Transcript Case Study: Airline Revenue Management

Demand and Revenue Management
Anton J. Kleywegt
April 2, 2008
1
Revenue Management




What is Revenue Management
Why do Revenue Management
Pricing Optimization
Demand Modeling and Forecasting
2
What is Revenue Management


Management of inventory, distribution channels
and prices to maximize profit over the long run
Selling the right product to the right customer at
the right time at the right price
3
What is Revenue Management

Revenue Management involves the
following activities





Demand data collection
Demand modeling
Demand forecasting
Pricing optimization
System implementation and distribution
4
What is Revenue Management

Airline industry




How many seats to make available at each of the listed
fares, depending on the OD pair, time of year, time of
week, remaining seats available, remaining time until
departure
What contracts and prices to provide to corporations
How many seats to make available to consolidators and
travel agents (if at all), and at what prices
How much capacity to make available to cargo shippers
and freight forwarders, and at what prices
5
What is Revenue Management

Hotel industry

How much to charge for a room depending on
the location, type of room, time of year, time of
week, duration of stay
6
What is Revenue Management

Ocean cargo industry




Which types of contracts to enter into with
shippers
How much capacity to commit to each shipper
Which contract prices to have for each shipper
How to vary prices as a function of direction of
trade, commodity, and time of year
7
What is Revenue Management

Car rental industry


How much to charge for a rental car depending
on the class of car, time of year, time of week,
duration of rent
Restaurant industry

How much to charge for lunch vs dinner
8
What is Revenue Management

Manufacturing industry



Make-to-stock: dynamic pricing of inventory
Make-to-order: dynamic pricing of orders, how
much discount to give for orders in advance
Make-to-stock and make-to-order: prices of
advance orders vs prices of inventory
9
What is Revenue Management


Retail industry
Example: fashion apparel industry




Products in fashion for a single season
Retailer wants to sell available inventory for
maximum profit
Prices higher at start of season
Retailer has to decide when to mark prices
down, and by how much
10
What is Revenue Management


Entertainment ticket pricing
Example: opera houses let their ticket prices
depend on








The performance
The reviews received so far
Location of seat in opera house
Day of the week of the performance
Time of the day of the performance
Time of performance in the season
Remaining time until the performance
Number of remaining seats available
11
What is Revenue Management

Golf courses

Variable pricing: Choose prices to vary by




time of day
day of week
season of year
Round duration control



control tee-time interval
control uncertainty in arrival time
control uncertainty in duration
12
Hospital Contract Case Study

Major customers of hospitals





Insurance companies
Medicare
Medicaid
Individuals
Hospital contracts with major customers




Discount-off-listed-charges contracts
Per-diem contracts
Case-rate contracts
Capitation contracts
13
Hospital Contract Case Study
Example of setting per-diem rates
ICU Patient Length of Stay
% of Patients

16%
14%
12%
10%
8%
6%
4%
2%
0%
1
2
3
4
5
6
7
8
9 10 11 12 13 14
Number of Days
14
Hospital Contract Case Study

Example of setting per-diem rates


Observe that most patients stay for only a few
days, although a few patients make the average
length of stay quite high
Stratified per-diem rates


Charge more per day to patients who stay for only a
few days
Results


Higher average revenue
Lower standard deviation of revenue
15
Hospital Contract Case Study


Higher average revenue clearly beneficial to
the hospital
Lower standard deviation of revenue

Beneficial to the hospital?


Yes. More predictable revenue
Beneficial to the insurance company?

Yes. More predictable costs
16
What is Revenue Management


Overbooking may be part of revenue management
Overbooking important practice in many
industries that use reservations, and where
cancellations or no-shows may occur






airlines
hotels
car rental
cruise lines
restaurants
contractors (construction etc)
17
What is Revenue Management

Overbooking




Important trade-off between opportunity cost of unused
resources if cancellations or no-shows cause resources
to be wasted, and cost of oversales
In 1960’s, Simon and Vickrey proposed the use of
auctions to allocate airline seats in case of oversales
Airlines rejected idea for many years
Nowadays, reverse Dutch auctions are widely used to
allocate airline seats in case of oversales, and seem to
be widely accepted
18
What is Revenue Management


Dynamic pricing and the bullwhip effect
Dynamic pricing can increase demand
variability

The case of Campbell Soup

Wild swings in demand and in shipments of chicken
noodle soup from the manufacturer to distributors
and retail stores



Increase in production, storage and logistics costs
Frequent stockouts resulting in lost sales
The culprit: Trade promotions!
19
What is Revenue Management


Dynamic pricing and the bullwhip effect
Dynamic pricing can be used to decrease demand
variability

Peak load pricing: lower prices during off-peak times,
higher prices during peak times





Airlines
Hotels
Golf courses
Electricity wholesale market
Oil/gasoline?
20
What is Revenue Management

130
P
70
Revenue Management may involve price
discrimination, but it does not have to
P=130-Q
Consumer surplus=1800
Unit cost = 10
Firm’s profits
under single price:
Firm profits=3600 (130-Q-10)Q
Deadweight loss=1800
60
130
MC=10
q
21
Price Discrimination (continued)
P=130-Q
Unit cost = 10
What if the firm
could segment
the market and
charge two
different prices?
130
Consumer surplus=1600
90
P
Firm profits=4800
50
Deadweight loss=800
MC=10
40
80
130
q
22
Price Discrimination (continued)
130
110
P
Consumer surplus=1000
90
70
Firm profits=6000
50
Deadweight loss=200
30
MC=10
20 40 60 80 100
130
q
23
Price Discrimination (continued)
Perfect price
discrimination
130
Consumer surplus=0
Deadweight loss=0
Firm profits=7200
P
MC=10
130
q
24
What is Revenue Management

The same product sold at different times for different
prices is not necessarily price discrimination, because at
different times...







the production or distribution costs may be different
inventory costs were incurred to keep the product in stock until a
later time
the product value may change over time, such as perishable or
maturing or seasonal products, fashion goods, antiques.
the remaining inventory may be different
interest is earned if product is sold at an earlier time
consumers value products differently at different points in time
locking sales in early reduces uncertainty
25
What is Revenue Management

It is not spam
26
Fairness and Legal Issues


Depending on the industry, there may be legal
obstacles to revenue management
Examples




Regulated prices of utilities (this is changing)
Prices in airline industry were regulated until 1978 price and quantity changes had to be approved by CAB
Pricing in ocean cargo industry was regulated until
1999 - carriers had to provide all shippers with the
same essential contract terms
Spot market pricing in ocean cargo industry is still
regulated - 30 days notice required for price increases
27
Fairness and Legal Issues

Golf course examples

Kimes and Wirtz survey results (1 = extremely
fair, 7 = extremely unfair)






Time-of-day pricing: 3.41
Varying price (for example, as function of bookings
on hand): 6.16
Two-for-one coupons for off-peak use: 1.80
Time-of-booking pricing: 5.12
Reservation fee/Charge for no-shows: 3.19
Tee-time interval pricing: 3.95
28
Fairness and Legal Issues

Amazon.com example





Fall 2000, Amazon conducted experiment to try to
determine price sensitivity of demand for DVDs
Discounts between 20% and 40% offered randomly
Customers who visited amazon.com multiple times
noticed changing prices
Furious response by customers and press, suspecting
Amazon varied price by demographics
Why are varying airline prices accepted by most, and
not varying DVD prices?
29
Why do Revenue Management

Success stories





American Airlines increased annual revenue with $500
million through revenue management
Delta Airlines increased annual revenue with $300
million through revenue management
Marriott hotels increased annual revenue with $100
million through revenue management
National Car Rental was saved from liquidation with
revenue management
Canadian Broadcasting Corporation increased revenue
with $1 million per week
30
Why do Revenue Management

Increasing competition





Fewer restriction on international trade
More efficient international transportation
Low cost foreign competitors
Competitors use revenue management
Use revenue management to stay on top
31
Demand Forecasting


The first law of forecasting: The
forecast is always wrong
Sources of forecast error:



Modeling error
Parameter error
Measurement error
40
Demand Modeling


It is very important to understand and model
customer behavior accurately
Incorrect models of customer behavior can
lead not only to suboptimal prices, but can
lead to the systematic deterioration of
models, prices, and profits over time – the
spiral-down effect
44
Demand Modeling

Spiral-down effect in airline revenue management

For many years, airlines have used following simple
model of customer behavior




Some time before departure, customer requests a ticket in a
particular fare class
Airline accepts or rejects the request
Above model describes the way airline reservations
systems work
However, it does not accurately describe the way
customers behave
45
Demand Modeling

Spiral-down effect in airline revenue
management



Low fare tickets and high fare tickets
Airlines set aside chosen number of seats for
high fare tickets
Airlines use observed sales to estimate the
supposed “demand for high fare tickets”
46
Demand Modeling

Spiral-down effect in airline revenue management

Spiral-down effect:






Airline allows some low fare sales
Some flexible customers (not modeled by the airlines) willing
to buy high fare if that is the only option, now buy low fare
tickets
Airlines observe more low fare sales and less high fare sales –
decrease their estimate of “high fare demand”
Airlines set aside fewer seats for high fare tickets, and allow
more low fare sales
More customers buy low fare tickets, and the spiral down
continues
Spiral-down effect is the consequence of an incorrect
model of customer behavior
47
Demand Forecasting

Forecasting methods


Judgmental methods
Statistical forecasting methods
48
Demand Forecasting

Judgmental forecasting methods


“Expert” opinion
Questionable: See the articles



Armstrong, J.S., “How Expert Are the Experts?”,
Inc, pp.15-16, 1981
Armstrong, J.S., “The Seer-Sucker Theory: The
Value of Experts in Forecasting”, Technology
Review, pp.16-24, 1980
Consensus methods, such as Delphi technique
49
Demand Forecasting

Statistical forecasting methods

Non-causal methods



Causal methods




Exponential smoothing
Time series methods
Linear regression
Nonlinear regression
Discrete choice models (logit, probit, etc)
Whatever the method, the basic approach is to find
systematic behavior in data that one has reason to
believe will continue in the future
50
Demand Forecasting

Forecasting software surveys:




Yurkiewicz, J., “Forecasting: Predicting Your Needs”, OR/MS
Today, volume 31, number 6, pp. 44-52, December 2004,
<http://lionhrtpub.com/orms/surveys/FSS/fssfr.html>.
Swain, J. J., “Desktop Statistics Software: Serious Tools for
Decision Making”, OR/MS Today, volume 26, number 5, pp. 5061, October 1999.
Swain, J. J., “Looking for Meaning in an Uncertain World”,
OR/MS Today, volume 28, number 5, pp. 48-49, October 2001.
Swain, J. J., “2005 Statistical Software Products Survey: Essential
Tools of the Trade”, OR/MS Today, volume 32, number 1, pp. 4251, February 2005,
<http://lionhrtpub.com/orms/surveys/sa/sasurvey.html>.
51
Questions?
54