway A summary of notation isprovided in Appendix A for reference The consistency of notation andpresentation‚ we believe‚ makes reading the book much easier than look-ing at the corresponding collection of original-source articles‚ and it alsohighlights the connections among topics

This book has been a long time in the making‚ and writing it wouldnot have been possible without strong support from our institutions‚colleagues and family.

Jointly we would like to acknowledge the following colleagues for theirtime in reading‚ either sections‚ or significant parts of the book‚ andproviding us with valuable feedback: Antonio Cabrales (UPF)‚ JamesDana (Northwestern)‚ Srinivas Bollapragada (NBC)‚ and the graduate

OM class of UPF (2002); Gustavo Vulcano (NYU)‚ Itir Karaesmen (U

of Maryland)‚ Sanne de Boer (MIT)‚ Michael Harrison (Stanford) (anddoctoral students in Mike’s 2002 Ph.D seminar on RM)‚ Costis Maglaras(Columbia)‚ Serguei Netessine (Wharton)‚ Qian Liu (Columbia)‚ YannisPaschalidis (Boston U.) and Andy Philpott (U of Auckland) and thegraduate students of the seminar taught at the Auckland University in

2002 The book has benefited greatly from their comments Certainly‚all remaining errors and obfuscations are our responsibility

Interactions with many industry colleagues over the years‚ especiallythose with Surain Adyanthaya‚ Andy Boyd‚ Sebastian Ceria‚ Ren Curry‚Mark Diamond‚ Kevin Geraghty‚ Craig Hopperstad‚ Bob Philips‚ AnandRao‚ John Salch‚ Barry Smith‚ and Ben Vinod‚ have also greatly bene-fited the book Dr Rama Ramakrishnan of Profitlogic was kind enough

to provide screenshots of markdown pricing software for use in the plementation chapter

im-Kalyan Talluri would like to thank the Department of Economics andBusiness of the Universitat Pompeu Fabra for their support and healthyresearch environment‚ and the Deming Center of Columbia BusinessSchool for funding many trips to New York to work on the book Healso would like to acknowledge that his knowledge of RM‚ and his re-search‚ benefited from his long collaboration with the Pricing and RMdepartment at Iberia airlines‚ specifically working for many years with

Fernando Castejon and Juan Magaz On the personal side‚ the stressand labors of writing a long book like this‚ he would like to acknowledge‚were vastly mitigated by the love and joy of companionship of CristinaFerrer and Uma Talluri Ferrer‚ both of whom no doubt greet this bookwith a big sigh of relief.

Garrett van Ryzin would like to thank Columbia Business School forsupporting this project over many years‚ and in particular the DemingCenter and its director‚ Nelson Fraiman‚ who provided travel funds andresearch support which helped make writing this book possible Sig-nificant portions of this book were written during a sabbatical visit tothe University of Auckland in 2001-2002‚ and the support of the MSISDepartment and especially its then head-of-department‚ Justo Diaz‚ isgreatfully acknowledged A course taught at Auckland also helped im-prove early drafts of the book‚ as did input and discussions with AndyPhilpott of the Engineering Science Department at Auckland Much ofthe content of this book is the result of research collaborations with anumber of colleagues‚ including Guillermo Gallego‚ Aliza Heching‚ ItirKaraesmen‚ Costis Maglaras‚ Siddharth Mahajan‚ Jeff McGill and Gus-tavo Vulcano It has been a privilege to work with such a talented group

of colleagues‚ and this book has benefited greatly from their collectivecontributions Finally‚ writing this book would not have been possiblewithout the patience‚ love and support of Mary Beth‚ Stephanie‚ Claireand Andrea—who generously (if not joyfully) tolerated Dad’s many longhours of isolation in his office Like Cristina and Uma‚ they too verymuch deserve to celebrate the completion of this book

This chapter provides an introduction to the topic of revenue agement (RM) We begin with an explanation of RM and its historyand origins We then provide a conceptual framework for understandingthe objectives of RM, the types of business conditions under which it isapplied, and the ways RM systems work Finally we conclude by giving

man-an outline of the remaining chapters of the book

1.1 What Is “RM”?

Every seller of a product or service faces a number of fundamentaldecisions A child selling lemonade outside her house has to decide onwhich day to have her sale, how much to ask for each cup, and when todrop the price (if at all) as the day rolls on A homeowner selling a housemust decide when to list it, what the asking price should be, which offer

to accept, and when to lower the listing price—and by how much—if nooffers come in A stamp dealer selling on an Internet auction site has toselect the duration of the auction, what reserve price to set (if any), and

so on

And anyone who has ever faced such decisions knows the uncertaintyinvolved You want to sell at a time when market conditions are mostfavorable, but who knows what the future might hold? You want theprice to be right—not so high that you put off potential buyers and not

so low that you lose out on potential profits You would like to knowhow much buyers value your product, but more often than not you mustjust guess at this number

Indeed, it is hard to find anyone who is entirely satisfied with theirpricing and selling decisions Even if you succeed in making a sale,

you often wonder whether you should have waited for a better offer orwhether you accepted a price that was too low.

Businesses face even more complex selling decisions For example,how can a firm segment buyers by providing different conditions andterms of trade that profitably exploit their different buying behavior orwillingness to pay? How can a firm design products to prevent cannibal–ization across segments and channels? Once it segments customers, whatprices should it charge each segment? If the firm sells in different chan-nels, should it use the same price in each channel? How should prices beadjusted over time based on seasonal factors and the observed demand

to date for each product? If a product is in short supply, to which ments and channels should it allocate the products? How should a firmmanage the pricing and allocation decisions for products that are com-plements (seats on two connecting airline flights) or substitutes (differentcar categories for rentals)?

seg-RM is concerned with such demand-management decisions1 and themethodology and systems required to make them It involves managingthe firm’s “interface with the market” as it were—with the objective

of increasing revenues RM can be thought of as the complement of

supply-chain management (SCM), which addresses the supply decisions

and processes of a firm, with the objective (typically) of lowering the

cost of production and delivery.

Other roughly synonymous names have been given to the practice

over recent years—yield management (the traditional airline term),

pric-ing and revenue management, pricpric-ing and revenue optimization, revenue process optimization, demand management, demand-chain management

(favored by those who want to create a practice parallel to supply-chainmanagement)—each with its own nuances of meaning and positioning.However, we use the more standard term revenue management to re-fer to the wide range of techniques, decisions, methods, processes, andtechnologies involved in demand management

estimating demand and its characteristics and using price and capacity control to “manage”

demand) We use the latter consistently and use the shorter demand management whenever

appropriate.

Structural decisions: Which selling format to use (such as postedprices, negotiations or auctions); which segmentation or differentia-tion mechanisms to use (if any); which terms of trade to offer (in-cluding volume discounts and cancellation or refund options); how tobundle products; and so on.

Price decisions: How to set posted prices, individual-offer prices, andreserve prices (in auctions); how to price across product categories;how to price over time; how to markdown (discount) over the productlifetime; and so on

Quantity decisions: Whether to accept or reject an offer to buy;how to allocate output or capacity to different segments, products orchannels; when to withhold a product from the market and sale atlater points in time; and so on

Which of these decisions is most important in any given business pends on the context The timescale of the decisions varies as well.Structural decisions about which mechanism to use for selling and how

de-to segment and bundle products are normally strategic decisions takenrelatively infrequently Firms may also have to commit to certain price

or quantity decisions, for example, by advertising prices in advance ordeploying capacity in advance, which can limit their ability to adjustprice or quantities on a tactical level The ability to adjust quantitiesmay also be a function of the technology of production—the flexibility ofthe supply process and the costs of reallocating capacity and inventory.For example, the use of capacity controls as a tactic in airlines stemslargely from the fact that the different “products” an airline sells (differ-ent ticket types sold at different times and under different terms) are allsupplied using the same, homogeneous seat capacity This gives airlinestremendous quantity flexibility, so quantity control is a natural tactic inthis industry Retailers, in contrast, often commit to quantities (initialstocking decisions) but have more flexibility to adjust prices over time.The ability to price tactically, however, depends on how costly pricechanges are, which can vary depending on the channel of distributionsuch as online versus catalog

Whether a firm uses quantity or price-based RM controls varies evenacross firms within a given industry For instance, while most airlinescommit to fixed prices and tactically allocate capacity, low-cost carrierstend to use price as the primary tactical variable

Firms can also find innovative ways to increase their ability to makeprice or quantity recourse decisions For example, retailers may holdback some stock in a centralized warehouse and then make a mid season

replenishment decision rather than precommit all their stock to stores

up front Some major airlines have experimented with movable tions that allow them to reallocate seats from coach to business cabins

parti-on a short-term basis And other major airlines have recently

experi-mented with a practice called demand-driven dispatch in whichaircraft of different sizes are dynamically assigned to each flight depar-ture in response to fluctuations in demand, and are not precommitted toflights [50] Car rental companies also may reallocate their fleet from onecity to another In terms of pricing, using online channels or advertis-ing products without price (“call for our low price”) provides firms withmore price flexibility All these innovations increase the opportunity forquantity and price-based RM

Broadly speaking, RM addresses all three categories of management decisions—structural, pricing, and quantity decisions We

demand-qualify RM as being either quantity-based RM or price-based RM if it

uses (inventory- or) capacity-allocation decisions or prices as the mary tactical tool respectively for managing demand Both the theoryand practice of RM differ depending on which control variable is used,and hence we use this dichotomy as necessary

pri-1.1.2 What’s New About RM?

In one sense, RM is a very old idea Every seller in human history hasfaced RM-type decisions What price to ask? Which offers to accept?When to offer a lower price? And when to simply “pack up one’s tent”

as it were and try selling at a later point in time or in a different market

In terms of business practice, the problems of RM are as old as businessitself

In terms of theory, at a broad level the problems of RM are not new ther Indeed, the forces of supply and demand and the resulting process

ei-of price formation—the “invisible hand” ei-of Adam Smith—lie at the heart

of our current understanding of market economics They are embodied

in the concept of the “rational” (profit-maximizing) firm, and define themechanisms by which market equilibria are reached Modern economictheory addresses many advanced and subtle demand-management deci-sions, such as nonlinear pricing, bundling, segmentation, and optimizing

in the presence of asymmetric information between buyers and sellers

What is new about RM is not the demand-management decisions themselves but rather how these decisions are made The true inno- vation of RM lies in the method of decision making—a technologically

sophisticated, detailed, and intensely operational approach to makingdemand-management decisions

This new approach is driven by two complementary forces First,scientific advances in economics, statistics, and operations research nowmake it possible to model demand and economic conditions, quantifythe uncertainties faced by decision makers, estimate and forecast marketresponse, and compute optimal solutions to complex decision problems.Second, advances in information technology provide the capability toautomate transactions, capture and store vast amounts of data, quicklyexecute complex algorithms, and then implement and manage highlydetailed demand-management decisions This combination of scienceand technology applied to age-old demand management is the hallmark

of modern RM

And both the science and technology used in RM are quite new Much

of the science used in RM today (demand models, forecasting methods,optimization algorithms) is less than fifty years old, most of the infor-mation technology (large databases, personal computers, Internet) isless than twenty years old, and most of the software technology (Java,object-oriented programming) is less than five years old Prior to thesescientific developments, it would have been unthinkable to accuratelymodel real world phenomena and demand-management decisions With-

out the information technology, it would be impossible to operationalize this science These two capabilities combined make possible an entirely

new approach to decision making—one that has profound consequencesfor demand management

The first consequence is that science and technology now make it

possible to manage demand on a scale and complexity that would be

un-thinkable through manual means (or would require a veritable army ofanalysts to achieve) A modern large airline, for example, can have thou-sands of flights a day and provide service between hundreds of thousands

of origin-destination pairs, each of which is sold at dozens of prices—andthis entire problem is replicated for hundreds of days into the future! Asimilar complexity is found at most large retail chains, which can havetens of thousand of SKUs2 sold in hundreds of stores and over the Webwith prices monitored and updated on a daily basis The sheer scaleand complexity of the decision-making task in these cases is beyond theability of human decision makers And if not automated, the task has

to be so highly aggregated and simplified that significant opportunitiesfor incremental gains—on particular products, at particular locations,

at specific points in time—are simply lost

2 A SKU (stock-keeping unit) is the lowest level at which we identify inventory—such as men’s

Arrow blue Oxford shirts, long sleeves, size medium.

The second consequence of science and technology is that they make

it possible to improve the quality of demand-management decisions The

management tasks that are involved—quantifying the risks and rewards

in making demand-management decisions under uncertainty; workingthrough the often subtle economics of pricing; accurately interpretingmarket conditions and trends and reacting to this information withtimely, accurate, and consistent real-time decisions; optimizing a com-plex objective function subject to many constraints and business rules—are tasks most humans, even with many years of experience, are sim-ply not good at Models and systems are better at separating marketsignals from market noise, evaluating complex tradeoffs, and optimiz-ing and producing consistent decisions The application of science andtechnology to demand decisions often produces an improvement in thequality of the decisions, resulting in a significant increase in revenues

Of course, even with the best science and technology, there will ways be decisions that are better left to human decision makers Modelscan detect only what’s in the data They cannot reason through theconsequences of a demand shock, new technologies, a sudden shift inconsumer preferences, or the surprise price war of a competitor Thesehigher-level analyses are best left to experienced, human analysts Most

al-RM systems recognize this fact and parse the decision-making task, withmodels and systems handling routine demand-management decisions on

an automated basis and human analysts overseeing these decisions andintervening (based on flags or alerts from the system) when extraordi-nary conditions arise Such man-machine interaction offers a firm thebest of both human and automated decision making

The process of managing demand decisions with science andtechnology—implemented with disciplined processes and systems, andoverseen by human analysts (a sort of “industrialization” of the entiredemand-management process)—defines modern RM

1.2 The Origins of RM

Where did RM come from? In short, the airline industry There arefew business practices whose origins are so intimately connected to asingle industry Here we briefly review the history of airline RM andthen discuss the implications of this history for the field

1.2.1 Airline History

The starting point for RM was the Airline Deregulation Act of 1978.With this act, the U.S Civil Aviation Board (CAB) loosened control ofairline prices, which had been strictly regulated based on standardized

price and profitability targets Passage of the act led to rapid changeand a rash of innovation in the industry Established carriers were nowfree to change prices, schedules, and service without CAB approval.Large airlines accelerated their development of computerized reservationsystems (CRSs) and global distribution systems (GDSs), and the CDSbusiness became profitable in its own right The majors developed hub-and-spoke networks, which allowed them to offer service in many moremarkets than was possible with point-to-point service but also madepricing and operations more complex.

At the same time, new low-cost and charter airlines entered the ket Many of these upstarts—because of their lower labor costs, simpler(point-to-point) operations, and no-frills service—were able to profitablyprice much lower than the major airlines These new entrants tappedinto an entirely new and vast market for discretionary travel—families

mar-on a holiday, couples getting away for the weekend, college students iting home—many of whom might otherwise have driven their cars ornot traveled at all It turned out (quite surprisingly to some at the time)that air travel was quite price elastic; with prices sufficiently low, peopleswitched from driving to flying, and demand from this segment surged.The potential of this market was embodied in the rapid rise of People-Express, which started in 1981 with cost-efficient operations and fares

vis-50 to 70% lower than the major carriers By 1984, its revenues wereapproaching $1 billion, and for the year 1984 it posted a profit of $60million, its highest profit ever (Cross [137])

While these developments resulted in a significant migration of sensitive discretionary travelers to the new, low-cost carriers, the majorairlines had strengths that these new entrants lacked They offered morefrequent schedules, service to more city pairs and an established brandname and reputation For many business travelers, schedule convenienceand service was (and still is) more important than price, so the threatposed by low-cost airlines was less acute in the business-traveler segment

price-of the market Nevertheless, the cumulative losses in revenue from theshift in traffic were badly damaging the profits of major airlines

A strategy to recapture the leisure passenger was needed Yet, for themajors, a head-to-head, across-the-board price war with the upstartswas deemed almost suicidal; with their much lower costs, airlines likePeopleExpress could earn a profit at the new low prices, while mostmajors would lose money at a staggering rate

Robert Crandall, American Airline’s vice president of marketing at thetime, is widely credited with the breakthrough in solving this problem

He recognized that his airline was already producing seats at a marginalcost near zero because most of the costs of a flight (capital costs, wages,