Technical Program
Revenue
Management Presentations ( Based on submitted
abstracts to date* )
Technical Presentation
Theory
vs Practice?
Byron Gamble, Soufiane Ahallal,
and Sunny Ja – American Airlines
This
presentation will explore the complexity of real-world Revenue Management
systems that is sometimes overlooked in academic literature. We aim to test
the legitimacy of certain academic assumptions using practical airline data
and to show the consequences that may occur when they do not hold. We will
focus on three topics. First, we examine the assumption that demand for
lower fares arrives earlier than demand for higher fares. We will also
discuss the flaws of forecasting demand by RBD class. Finally, we highlight
the impact of schedule changes on the forecasting and optimization
processes.
Measuring
Performance of an Airline’s Fare Class Realignment
Goutham
Komirishetty – Etihad Airways
Fare
class realignment (FCR) could be undertaken to address one or more of the following
strategies, to introduce or remove fare levels to match market, to readjust
fare levels across cabins, to restructure the sell-up between the classes.
An Airline migrating from a leg based to an O&D based RM solution would
also engage in fare class realignment. As the current commercial RM systems
forecast at a class level rather than at a price point level, demand
adjustments are required to enable the RM system to optimize with regards
to the new fare levels. Every airline goes through FCR in its lifetime, may
be multiple times to adhere to the market conditions, and measuring the
effectiveness of such an exercise has always been a big concern in the
industry. This paper outlines a method utilizing market theory of
price-demand relationship and peak detection algorithms to effectively
measure the performance of FCR.
Unlocking
the value from Origin and Destination Revenue Management
Sumala, T., Goda
Doreswamy, Mukund
Shankar, Vamsidhar Kodam
and Sergey Shebalov – Emirates and Sabre Airline
Solution
“O&D
RM provides incremental revenue benefits compared to Leg RM” – the
statement has been proven through many simulation studies and many airlines
have implemented and reported the benefits from O&D RM. The key value
driver of this benefit is the network optimization model that uses O&D
forecasts and market representative fares to do effective trade-offs. There
are a multitude of factors that are at play in ensuring the proposed
benefits from O&D RM are materialized. Some of the key challenges that
airlines face today are the complexity of forecast management, maintaining
accurate fares for optimization and advanced availability processors to
achieve this goal. The other major challenge is the big organizational
change from Leg mode to O&D-mode among the managers and RM analysts.
Additionally, distribution aspects such as seamless availability, married
segments and journey controls also play an important role to effectively
reap the benefits of O&D RM. In this study, we leverage APOS (Airline
Planning and Operations Simulator) of Sabre with real airline data to
simulate and analyse how each of these factors
impact the O&D RM value.
Solving
the primary objective of RM - Revenue maximization through revenue
forecasting
Burak
Ozdaryal – United Airlines
One of
the primary focus areas for RM in the last 10-15 years has been on demand
forecasting in an environment with loose fare restrictions. Multiple
solutions have been proposed and built to model real world “unconstrained”
demand during this time. We suggest an approach that is slightly different,
focusing on estimation of constrained demand – realizable demand given an
availability condition. In this presentation, we will briefly
touch up on this “constrained” or rather conditional demand forecast idea,
but focus primarily on conditional revenue forecasting. We will highlight
the mechanical similarities with forecasting of conditional demand and talk
about its potential use cases.
Product
bundling for airline customers
Manini
Madireddy, Goda Doreswamy, Ramasubramanian Sundararajan, and Meisam Hejazi Nia - Sabre
Airline Solution
We
consider the problem of product bundling (seats and ancillaries) in the
context of offering the right products to airline customers at the right
price and in the right manner, so as to best satisfy customer needs and
maximize airline revenue. This problem falls on the cusp of airline revenue
management (apropos controlling price and availability) and retail
e-commerce (apropos bundle design and shopping session management); therefore,
we synthesize ideas from both domains to devise a solution framework. Our
proposed solution is designed in a modular manner, so as to allow
incremental and independent improvements to product design, pricing and
shopping session management. We demonstrate the utility of our approach
through illustrative results on a sample airline dataset.
Moving towards personalization in
airline RM: Dynamic adjustments of availability and fare quotes
Michael D. Wittman
and Peter Belobaba - MIT International Center for
Air Transportation
Improvements
in airline IT, as well as IATA's New Distribution Capability, would allow
airlines to make customized fare offers to specific passengers or passenger
segments. By strategically targeting offers to the right customers,
airlines could gain new bookings from price-sensitive passengers while
encouraging more price-inelastic travelers to buy-up to higher price
points. We propose two "dynamic availability" heuristics to
adjust fare quotes or class availability in certain situations as a
function of a simple willingness-to-pay estimate. Simulations in the
Passenger Origin-Destination Simulator (PODS) show that the heuristics can
be revenue-positive for airlines and stable in a competitive revenue
management environment.
Incorporating
Ancillary Services in Passenger Choice: Revenue and Booking Impacts
Adam Bockelie
– MIT International Center for Air Transportation
This work
extends the passenger choice process in the Passenger Origin-Destination
Simulator to account for ancillary services. Two extremes of rational choice are
considered: simultaneous behavior, in which passengers choose an itinerary,
fare class, and set of ancillary services at the same time; and sequential
behavior, in which passengers choose an itinerary and fare class in one
phase and then choose a set of ancillary services in a second phase. The sensitivity of revenues and bookings
to various ancillary fee structures are illustrated by simulations in
PODS. Revenue results are presented
comparing traditional leg or network revenue management optimization to an
optimization that incorporates estimates of ancillary revenue.
Use of Dynamic User Influence and
Forecast Multipliers in a Perfectly Tuned PODS Network U10
Larry Weatherford - University of
Wyoming
We have
seen many results over the years where various ‘intelligent aggressiveness’
levers (e.g., Forecast Multipliers, Dynamic User Influence, Hybrid
Forecasting and/or Fare Adjustment) have been used effectively by an
individual airline in a competitive environment to increase its
revenues. The question is whether an
airline can still use these methods to increase revenue in a competitive
environment where all the airlines are perfectly aggressive? This presentation will answer the
question in PODS Network U10 (a large “international” network with 572
O&D markets with four airlines competing for passengers, including an
LCC).
"User
Influence. What is it good
for?"
Bill Brunger - PODS Research LLC
The
author will discuss what RM Users say they are trying to achieve when they
"influence" forecasts and/or adjust fare availability levels
post-optimization. Using results from the PODS Simulator, he will then
examine user strategies, and discuss how airlines should rethink their RM
training & process as they adopt more sophisticated RM algorithms.
Airline
seat allocation with overbooking for unrestricted and fare family
fare-structures
Thomas Fiig (a) (Presenter),
Robert Hjorth (b), Nicolas Bondoux (a), and
Jesper Larsen(b) - (a)Amadeus
Airline IT (b)DTU Management Science
RMS
traditionally solve the seat allocation problem separately from the
overbooking problem. Overbooking is managed by inflating the authorization
levels obtained from seat allocation by various heuristics. This approach
although suboptimal, is necessitated because of the complexity and
dimensionality of the Dynamic Program (DP), which prohibits computation for
realistic size problems. In this paper we solve exactly the combined seat
allocation and overbooking problem for a fully unrestricted and for fare
family fare structures, that have become prevalent in the
recent past, by transforming the problem to an equivalent independent
demand model, thereby avoiding computational complexity. The
resulting availability control - can easily by implemented in existing RM
systems. We investigate the behavior and performance of our joint
seat-allocation, overbooking model and demonstrate that it can deliver
incremental revenue benefits.
Accounting
for price endogeneity in airline itinerary choice models
Laurie A Garrow, Matt Higgins,
Jeff Newman (Georgia Institute of Technology) and Virginie
Lurkin and Michael Schyns
(University of Liege)
This study
formulates an itinerary choice model that is consistent with those used by
industry and corrects for price endogeneity using a control function that
uses several types of instrumental variables. We estimate our models using
database of more than 3 million tickets provided by the Airlines Reporting
Corporation. Results based on Continental U.S. markets for May 2013
departures show that models that fail to account for price endogeneity
overestimate customers’ value of time and result in biased price estimates
and incorrect pricing recommendations.
Extensions to advanced discrete choice models show the importance of
accounting for inter-alternative substitution for products that share
similar departure times.
Using Automation, Simplification and People to reduce
RM and passenger experience conflicts
Eric Boromisa – Avisell
The future of airline travel is changing, and
airlines are investing in improving the customer experience through
creative retailing and merchandising.
We have interviewed thousands of passengers and dozens of airline
executives to get their take on where the industry is heading. We assembled a shortlist of major trends
which should be on the roadmap of every airline - and how RM stakeholders can
take a role in making them a reality.
Do Consumers Benefit from Revenue Management and
Dynamic Pricing?
Guillermo Gallego
(Hong Kong University of Science and Technology) and Ningyuan
Chen (Yale University)
We have asked this question to seasoned practitioners
and leading academics and the answer is often "I don't know, but my
intuition is that if the seller is benefiting it may be because more surplus is being extracted from consumers." This
zero-sum game perception may even prevent the use of RM by companies that
fear it may cause consumer retaliation. It was the expression of this
sentiment by an industry executive that leads us to this investigation. We
asked ourselves: Do consumers prefer benefit from dynamic pricing? We show
that the answer is that dynamic pricing is win-win most of the time, but
there are some exceptions as we will describe in our talk.
Nonparametric Demand Estimation in Airline Revenue
Management
Johannes Jörg, Catherine Cleophas - RWTH Aachen University
Analyzing historical sales data to draw conclusions on
the underlying demand structure is a central theme of revenue management.
This contribution focuses on the estimation of demand segments present in a
market using nonparametric methods on panel data. We employ finite mixtures
to model booking events over time frames and obtain an estimator for the
number of demand segments. Furthermore, we extend the procedure to include
availability data to improve the estimation for an airline application
setting. The approach shows promising performance when applied to simulated
and real panel data sets.
New perspectives on airline ticket data: A
data-driven approach to cluster air travel markets
Sebastian Vock,
Catherine Cleophas, Laurie Garrow - RWTH Aachen University and Georgia Tech
Data-driven clustering can provide new perspectives
on airline ticket data, transcending a priori assumptions about market
differentiation. To that end, we propose and evaluate multiple
preprocessing and clustering approaches. We analyze a large set of data
provided by the Airline Reporting Corporation, including data from six
carriers, 945 routes, and more than 13,100 itineraries.
By benchmarking the results against a more
traditional geographical grouping, we demonstrate the potential for
data-driven analysis in complex transportation networks. The resulting
clusters could support revenue management demand forecast initialization,
performance evaluation, and targeting of analyst interventions.
Dynamic Learning Approach for Personalized Offer
Recommendations
Markus Ettl (IBM
Research), Adam Elmachtoub (Columbia University),
Sechan Oh (IBM Research), Marek Petrik (IBM
Research) and Rajesh Ravi (IBM Research)
Todays empowered consumers are pledging their loyalty
to airlines that provide more information, control and personalized
service. Airlines are increasingly differentiating themselves based on the
experience they offer from the moment passengers start researching a trip
until they arrive at their final destination. With insights derived from
analytics, airlines can deliver personalized services based on customer
preferences which can improve cross-selling or up-selling of other
services.
Consider the following scenario: a traveler is
searching an airline’s website (or mobile sales channel) for a particular
flight itinerary. To increase the odds of the customer making a ticket
purchase in the current session, the airline may issue promotional offers
in real-time, for example, bonus miles, discounted lounge access, or
additional checked baggage. A major caveat is that there is limited
historical data to estimate the success rates of such offers. Thus, the
airline is tasked with the challenge of leveraging customer data to make
personalized promotional offers, but in the absence of any historical data
on how these promotions may perform on specific customers, personas or
markets.
In this talk, we describe an approach, based on
multi-armed bandit and propensity-to-pay models, for addressing this
problem. For each potential promotion, we estimate the unknown parameters
of a logit model using standard techniques with regularization
capabilities. Based on these estimates, we predict the probability the
customer will accept each promotion. In addition, we compute a variance
term that captures the value of learning, and compute an expected reward if
the customer accepts the promotional offer. We show that our algorithm
efficiently learns how to deliver the right promotions to the right types
of customers and itineraries. We evaluate the performance by measuring increase
in profit, regret compared to a clairvoyant solution, and the value of
personalization.
Partnership Revenue Management : Impact of
forecasting, information exchange and revenue sharing agreements
Kavitha
Guddeti, Goda Doreswamy, and Amritraj Misra – Sabre
Airline Solution
Airline partnerships have become one of the major
trends in the recent years. Global demand data has reported more than 40%
increase in code share traffic since 2010. The total benefit of an Alliance
can be achieved when partners in an alliance operate as a single virtual
entity. Various systems of the partner airlines need to interface and
exchange information to achieve the benefit in a decentralized world.
In this study, we leverage APOS (Airline Planning and
Operations Simulator) of Sabre to conduct experiments on real data from two
partner airlines. We evaluate various levels of collaboration between two
airlines from pure competition to total revenue sharing. We consider key
factors such as bid price exchange, forecasting true O&D demand and
structure of revenue sharing agreement to gain insight on opportunities to
align individual airline objectives with the alliance goals.
Robust Revenue Opportunity Modeling
Dong Liang, Richard
Ratliff, Norbert Remenyi - Sabre
This presentation describes research work in progress
on a new, QP-based revenue opportunity model intended to optimize revenue
while providing market-level allocations which are more stable and robust
over time than traditional, LP-based ones.
Because airline O&D networks foster passenger connections, it
results in more markets served than flights operated; this structure
provides additional degrees of freedom for RM bid-price control solutions
with alternate optima (or near optimal) revenue. Although these different alternate
solutions can lead to the same (or nearly the same) network revenue
outcome, they cause manageability issues for airline RM analysts in
practice. A desirable feature of a
ROM solution is, to the extent possible, to generate similar types of
market-level controls over time (e.g. in a market such as JFK-FRA on
Tuesdays, keep the local traffic closed and the flow markets open). Such stability aids RM analysts in
setting effective default allocations and monitoring outliers; it is also a
consideration when holding RM analysts accountable to market-level ROM
performance metrics.
IATA NDC & ONE Order initiatives: industry
opportunities for simplification and value creation
Bryan Wilson and
Sebastien Touraine - IATA
In line with IATA vision to “ be
the force for value creation and innovation” , IATA is driving major
industry initiatives to modernize airline distribution ecosystems. The NDC
standard allows an airline to make sales offers to agents without them
being prepared by third party as an intermediary. It also unlocks
opportunities for the airline to manage other components throughout the
indirect distribution process such as the ability to fulfill the
transaction, create the booking record, issue the document(s) and respond
with confirmations. NDC is the catalyst to transform current airline
shopping and merchandising with new offer and order management workflow
integrated into today artifacts like PNR, ET or EMD. Leveraging NDC, the
ONE Order initiative aims to modernize the booking, ticketing, delivery and
accounting processes with one single and flexible order management process.
It will move from 40 year-old airline-specific processes to current retail
industry order management concepts.
IATA will firstly present a complete end-to-end
architecture of the modernization made possible with NDC & ONE Order.
Beyond programs benefits, IATA will then debate
around 4 areas of opportunities relevant to the Pricing & Revenue
Management , Distribution and e-commerce communities:
1.
Processes Simplification
2.
Optimization Capability
3. Data
Quality & Analytics
4.
Organization Alignment
IATA would like to challenge the AGIFORS RM community
to consider the revenue opportunities that these new processes will
bring. What would be the best RM
practices in a world where airlines make every offer for all passengers
through all channels? And perhaps
most critically at this time of inception – how big is the revenue prize
through eliminating revenue leakage and optimizing every sales opportunity?
A
Capacity Sharing Model with Movable Curtain
Ang Li and Darius
Walczak - PROS
In
this talk, we introduce a single-leg RM model in which the aircraft’s
seating capacity is shared between business and economy compartments. In
particular, a curtain that separates the two compartments can be installed
on the day of flight departure. The control problem is solved by a DP that
jointly considers the booking levels on both sides of the curtain. We
discuss the complexity of this model and show that the increase in
computational burden is moderate over a single-compartment DP. We then
present one enhanced model and two heuristics for the problem, and also
show how to modify the formulation to account for restrictions in where the
curtain can be placed. With actual data, we compare the performances of the
various models as well as show some structural properties.
Leg
revenue management with downsell and delayed
decision-making
Daniel
Hopman – Emirates Airlines
In
this article, we study the impact of downsell in
leg revenue management. Downsell happens when a
customer purchases a lower fare than she was looking for. We aim to
minimize the losses in revenue that arise from this situation. We
reformulate the traditional DP formulation to account for this behavior,
and show substantial revenue gains compared to the traditional DP
formulation. Next, we aim to improve customer booking simulation by
assuming customers may postpone their decision to book. Using a surprisingly
easy reformulation of our DP formulation we ensure cheaper classes will
never open after they get closed, guaranteeing customers booking now is
better than doing so in the future. When more than one eighth of passengers
wait with booking, revenue gains are reported
Modeling
customer behavior for advanced seat reservation
Shuai Shao and Jonas
Rauch - Lufthansa German Airlines
Motivated
by the increasing importance of ancillary services in airline business, we suggested
a statistical approach to modeling the behavior of customers using Advanced
Seat Reservation. By taking a closer look at the data generating process we
decompose this into subsequent decision situations, where different models
such like logit-, discrete-choice- and
time-discrete-survival-model are used to examine the relation between using
ASR and various flight-, booking- and seat-specific covariates.
Vendor Presentations
Big
Data & Predictions
Pablo Fernandez Cardoner - INFARE
|