Feb 2012

The Newsletter of the

INFORMS Revenue Management and Pricing Section

New book: Pricing Segmentation and Analytics
by Tudor Bodea and Mark Ferguson, Business Expert Press
Pricing analytics uses historical sales data with mathematical optimization to set and update prices offered
through various channels in order to maximize profit. A familiar example is the passenger airline industry, where
a carrier may sell seats on the same flight at many different prices. Pricing analytics practices have transformed
the transportation and hospitality industries, and are increasingly important in industries as diverse as retail,
telecommunications, banking, health care and manufacturing. The aim of this book is to guide students and
professionals on how to identify and exploit pricing opportunities in different business contexts. The first chapter
looks at pricing from an economist's viewpoint, beginning with the basic concept of price elasticity and how it
differs at the product, firm, and industry levels as well as the short term versus long term. Next, the common
assumptions regarding the customer population's willingness-to-pay is discussed along with the price response
curves that result from this assumption. Basic price optimization techniques are then explored with extensions
provided for alternative objective functions and constrained supply. The second chapter looks at these same
topics, but from a more practical standpoint, with examples provided from several consulting projects. The third
chapter is on dynamic pricing, with a special emphasis on the most common application: markdown pricing.
Similar to the first two chapters, both the theory and the application aspects will be covered. The fourth chapter
covers the new field of customized pricing analytics, where a firm responds to a request-for-bids or request-forproposals with a customized price response. In this situation, the firm will only have historical win/loss data and
traditional methods involving price elasticity do not apply. The pricing analytics methodology along with several
case studies are provided. The final chapter covers the relevant aspects of behavioral science to pricing.
Examples include the asymmetry of joy/pain that customers feel in response to price decreases/increases. A
set of best pricing practices are presented that are based on these behavioral responses. Finally, the appendix
contains the details needed to build and implement a pricing analytics system in practice.
Evaluation copies are available from this address:
http://www.businessexpertpress.com/books/pricing-segmentation-and-analytics

Upcoming Workshop
The University Transportation Center (UTC) at the Georgia Institute of Technology will be sponsoring a
workshop focused on integrating discrete choice and other models ground in behavioral theory with revenue
management. The goal of the workshop, to be held in May in Atlanta, Georgia, is to bring together experts from
discrete choice modeling and revenue management, present current research efforts in this area, and
determine future research opportunities. The workshop will review current research, identify future research
opportunities, explore collaborations with industry experts and software vendors to obtain data needed to
successfully implement choice-based revenue management models, and gain perspectives from Associate
Editors of top discrete choice modeling and operations research journals on ways to best position papers for
journals from this inter-disciplinary topic.

Page | 1

To encourage research in the choice-based revenue management area, UTC will be providing up to five full
scholarships to doctoral students interested in attending the workshop. To apply for a scholarship, students
should provide the following information:
1. A CV (no more than two pages). The CV should clearly indicate your program of study, prior
educational background, and expected graduation date.
2. A short personal essay (no more than one page). The essay should address your interest in choicebased revenue management, prior research experiences relevant to revenue management and/or
discrete choice modeling, and your proposed dissertation topic.
3. Nomination letter from your dissertation advisor (this should be submitted electronically directly by the
dissertation advisor to maintain confidentiality of the nomination letters).
Nomination packages should be emailed to Laurie Garrow at [email protected] by 5 PM eastern
time on Friday, March 2, 2012. Preference will be given to those doctoral students who have completed the first
two years of their coursework and are in the process of starting their doctoral research.
Please stay tuned for the exact dates of the conference which will be finalized soon.

News from the Journal of Revenue and Pricing Management

Special Issue on Strategic B2B Pricing
Don't miss the latest issue of the Journal of Revenue and Pricing Management (RPM), which features research
summarizing how firms treat the pricing function in industrial companies as a strategic activity both as an
integral part of firm strategy, and as a valuable, difficult to imitate way to build competitive advantage and to
achieve superior profitability as a result of pricing activities.
These featured articles from the issue are now available free-to-view for a limited time:
The conceptualization of value-based pricing in industrial firms
By Stephan M Liozu, Andreas Hinterhuber, Richard Boland and Sheri Perelli
Optimal pricing models in B2B organizations
By Rafael Farres
Order RPM at a reduced subscription rate!
Members of the INFORMS Revenue Management & Pricing Section receive a 50% reduced subscription rate to
this Journal. For additional details and to place your order, please visit: www.palgrave-journals.com/rpm

RPM on Twitter
Looking for more RPM news and updates? Follow @RevManJournal on Twitter!

Page | 2

Keynote at the 2011 INFORMS annual meeting by Scott Nason

Jointly sponsored by the Aviation Applications Section and the Revenue Management
and Pricing Section
The Airlines Evolving Revenue Models
The airline industry has seen many changes during the past year. The merger of United and Continental
Airlines, forming the worlds largest airline, followed by the acquisition of AirTran Airways by Southwest Airlines,
are major developments that will have significant impacts on airline business models. On another front, airlines
and global distribution systems (GDSs) are suing each other and the Department of Justice is investigating the
air travel distribution business. These may affect the structure of the GDS model that has long been in use. This
dynamic business environment may require changes in the revenue models employed by the airlines. Scott
Nason, former vice-president of revenue management at American Airlines, discussed these and other topics
as part of his keynote presentation at the 2011 INFORMS meeting in Charlotte. In particular, Scott offered his
perspectives on airline consolidation trends, distribution wars, and how revenue management models will need
to evolve.
Why Airlines Merge Or Why Not
I can recall being at a conference more than 20 years ago at the FAA with a group of industry experts. The
consensus of that group was that within a few years, the airline industry would consolidate down to three big
airlines. I wasnt part of that consensus, as I didnt see consolidation happening that extremely or that quickly.
Thats not to say there havent been a lot of mergers: there have been, particularly lately with United and
Continental; Delta and Northwest; US Airways and America West; American and TWA. These mergers have
helped create the very large airlines we have today.
We have also seen a lot of airlines come and go. Some of the largest legacy airlines to disappear are now a few
decades old and include Eastern and Pan Am. There have also been multiple airlines that have tried to provide
a much better product at a competitive price. This includes F or J airlines that offer flights will all business
class service to Europe. There have been three incarnations of this idea, the largest being EOS. All three of
these airlines have disappeared and have followed the tradition of other airlines such as Air One, MGM Grand
Air, and First Air that decided to provide a business class product at a competitive price. Maybe one day the
idea will succeed, but not now.
On the other end, several low cost airlines including Virgin America and JetBlue have emerged and have
succeeded by providing a higher-quality coach product. Spirit and Allegiant Airlines are examples of airlines
that have provided low fares to consumers by introducing a very low cost / bare bones product.
So why do airlines merge, and why havent we seen more mergers? There are three reasons why airlines
merge: network benefits, cost synergies, and to gain competitive power. In some cases, mergers have been
able to achieve one or more of these benefits, although the jury is still out on whether recent mergers are
successful. Successful mergers include America West and US Airways, Delta and Northwest, and United and
Continental. All of these mergers resulted in larger networks (which created competitive power) and cost
reductions. There are lots of examples of failures though, including American Airlines and TWA. American
closed the deal with TWA in spring of 2001, just a few months prior to 9/11. American immediately wished they
hadnt purchased TWA, and quickly eliminated remnants of the old airline. Earlier on, US Airways purchased
several smaller airlines including Piedmont Airlines and Pacific Southwest Airlines, but never achieved network
benefits from these acquisitions. Delta purchased Western Airlines, a smaller west coast airline. From an

Page | 3

operational perspective, this merger was a success, but they were unable to achieve cost synergies due to the
amount of money they had to spend to keep their pilots happy.
There are several reasons why airlines dont merge. One of the most difficult aspects involves labor. Seniority
is very important to airline employees and determines not only their pay but the shifts they work. Pilots in
particular get very adamant about seniority rights. There arent enough good captain jobs to go around and
some employees feel very disadvantaged by the merger. Pilots have a view of what their job trajectories are
going to be through internal growth, and mergers upset that trajectory. Often airlines need to give pilots a lot of
money in mergers to keep them happy.
Another reason why airlines dont merge is due to difficulties associated with systems and process integration.
Most major airlines have a rather unique set of systems, and need to spend hundreds of millions of dollars
trying to make these systems talk to each-other, or replacing one system with another. Process integration is
also difficult. Each airline has a lot of unique rules and procedures regarding how to handle passenger
disruptions, process refunds, etc.
The economies of scale are also not huge in mergers, and some aspects actually get more complex as airlines
grow. Pilot training is one example. For an airline that owns a single aircraft type, pilots go to training once in
their career (to move from co-captain to captain position). For an airline that owns two aircraft types, pilots now
go to training four times over their career as they switch positions and aircraft types. These diseconomies of
scale tend to occur with any employees who have equipment-specific qualifications.
Despite the disadvantages, there are benefits of merging, particularly as they related to corporate travel.
Consider two airlines: one that covers 80% of the travel needs of a corporation and another than handles 50%
of the travel needs of the corporation. This 50% includes the 20% that the first airline does not serve in addition
to 30% overlap. The first airline that can serve 80% of the corporations travel needs offers a 20% discount in
exchange for receiving 80% of the travel. The second airline offers 30% discount in exchange for receiving
50% of the travel. The airline that meets the greatest needs of the corporation tends to win the contract
negotiation. Through building larger networks, airlines are able to get corporate deals that they wouldnt get
otherwise.
Similar advantages are also present when airlines are negotiating fees with GDSs and travel agencies. Being
big is important in these negotiations. Another benefit of merging is that there are some economies of scale
merged airlines dont need two CIOs or presidents, and are able to achieve some cost savings by eliminating
headquarter staff.
So where are we now? Well, we havent gotten down to three airlines. Weve seen a lot of mergers lately, but I
dont think were done yet and that airlines will find opportunities to merge for one or more of the reasons I
stated earlier.
Distribution Wars
Customers can purchase airline tickets through a variety of distribution channels. They can buy direct by going
to the airline website or, for a fee, by calling the airlines call center. Customers can still use traditional travel
agents to book travel, although now they charge you for their service. Travel can also be booked via online
travel agencies. Up until a few years ago, online travel agencies used to charge a small booking fee, but these
fees have been eliminated and customers can book through sites such as Travelocity now without a fee. There
are also still lots and lots of corporate deals. Most medium-sized companies have corporate deals with one or
more airlines in which the corporation commits to providing a certain market share to the airline in exchange for
discounts, VIP memberships, and other benefits. These bookings are made through corporate travel systems.
Off-tariff distributions also exist, often in ethnic and overseas markets. There is some channel-differentiated

Page | 4

pricing, particularly for opaque products in which a customer commits to buying a fare without knowing detailed
airline brand and itinerary characteristics.
In order for a channel to be priced competitively, it is critical for the channel to have every airlines schedule and
fare data (including the lowest fares the airline offers). This ensures that travel agents have all of the information
they need when making bookings. The way this has played out is that the major GDSs such as Sabre,
Worldspan, and Amadeus have signed full content deals with airlines. If airlines commit to providing all of their
schedule and fares (including lowest fares) for their entire network, the GDSs lower the booking fees they
charge travel agents for making bookings through their systems. If the GDSs can get all airlines to sign full
content agreements, they can then approach travel agencies and argue that they should use their system since
they will be able to provide all of the information they need to serve their customers.
But what happens if an airline decides it doesnt want to commit to giving all if its services and fares, but wants
to save some for its own direct channels? What if an airline decides the booking fees being charged are too
high? Currently, were at the point in which airlines and GDSs are arguing over these points and airlines,
GDSs, travel agents, and online travel agencies are trying to decide who needs who more. If an airline pulls out
of an agency, who loses? Is the bigger problem due to the fact the airline loses sales because it is no longer in
that channel, or is the bigger problem due to the fact the agency loses sales because it no longer has that
airlines products in its system? In the short term, the airline will probably lose more business as travel agents
didnt realize that airline was an option. In the long term, airlines will likely be able to turn this around, and direct
travel agencies to other channels that have their full schedules.
The increasing reliance on ancillary revenue streams is also contributing to the distribution wars. Ancillary
revenues refer to fees that airline charge for selling services (baggage fees, ticket exchange fees, early standby
fees, seat reservation fees). Ancillary fees greatly complicate the purchase process. It has been difficult for
generic sites to compare these fees across airlines as there is a lot of fine print associated with ancillary
revenues that vary from airline to airline. The standards for filing fares are well established in the airline
industry, but the standards for filing ancillary fees are in their infancy. Airlines will also need to develop better
databases to keep track of who paid ancillary fees, which flight it was tied to, etc. This information will become
increasingly important from a RM perspective.
All of these factors have given rise to distribution wars. The wars have gone through several phases. In the
early part of last decade, some airlines that tried to pull out of systems for example, Northwest pulled out of
Expedia for a while. These disagreements were mostly put to bed when GDSs signed three-year full content
deals. When these contracts first came up for renewal, a few airlines pushed back and fought hard, but in the
end renewed their three-year deals. All this fell apart about a year ago when American Airlines decided not
renew their contract with Orbitz because Orbitz did not agree to go through Americans direct booking link.
American wanted to have customers make bookings directly on their site so that they could customize
information on fees to each customer. Expedia then told American: if youre not in Orbitz, youre not in Expedia
either. Sabre told American they would raise their booking fees and bias the display against them. The
outcome of the current distribution wars is awaiting court cases. In the meantime, the Department of Justice
decided this was pretty interesting stuff, and is doing an investigation to determine if they are happy with the
model in which GDSs pay travel agents a fee for making bookings on their system. That is, GDSs convince
travel agents to use their system, collect fees from airlines when these bookings are made, and give some of
these fees back to the travel agency as a payment for making the booking in their GDS system. Airlines view
this as effectively paying travel agents to make a booking in a particular distribution channel. Airlines have the
option of not selling to that channel, but if the channel is very large (such as Expedia), this will be a problem.
The Department of Justice is looking at the relationships among travel agencies, GDSs and airlines to
determine if we have a free marketplace on sale and distribution of airline tickets. Looking ahead, American

Page | 5

Airlines has set the stage for the ongoing lawsuits, but other airlines have hinted that they may follow
Americans lead.
Evolving Revenue Management Models
Revenue management models are designed to determine what price to offer to each customer. Prices have to
be set independently of who the customer is. The same price, depending on who the customer is, can be too
low and too high. Its too low if the customer would have paid more. Its also too low if the airline could have
sold it later to a different customer who was willing to pay more. The other way price is too low is if the airline
could have moved the passenger to a less valuable itinerary. Assume the price on two flights that depart at
noon and 5 pm are both $300, but the 5 pm flight is more valuable. The $300 fare is too low if, by raising the
fare, the customer would have purchased the noon flight that wasnt worth as much to the airline. On the other
hand, the fare is too high if the customer doesnt buy and, in retrospect, the airlines end up sorry they didnt buy
a ticket on one of their flights. That is, the fare is only too high if they didnt buy from the airline and that seat
ended up not being sold.
As in many industries, airlines cant price individually for each customer yet. The optimal price for a particular
flight or itinerary is not the same for every customer, but the customer has to post the same price for each
customers. In order to offer different prices to different customer segments, airlines have created different
products by using fare rules and have offered channel-specific prices. Fare rules refer to restrictions that must
be met by the consumer in order to purchase a lower fare. Lower fares are offered to customers who buy a
round-trip ticket, purchase 14 days in advance of departure, and stay over a Saturday night. These restrictions
are designed to try to make prices higher for business customers who are willing to pay more than leisure
travelers who will not purchase if the fare is too high. Airlines are also willing to sell their inventory for deeply
discounted (and unpublished fares) through opaque distribution channels, such as Priceline or Hotwire.
Customers using these channels are more cost-conscious than other travelers. Coupons are also used to try to
create different prices for different people.
Airlines first began to price discriminate more than 30 years ago. Fare restrictions such as minimum/maximum
stay rules and advance purchase requirements were used as a way to try to get price-inelastic business
customers to pay more than price-elastic leisure customers. Yield management became a science designed to
try to optimize price controls using the fare restrictions and inventory controls to maximize revenues. Yield
management started out in the early 1980s and has evolved into an elaborate science over the past 30 years.
Yield management quantifies the trade-off between selling a lower-priced seat in advance or saving that seat for
a potential customer who is willing to pay a higher price.
The first yield management models were leg-based. Initial efforts were focused on developing demand forecast
models to understand how much customers were willing to pay for each booking class. Historical data was
used as a basis for these forecasts, and were used to optimize inventory controls. Over time, the optimization
algorithms also got better and began to incorporate key network linkages in the system. For example, a flight
from Savannah to Charlotte may be used to get people from Savannah to London or Savannah to Mexico. If I
sell the Savannah-Charlotte flight to a local passenger for $100, I wont be able to sell a more valuable $700
itinerary from Savannah to London through Charlotte. I need to save inventory not only for the highest classes,
but also for the more valuable itineraries. However, network optimization algorithms currently do not determine
the absolute best use of the seat on a flight. In the Savannah to London example, I didnt consider that there
was an alternate path available from Savannah to London through Philadelphia. So I could have sold the
Savannah-Charlotte leg to a local passenger, and accommodated the international passenger through
Philadelphia instead. Incorporating these network-level interactions are in the early stages of development.
Early demand models estimated how many people want to book a fare class using historical data. These
models basically took demand as saying: this is a person who buys full fare Y inventory or this is a person why

Page | 6

buys a restricted V class inventory. In reality, though, people buy the cheapest seat they can. A person willing
to purchase Y will buy V if it is cheaper and available. Demand models began to evolve to bring more complex
interactions among booking classes, and recognize that some customers who bought V would have purchased
Y.
Airlines have spent a lot of time improving demand forecasts. Demand forecasts are the heart of a RM
system and the optimization models are heavily reliant on the accuracy of the demand model. Airlines
improved their demand forecasts by incorporating passenger characteristics and detailed seasonality
characteristics: When did the passenger book? What month? What day of the week? What time of day? The
better an airline gets at understanding which types of customers are purchasing its flights, the better the airline
gets as forecasting demand for its flights.
Demand forecasts are still based primarily on history, although how airlines slice and dice history has gotten
more sophisticated. The focus is still on ensuring airlines stop selling cheap seats in order to save for high-yield
late arriving customers (along with other subtleties mentioned earlier). Demand models have also gotten better
at overbooking and understanding passenger no show and cancellation behavior. No show models account for
both passenger behavior as well as operational no shows caused with passengers misconnects due to flight
delays or flight cancellations.
To date, price elasticity hasnt been explored too much. Airlines havent gotten good at recognizing price
elasticity or modeling the passenger purchase decisions. How customers decide to purchase a certain ticket,
and make tradeoffs among price, carriers, time of day, day of week, etc. is a complicated process. However, as
the passenger choice process changes (due to availability of products, or increased visibility of fares in the
market), existing demand models have a hard time catching up based on history.
Other behavioral aspects of the customer booking prices are also not captured in the demand models. Today,
many customers go to metasearch engines and online travel agencies to do their initial searches, then book
directly on airline websites. Airlines also have lots and lots of data about their customers and detailed flight
histories for their frequent flyer customer databases. This information has not been used extensively in the RM
process.
Demand for airline flights also relies, at least on part, on competition and who is flying itinerary. Current
demand models dont really capture this information. To the extent that future competition is the same as
historical competition, my models are fine but what if another airline adds a nonstop flight, or cancels a flight.
What my competitors are doing is relevant, but to date there has not been a lot of attention paid to what my
competitors are doing. There are multiple aspects of this competition: what schedules are my competitors flying
(and is it competitive to mine), what fares have been filed. However, there is a distinction between fares that
are filed (which may potentially be sold) and what fares are actually available. Inventory controls decide which
fares are available for particular itineraries. Airlines have historically monitored what flights competitors offer,
and what fares they have filed, but only now are they monitoring whether competitors are offering particular
fares (that is, a filed fare is irrelevant if no one can buy it). Airlines are developing systems based on online
screen scrapes and other sources that automatically collecs information about selling fares on specific
itineraries and dates. Airlines are beginning to ask what they can do with this competitive information.
Looking ahead, what are airlines trying to do to solve these problems and where are future RM models headed?
Airlines will begin to develop behavioral demand models that are less reliant on history. I cant tell you how
soon, but some in the audience will see choice models adopted as the basis of RM demand forecasts during
their lifetimes. Airlines will continue to monitor competitive information and get a better sense of customers
price elasticities. How airlines use competitor price information remains to be seen. What is one airline

Page | 7

monitors competitive prices and offers a price that is $10 lower than the competitor? What if two airlines do
that? Is it beneficial?
Airlines will begin to personalize prices and offers. To date, personalization has focused on showing customers
different information when they visit a website or customizing emails to different customers, but there is a lot of
potential for using data from frequent flyer databases to customize offers to various people. Airlines will
continue to aggressively merchandise and sell not only air travel, but other components of a trip.
At least one airline is going to try to personalize price by taking into account who you are when they offer you a
price. I think airlines will fall into a trap, though, as it is their best customers who are probably willing to pay the
most. Charging your best customers higher prices is not sustainable, although bland economic theory will tell
you that is the way to go. I think personalized pricing is headed for some big mistakes before airlines get it
right.
We will also begin to see airlines differentiate price in terms of where you buy a ticket. Today, differences in
prices across channels are caused by accident or system glitches. I think differentiation across channels will
become more intentional in the future.

Scott Nason is currently a freelance consultant specializing in revenue management, IT systems, operations,
and customer relationship management for the airline industry. He was formerly the vice president of Revenue
Management at American Airlines, where he worked for almost 30 years at various positions, including chief
information officer and vice president of Operations Planning and Performance. Nason holds MS and BA
degrees from Massachusetts Institute of Technology and University of California, Berkeley, respectively.

Page | 8