Page1utdallas.edu/~
metin
Revenue Management
OutlineHistory Levels of Decisions
– Strategy– Tactics
Net Contribution
Prof. Metin Çakanyıldırım used various resources to prepare this document for teaching/training. To use this in your own course/training, please obtain permission from Prof. Çakanyıldırım.
If you find any inaccuracies, please contact [email protected] for corrections.Updated in Fall 2018
Page2utdallas.edu/~
metin
Some U.S. airline industry observations
Since deregulation in 78, 137 carriers have filed for bankruptcy until 2006.
Over 95-99 (the industry’s best 5 years ever) airlines earned 3.5 cents on each dollar of sales:
– The US average for all industries is around 6 cents.– Over 90-99 the industry earned 1 cent per $ of sales.
Carriers typically fill 72.4% of seats while the break-even load is 70.4%.– Utilization is about 3% higher in 2007-08 perhaps due to flight cancellations.
Gas prices vary.– Downsize airlines– Merge. Domestic/international
» UnitedContinentalHoldings.com» AA+USAir
– Invest into alternative transportation modes» High speed train
Page3utdallas.edu/~
metin
Matching supply to demand when supply is fixed
Examples of fixed supply:– Travel industries (fixed number of seats, rooms, rental cars, etc).– Advertising time (limited number of time slots).– Telecommunications bandwidth.– Size of the MBA program.– Doctor’s availability for appointments.
Revenue management is a solution:– If adjusting supply is impossible – adjust the demand!– Segment customers into
» High willingness to pay » Low willingness to pay.
– Limit the number of tickets sold at a low price, » i.e., control the average price by changing the mix of customers.
Page4utdallas.edu/~
metin
Suitable Contexts for Revenue Management
Demand-side– High gross margins: The variable cost of an additional sale is low. – Homogenous capacity: The same unit of capacity (e.g., airline seat) can be used to deliver
services to different segments (e.g., business & leisure customers buying economy seat).– Customer segmentation: Different prices charged to different segments that have different WTP.
» It is not illegal or morally irresponsible to discriminate the customers.
Supply-side– Perishable capacity: Non-storable, degrading/unavailable units. – Limited capacity: All customers cannot be always served.– Costly adjustment of sold capacity: Capacity is sold in advance of demand.
» In contrast, moving curtain between economy and business sections on domestic flights of small European airlines.
Is revenue management for incoming MBA class possible?
Page5utdallas.edu/~
metin
History of Airline Travel in the USA
This picture of manual reservation system is taken from Barry Smith’s AGIFORS presentation.
1950’s, 20% of Americans have ever flown on an airplane. 1964, Sabre (reservation system) became operational Flights used to feature good service, even live music by
– Professionals: At the beginning of his career Frank Sinatra is said to play on NYC-LA flights. Read in American Way Magazine of American Airlines but could not confirm independently.
– Passengers as in the picture on the right. Air travel was an exciting experience but now became a torture
to be endured, except for some Southwest flights:– See flight attendant M. Cobb https://www.youtube.com/watch?v=07LFBydGjaM
A phonograph cover
Page6utdallas.edu/~
metin
Published Fares – Controlled by Civil Auronautics BoardDeregulation in 1978 removed the restrictions on schedules and fares
Good Service & High Cost ThenIn 1950s, OW (one way) ticket
Dallas DC for $70. This corresponds to $700 over 60
years with 4% inflation.This is from Barry Smith’s AGIFORS presentation.
Page7utdallas.edu/~
metin
American vs. PeopleExpress
After deregulation, low-cost carrier PeopleExpress was established circa 1980– PeopleExpress charged 70% less than major airlines – Fast growth over 1980-1984 by serving underserved markets: Leisure passengers– In 1984, attacked AA’s bread-and-butter routes: Newark-Chicago and New Orleans-LA
January 1985, American’s counterattack: Ultimate Super Saver– Discount fare if reserving 2 weeks before departure and staying over a Saturday night.– Restrictions on the number of seats available under Ultimate Super Saver program
Leisure passengers used Ultimate Super Saver while Business travelers paid the full fare.
September 1985, PeopleExpress on the verge of bankruptcy– We had great people, tremendous value, terrific growth. We did a lot of things right. But
we didn’t get our hands around the yield management and automation issues. – Donald Burr, CEO of PeopleExpress
Page8utdallas.edu/~
metin
American’s Experience with RM in 1980’s
MOMS: Multiclass Optimization Modeling System, 1979 DADS: Discount Allocation Decision System, 1980 CARS: City Allocation Report, 1981 SCARS: Super City Analysis System, 1982 DINAMO: Dynamic Inventory Allocation Modeling Optimizer, 1986
Airlines have been in the RM game for a long while. Hotels and Car Rental companies came afterwards.
Page9utdallas.edu/~
metin
Airline Mergers in the Wake of Low ProfitsTrans World
AA
US Air
America West
Continental
United
Northwest
Delta
Western
Southwest
Value Jet
AirTran
TWA
AA
US Air
America West
Continental
United
Northwest
Delta
Western
Southwest
Value Jet
AirTran
1985 1990 1995 2000 2005 20102007 201120131978
Deregulation
Based on the data in S. McCartney, 2013. How the AMR Deal Might Affect Airline Service in US. February 14 issue of WSJ.
Time axis is not drawn to scale. Post-merger horizontal lines are thicker.
Page10utdallas.edu/~
metin
US Airlines when AA & US Air Merge2012 Revenue in $B # of Aircrafts
AA+US Air 24.9+13.8=38.7 898+624=1,522
United 37.2 1239
Delta 36.7 1289
Southwest 13.8 692
Airport Flights/ week
Departing seats / wk
AA+US Air %
United%
Delta%
Southwest %
Other%
CLT 4,791 473,062 2.2+89.9 1.7 4.2 - -
DFW 6,053 685,666 83.6+02.7 - 3.7 - Spirit 3.3
PHL 3,809 352,904 3.1+75.5 - 5.9 6.8 -
MIA 3,111 472,913 68.1+02.3 3.0 5.2 - -
PHX 3,535 459,844 3.9+47.0 - 5.5 32.7 -
ORD 7,541 705,721 35.9+02.8 46.7 3.9 - -
LAX 4,905 693,485 19.0+02.3 18.3 13.7
JFK 3,498 557,302 18.0+01.4 - 21.1 - JetBlue 23.4
Based on the data in S. McCartney, 2013. How the AMR Deal Might Affect Airline Service in US. February 14 issue of WSJ.
More than 50% share in CLT (Charlotte, NC), DFW, PHL, MIA, PHX.
Page11utdallas.edu/~
metin
Airline Mergers: American + US Air = American over 2013-2015
AA and US Air Merger started in 2013, ended on Oct 29, 2015
with a victorious email announcement to passengers
Earlier mergers suffered from integration problems:– “[United and Continental] merged in 2010 and made the system
transition in early 2012, disrupted passengers and employees alike for weeks.”
– “US Airways and America West Airlines linked up in 2005, executed poorly on their IT transition two years later.”
AA’s gradual transition Oct 17 Sat morning integration date– Before March 2015: “The US Airways cargo, revenue
accounting, and employee- and retiree-travel functions already switched to American versions”
– March 2015: “US Airways’ frequent flier plan was folded into the larger American AAdvantage program”
– July: “American … started transferring reservations made on US Airways for travel on or after Oct 17 using American’s flight codes and to its Sabre Corp. reservations-system provider.” “American also stopped offering a US Airways schedule beyond Oct. 17, forcing passengers to shift bookings to American.”
– On Oct. 17, US Airways operates ~200 fewer flights to its three hubs (11% less than usual) to lighten the load on the system.
– “For the transition period, [American] raised airport staffing levels by more than 20%.”
– “American’s big switch [was] run from a command center near its headquarters... There also [were] 23 satellite offices, including one at Sabre’s reservations database in Tulsa, Okla. The centers went live on Oct 14 and [were] expected to stay open around the clock until Oct 27.”
Based on S. Carey. 2015. American Airlines’ Reservations Test:
Smooth Travel This Weekend. Oct 15 issue of WSJ.S. Carey. 2015. Few Glitches in American Switchover. Oct 18
issue of WSJ.
Page12utdallas.edu/~
metin
Revenue Management StrategyCustomer Segmentation
Leisure Passengers vs. Business Passengers– Leisure passengers are highly price sensitive– They book earlier– They have more flexibility in departure and arrival times– They accept or prefer Saturday night stayover– Finer segmentation is also possible among Leisure and Business Passengers depending on
price sensitivity, schedule flexibility, etc. Other airline customer segments include
– Government employees, senior citizens, children, frequent flyers, etc.– International pricing: Same round trip route and same starting points could lead to different
prices in different countries. » Any logic here? This may be a relic of the past as it does not make much sense in the age of Internet.
– Origin based segmentation: Same round trip route but different starting points can lead to different prices in different countries.
– Channel based segmentation, see next page: Travel agent in person vs. Travel agent software vs. Airline’s website.
In addition to these, hotels, theme parks and cruise lines can have city-of-origin based pricing
– Hotels: Kamaaina rates for Hawaii residents– Theme Parks: Disney’s resident rates for Orlando residents– Cruise Lines: City-of-origin needed to fly the passengers from out of town
Page13utdallas.edu/~
metin
Reservation and Distribution System
Leis
ure
Cus
tom
ers
Global Distribution
SystemAmadeus, Galileo, Sabre, Worldspan
LargeCorporate
Travel Agent
FrequentFlyer Program
Air lines.com
Travel agent.com
Distressed.com
Bus
ines
s Cus
tom
ers
800Number
Opaque.com
ComputerizedReservation
System
Page14utdallas.edu/~
metin
Booking ControlBooking Limits & Protection Levels
Booking control: Controlling the number of products (seats, rooms, rental cars)sold to customer classes
Booking limit– To ensure that limited number of products sold to low-paying customers
Booking limit (b) for a fare class is the maximum number of products that can be sold in that class or in lower classes.
Protection level (y) for a fare class is the number of products reserved for that class or higher classes.
Protection levels – To ensure that enough products are sold to high-paying customers
Page15utdallas.edu/~
metin
Booking Limit vs. Protection Level 2 Fare Classes at a Hotel
2 classes: low paying and high-paying customers for the same room. Limits and levels for nested classes.
The booking limits ,– one for the low-paying customers, is the maximum number of rooms
to sell to low-paying customers.– booking limit for high-paying customers? No!– another for the low-paying and high-paying customers, is the
maximum number of rooms to sell to low-paying and high-paying customers. Effectively, the booking limit for these is the capacity.
The protection levels– one for the high-paying customers, is the number of rooms reserved
for high-paying customers. – protection level for low-paying customers? No!– one for the low-paying and high-paying customers, is the number of
rooms reserved for low-paying and high-paying customers. This is the capacity.
Only two numbers above differ from the capacity: booking limit for the low-paying customers and the protection level for high-paying customers:
0 200150 booking limit
Capacity200 rooms
50 roomsprotected for high-payingcustomers
200-50=150 rooms is
the booking limit for low-
paying customers
50 protection level
The rooms are not physically allocated to classes. Allocation is virtual. Room number
106 can serve both class R and class Q.
low-paying customers high-paying customers
Page16utdallas.edu/~
metin
Multiple Fare Classes Nested Booking Limits
Nested limits: A limit is for a class and all lower classes.– Booking limit for class {3} is 1; booking limit for classes {2,3} is 2.– Request Arrival Sequence A = Class 2 → Class 2 → Class 3
» Class 2 accepted, class 2 accepted and class 3 rejected» Sold capacity=2; Revenue = 2*[Class 2 price]
– Request Arrival Sequence B = Class 2 → Class 3 → Class 2» Sold capacity=2; Revenue = [Class 2 price]+[Class 3 price]
– Request Arrival Sequence C = Class 3 → Class 2 → Class 2» Sold capacity=2; Revenue=[Class 3 price]+[Class 2 price]
Partitioned limits (Allotments): A limit is set for a single class only.– Booking limit for class {3} is 1; booking limit for class {2} is 1.– Request Arrival Sequence A = Class 2 → Class 2 → Class 3
» Sold capacity=2; Revenue=[Class 2 price]+[Class 3 price] – Request Arrival Sequence B = Class 2 → Class 3 → Class 2
» Sold capacity=2; Revenue = [Class 2 price]+[Class 3 price] – Request Arrival Sequence C = Class 3 → Class 2 → Class 2
» Sold capacity=2; Revenue=[Class 3 price]+[Class 2 price]
b1
b2
b3
Sold capacity and revenues are the same except for Sequence A where nested limits are better. Allotment violates RM’s first principle: high-paying customers have priority over low-paying customers.
Suppose classes are 1,2,3, and 1 is the highest class. Nestedness can be illustrated with at least 3 classes.3 request arrivals{Class 2, Class 2, Class 3} in 3 sequences: A=223, B=232, C=322
Page17utdallas.edu/~
metin
Multiple Fare Classes: Nested Booking Limits with Another Set of Arrivals
Nested limits: – Booking limit for class {3} is 1; booking limit for classes {2,3} is 2.– Request Arrival Sequence A = Class 2 → Class 3 → Class 3
» Sold capacity=2; Revenue = =[Class 2 price]+[Class 3 price] – Request Arrival Sequence B = Class 3 → Class 2 → Class 3
» Sold capacity=2; Revenue = [Class 3 price]+[Class 2 price] – Request Arrival Sequence C = Class 3 → Class 3 → Class 2
» Sold capacity=2; Revenue = [Class 3 price]+[Class 2 price]
Partitioned limits (Allotments):– Booking limit for class {3} is 1; booking limit for class {2} is 1.– Request Arrival Sequence A = Class 2→ Class 3→ Class 3
» Sold capacity=2; Revenue=[Class 2 price]+[Class 3 price] – Request Arrival Sequence B = Class 3 → Class 2 → Class 3
» Sold capacity=2; Revenue = [Class 3 price]+[Class 2 price] – Request Arrival Sequence C = Class 3 → Class 3 → Class 2
» Sold capacity=2; Revenue = [Class 3 price]+[Class 2 price]
b1
b2
b3
Sold capacity and revenues are the same in all Sequences A, B, C. The other 3 request arrivals involving Classes 2 and 3 include only one class (e.g., 222 or 333) so they
result in the same revenues under both nested limits & partitioned limits. On this page and the previous one, we have considered all arrivals involving 3 requests of Class 2 and 3
and all possible sequences. Nested limits are better in one instance and never worse.
3 request arrivals{Class 2, Class 3, Class 3} in 3 sequences: A=233, B=323, C=332
In general, nested limits are better.
Page18utdallas.edu/~
metin
Multiple Fare Classes: Nested Booking Limits, Another Example with Batch Arrivals
Batch arrivals of customers to the reservation system, when they travel togetherBooking limits should be nested for batch arrivals as well
Nested limits: A booking limit is set for a class and all lower classes. Suppose classes are 1,2,3 & 1 is the highest class.
» Booking limit for class {3} is 8 and the booking limit for classes {2,3} is 20. » After selling 12 to class 2 customers,» Requests for 6 more class 2 or 3 customers are accepted,
If class 2: 0+0<8 limit for class 3; 12+6≤ 20 limit for class 2 and 3 If class 3: 0+6≤8 limit for class 3; 12+6≤ 20 limit for class 2 and 3
Partitioned limits (Allotments): A limit is set for a single class only.» Booking limit for class {3} is 8 and the booking limit for classes {2} is 12. » After selling to 12 class 2 customers, » Requests for 6 more class 2 or 3 customers,
Request for 6 more class 2 is rejected, 12+6>12. Request for 6 more class 3 is accepted. 0+6≤8.
» This is accepting low-paying customers while rejecting high-paying customers.
Page19utdallas.edu/~
metin
What are Fare Classes at a Hotel? Physically but minimally differentiating the product
Consider a standard room, not a deluxe or a suit. Class 1: Free Internet, Fruits at arrival, daily Newspaper, 2 bottles of
water everyday, choice of bed (2 queens or a king). These services are just add-ons that are not specific to the room size.
Class 2: Internet at $5/day. The rest is the same as Class 1. Class 3: No fruits; The rest is the same as Class 2. Class 4: No newspaper; The rest is the same as Class 3. Class 5: Water at $3/bottle; The rest is the same as Class 4. Class 6: No choice of bed type; The rest is the same as Class 5.
Prices: $250; $220; $190; $160; $130; $100. Cost of extra services cannot justify the price differential of $30. Prices are
not cost based but value based, and used to segment the customers.
Different distribution channels can use different selling practices: When minimally differentiated, show available classes whose prices differ
and ask customers to choose 60 days before check-in available classes={1,2,3,4,5,6}, show these 30 days before check-in available classes={1,2,3,4,5}, show these 10 days before check-in available classes={1,3,5}, show these 2 days before check-in available classes={1,3}, show these
Do not even minimally differentiate products: Exactly the same room. Show only the minimum priced available class and ask customers to buy it 60 days before check-in available classes={1,2,3,4,5,6}, show 6 30 days before check-in available classes={1,2,3,4,5}, show 5 10 days before check-in available classes={1,3,5}, show 5 2 days before check-in available classes={1,3}, show 3
b1
b2
b3
b4
b5
b6
Rooms are not physically allocated to classes. Allocation is virtual. For
example, room number 206 can serve class 1, 2, 3, 4, 5 or 6.
Page20utdallas.edu/~
metin
Multiple Fare Classes Nested Booking Limits & Protection Levels
Classes indexed by 𝑗𝑗 = 1,2, … ,𝑛𝑛. 𝑗𝑗 = 1 is the highest class, while 𝑗𝑗 = 𝑛𝑛 is the lowest class. Booking limit 𝑏𝑏𝑖𝑖 for class i limits bookings for classes 𝑗𝑗 = 𝑖𝑖, … ,𝑛𝑛 = 𝑖𝑖 ↔ 𝑛𝑛 Protection level 𝑦𝑦𝑖𝑖 for class i protects future reservations or classes 𝑗𝑗 = 1,2, … , 𝑖𝑖 = 1 ↔ 𝑖𝑖.
– Capacity = 𝑏𝑏1 = 𝑦𝑦𝑛𝑛; 0 = 𝑏𝑏𝑛𝑛+1 = 𝑦𝑦0Protection levelsBooking limits
{1,.,n}{3,.,n}
Class 1
Class 2
Class 3
Class n-1
Class n
{2,.,n}{n-1,.,n} {1}{1,.,n} {1,.,n-1} {1,2,3} {1,2}{n}
Because of nestedness, 0 = bn+1 < bn < bn-1 < …. < b2 < b1 and yn > yn-1 > yn-2 > …. > y1 > y0 = 0.
Page21utdallas.edu/~
metin
Capacity=Booking Limits + Protection Levels Booking limit 𝑏𝑏𝑖𝑖 for class i limits bookings for classes [𝑖𝑖:𝑛𝑛]. Protection level 𝑦𝑦𝑖𝑖−1 for class 𝑖𝑖 − 1 protects future reservations or classes [1: 𝑖𝑖 − 1]. Booking limit bi over [𝑖𝑖:𝑛𝑛] protects classes 1: 𝑖𝑖 − 1 :
– [Limiting classes[𝑖𝑖:𝑛𝑛]] ≡ [Protecting classes 1: 𝑖𝑖 − 1 ]. Number protected for 1: 𝑖𝑖 − 1 : 𝑦𝑦𝑖𝑖−1 = capacity − 𝑏𝑏𝑖𝑖 = 𝑏𝑏1 − 𝑏𝑏𝑖𝑖 .
{1,.,n}{3,.,n}
Class 1
Class 2
Class 3
Class n-1
Class n
{2,.,n}{n-1,.,n}
{1}{1,.,n} {1,.,n-1} {1,2,3} {1,2}
{n}
{1,.,n-2}
{4,.,n}
{}
{}
b1b2b3b4bn-1bnbn+1
y0y1y2y3yn-2yn-1yn
𝒃𝒃 𝒊𝒊+𝒚𝒚 𝒊𝒊−𝟏𝟏
=ca
paci
ty
Page22utdallas.edu/~
metin
Summary: Booking Limits and Protection Levels
Class 1
Class 2
Class 3
Class n-1
Class nb1 b2 b3 b4 bn-1bnbn+1
y0 y1 y2 y3 yn-2yn-1yn C
apac
ity
Page23utdallas.edu/~
metin
Class 1
5 Requested for Class 2
Class 3
Class 4
Class 5b1b2b3b4b5b6
y0y1y2y3y4y5
Protectionlevels
Bookinglimits
100 100 96 88 27 0
1007312400
Protect less for Class 2 and above
Reduce booking limit for class 2 and above
Example: Increasing Bookings
This approach is pursued in the appendix. Altering booking limits is confusing.
Page24utdallas.edu/~
metin
Example: Increasing Bookings while Keeping Booking Limits Constant
Suppose Booking limits=b=(b1, b2, b3, b4, b5) = (100, 73, 12, 4, 0). Current bookings=B=(B1, B2, B3, B4, B5) = (0, 0, 0, 0, 0) 5 requested for class 2; Let x=(x1, x2, x3, x4, x5) = (5, 5, 0, 0, 0) for the request.
Class 2 bookings are restricted by b2=73 (applies to classes 2..5) and b1=100 (applies to classes 1..5) . Both have enough room as b2-x2=68 ≥ 0 and b1-x1=95≥0, so accept the request.
Set the current bookings to B=(5, 5, 0, 0, 0)=(0, 0, 0, 0, 0)+(5, 5, 0, 0, 0). 1 requested for class 2; Let x = (1, 1, 0, 0, 0) for the request.
Since b2-B2-x2=67 ≥ 0 and b1-B1-x1=94 ≥ 0, accept the request. Set the current bookings to B= (6, 6, 0, 0, 0)=(5, 5, 0, 0, 0) + (1, 1, 0, 0, 0).
1 requested for class 4; Let x = (1, 1, 1, 1, 0) for the request. Since b4-B4-x4=3 ≥ 0 , b3-B3-x3=11 ≥ 0 , b2-B2-x2=66 ≥ 0 and b1-B1-x1=93 ≥ 0, accept the request. Set the current bookings to B=(7, 7, 1, 1, 0)= (6, 6, 0, 0, 0)+(1, 1, 1, 1, 0).
3 requested for class 3; Let x = (3, 3, 3, 0, 0) for the request. Since b3-B3-x3=8 ≥ 0 , b2-B2-x2=63 ≥ 0 and b1-B1-x1=90 ≥ 0, accept the request. Set the current bookings to B=(10, 10, 4, 1, 0)=(7, 7, 1, 1, 0)+(3, 3, 3, 0, 0).
4 requested for class 4; Let x = (4, 4, 4, 4, 0) for the request. Since b4-B4-x4=-1<0 , b3-B3-x3=4 ≥ 0 , b2-B2-x2=59 ≥ 0 and b1-B1-x1=86≥ 0, reject the request. Keep the bookings as B=(10, 10, 4, 1, 0).
2 requested for class 1; Let x = (2, 0, 0, 0, 0) for the request. Since b-B-x=(88, 63, 8, 3, 0) ≥ 0 , accept the request. Set the bookings to B =(10, 10, 4, 1, 0) + (2, 0, 0, 0, 0) =(12, 10, 4, 1, 0) .
Page25utdallas.edu/~
metin
Example: Increasing Bookings while Keeping Booking Limits Constant
b=(b1, b2, b3, b4, b5,)=(100, 73, 12, 4, 0). B=(B1, B2, B3, B4, B5)=(12, 10, 4, 1, 0). 30 requested for class 2; Let x=(x1, x2, x3, x4, x5,) = (30, 30, 0, 0, 0) for the request.
Since b-B-x=(58, 33, 8, 3, 0) ≥ 0 , accept. Set B =(12, 10, 4, 1, 0) + (30, 30, 0, 0, 0) =(42, 40, 4, 1, 0) .
20 requested for class 2; Let x = (20, 20, 0, 0, 0). Since b-B-x=(38, 13, 8, 3, 0) ≥ 0 , accept. Set B =(42, 40, 4, 1, 0) + (20, 20, 0, 0, 0) =(62, 60, 4, 1, 0) .
10 requested for class 3; Let x=(10, 10, 10, 0, 0). Since b-B-x=(28, 3, -2, 3, 0) ≥ 0 fails, reject. Keep B =(62, 60, 4, 1, 0) .
6 for class 3, x=(6, 6, 6, 0, 0). If b-B-x=(32, 7, 2, 3, 0) ≥ 0, accept. Set B =(62, 60, 4, 1, 0)+(6, 6, 6, 0, 0)=(68, 66, 10, 1, 0) .
3 for class 3, x=(3, 3, 3, 0, 0). Since b-B-x=(29, 4, -1, 3, 0) ≥ 0 fails, reject. Keep B =(68, 66, 10, 1, 0) .
6 for class 2, x=(6, 6, 0, 0, 0). If b-B-x=(26, 1, 2, 3, 0) ≥ 0, accept. Set B =(68, 66, 10, 1, 0)+(6, 6, 0, 0, 0)=(74, 72, 10, 1, 0).
1 for class 3, x=(1, 1, 1, 0, 0). If b-B-x=(25, 0, 1, 3, 0) ≥ 0, accept. Set B =(74, 72, 10, 1, 0)+(1, 1, 1, 0, 0)=(75, 73, 11, 1, 0).
Page26utdallas.edu/~
metin
Example: Increasing Bookings while Keeping Booking Limits Constant
b=(b1, b2, b3, b4, b5,)=(100, 73, 12, 4, 0). B=(B1, B2, B3, B4, B5)=(75, 73, 11, 1, 0). Since b-B=(25, 0, 1, 3, 0), all low fare classes are closed. Only the first class can accept from now on. 1 for class 5, x=(1, 1, 1, 1, 1).
Since b-B-x=(24, -1, 0, 2, -1) ≥ 0 fails, reject. Keep B =(75, 73, 11, 1, 0) .
1 for class 4, x=(1, 1, 1, 1, 0). Since b-B-x=(24, -1, 0, 2, 0) ≥ 0 fails, reject. Keep B =(75, 73, 11, 1, 0) .
1 for class 3, x=(1, 1, 1, 0, 0). Since b-B-x=(24, -1, 0, 3, 0) ≥ 0 fails, reject. Keep B =(75, 73, 11, 1, 0) .
1 for class 2, x=(1, 1, 0, 0, 0). Since b-B-x=(24, -1, 1, 3, 0) ≥ 0 fails, reject. Keep B =(75, 73, 11, 1, 0) .
25 for class 1, x=(25, 0, 0, 0, 0). If b-B-x=(0, 0, 1, 3, 0) ≥ 0, accept. Set B =(75, 73, 11, 1, 0)+(25, 0, 0, 0, 0)=(100, 73, 11, 1, 0).
Since b-B=(0, 0, 1, 3, 0), all classes are closed. Recover bookings
First class: B1-B2=27; Second class: B2-B3=62; Third class: B3-B4=10; Fourth class: B4-B5=1; Fifth class: B5=0.
Page27utdallas.edu/~
metin
Booking Algorithm with Booking Limits
{Initialization}B := 0. Booking limit b is given. {Both B and b are vectors of length n.}
{Iterative step}While Remaining_capacity=b1-B1 > 0 do
Suppose m requests for class i: x=[m, …, m, 0, …, 0]. {x is a vector (of length number of classes)
whose only first i elements are nonzero and equal to m.m<0 denotes a cancellation.}
If b≥B+x, Accept the request and set B:=B+x; else Reject the request.
EndWhile
Page28utdallas.edu/~
metin
Tactical Revenue Management Tactical RM is calculating and updating booking limits.
– Periodic checking/updates: Daily checking of expected vs. realized demand, when 15 or fewer days until departure. Less frequently if more days. Update booking limit if realized differs significantly from expected.
– Event-driven updates: NBA all-star game in Dallas in 2010; Aircraft change. Some terminology
– Resources: Units of capacity managed by a supplier – Products: What customers want to purchase.– Fare class: Group or resources allocated to a segment of customers.
Passenger Airline SOMResource unit Seat Course
Resource types Service classes OM, SCM, Finance, Energy…
Number of types 1-3 Many
Products Itineraries Degrees
Resources ina product
Many for hub-and-spoke airlines ;A few for point-to-point airlines.
Many for MBA;Fewer for Master.
– Load factor: Utilization of resources. Increase by booking more. – Yield: Revenue per passenger mile. Increase by booking less.– RASM/CASM: Revenue/Cost per available seat mile. RASM=Net yield * Load factor.– REVPAR: Revenue per available room night.
Page29utdallas.edu/~
metin
Executing Revenue Management
Capacity Allocation: Booking for various fare classes. Network management: Booking across the flight network of an airline or
multiple days at a hotel.– Network management is not relevant for
» Club-Med Resorts accepts check-in only on Sundays and a minimum stay of at least 7 days.» Point-to-point airlines. » Cruises depart on a certain day & everybody remains on the ship for the same number of days. » Concert and sport events are similar to cruise lines, the start and ending time of the event is the
same for everybody. Exception: Some jazz clubs run multiple sittings on a single night. Some patrons stay for 2 sittings in a row.
Overbooking: Booking extra to compensate for no-shows.– Resorts, cruise lines, events are not overbooked. In these industries, there is no viable
method of serving to an overbooked customer. – Some airline tickets are refundable and parts of customer’s hotel costs can be recovered
when one does not show up. Because of these, both airlines and hotels use overbooking.
Page30utdallas.edu/~
metin
Smith, Guenther, Rao & Ratliff (2001)“E-Commerce & Operations Research in Airline Planning, Marketing & Distribution”, INTERFACES 31:2
Systematic View of Revenue Management
Page31utdallas.edu/~
metin
Net Contribution=Price+Ancillary Fees−Incremental Cost
Incremental costs and fee obtained from ancillary products/services can change the net contribution.
Relative importance of incremental costs depends on the industry
Incremental Cost ImportanceCruise lines Commission, food, cleaning High
Container shipping Repositioning, handling, fuel High
Passenger airlines Commission, fuel, food, passenger fees Moderate
Rental cars Commission, processing, wear/tear Moderate
University Extra sessions, TAs, wear/tear Moderate
Hotels Commissions, cleaning, wear/tear Low
Casinos Commissions, cleaning, wear/tear Low
Events Commissions Very low
Page32utdallas.edu/~
metin
Net Contribution=Price+Ancillary Fees−Incremental Cost
Ancillary fees ImportancePassenger airlines Food, beverage, ear-phone, duty-free sales Low
University Activity fee, library fee Low-Moderate
Hotels Food, beverage, minibar, internet, phone fees Moderate
Container shipping Call before delivery, expediting Moderate
Rental cars Insurance, gasoline Moderate
Cruise lines Gambling, on-board sales Moderate-High
Casinos Gambling losses High
Events Food, beverage, merchandise, parking Very high
Ancillary fees becoming more important for passenger airlines.
Industry-dependent Relative Importance of Ancillary Fees.
What to expect in the future?
Page33utdallas.edu/~
metin
Recent Trends: RM in Practice Strategic customer: Because of fare visibility, ease of communication among customer
groups, customers are better informed than ever before. They may optimize the timing, the amount of their purchases.
– Case in point: Strategic retail customers wait for discounting season. A higher percentage of merchandise is sold every year through discounting. Can the retailers remain in business without resorting to discounts. Neiman Marcus (One Marcus Square, 1618 Main Street, Dallas) is trying.
– Yapta.com: Price drop alerts help you to time the purchase. Online travel agents/intermediaries (Expedia, Travelocity and Priceline) provide cheaper
distribution channels.– Attempting to book a hotel in San Diego: Google search for “cheap hotels San Diego” takes you to
Expedia.com. You book a hotel with Expedia and pay Expedia not the hotel. Expedia pays a wholesale rate to the hotel. The hotel will not reveal the wholesale rate to customers unless you insist. If you insist after the check-out, you will see that Expedia charges about 20% premium as distribution/commission fee.
Opaque products: Hidden product attributes. – Go to priceline.com bid for a hotel in San Diego for three nights, starting on coming Saturday.
» 1. Choose location: Downtown San Diego & Harbor Island. » 2. Choose star level: 3.5 stars or more. » 3. Name your price: $70 per night. (You are prompted with a median price of $187 per night – reference price.)» 4. Enter name/credit card info. What exactly did you buy?
Last-minute deals: See http://lastminute.travelocity.com Customer specific satisfaction measures as opposed to timely departure and arrival
– D0, A14: On-time departure and at most 14 minute late arrival per flight; binary data reported to FAA, DOT.– Delay in each customer’s arrival time per customer; data in minutes are integers, captured by AA but not
reported outside the company, according to Oct 14 2015 Informs DFW Chapter presentation by Jim Diamond.
Page34utdallas.edu/~
metin
When to Buy a Ticket
Source: S. McCartney. The Best Day to Buy Airline Tickets. WSJ The Middle Seat Column, Oct 22, 2014.
Page35utdallas.edu/~
metin
Summary
History Levels of Decisions
– Strategy– Tactics
Net Contribution
Follow-up reading: Phillips (2005) Chapter 6
Page36utdallas.edu/~
metin
Example: Decreasing Limits & Levels5 Fare Classes
Class 1
5 Requested for Class 2
Class 3
Class 4
Class 5b1b2b3b4b5b6
y0y1y2y3y4y5
Protectionlevels
Bookinglimits
100 100 96 88 27 0
1007312400
Affected
Affected
Page37utdallas.edu/~
metin
Example: Booking Limits and Protection Levels5 Classes
Class 1
1 for Class 2
Class 3
Class 4
Class 5
95 95 91 83 27 0
956812400
Class 1
Class 2
Class 3
1 for Class 4
Class 5
94 94 90 82 27 0
946712400
Class 1
Class 2
3 for Class 3
Class 4
Class 5
93 93 90 82 27 0
936611300
Class 1
Class 2
4 for Class 4
Class 3
Class 5
90 90 87 82 27 0
90638300
Request Rejected
Page38utdallas.edu/~
metin
Example: Booking Limits and Protection Levels5 Classes
2 for Class 1
Class 2
Class 4
Class 3
Class 5
90 90 87 82 27 0
90638300
30 for Class 2
Class 1
Class 4
Class 3
Class 5
88 88 85 80 25 0
88638300
20 for Class 2
Class 1
Class 4
Class 3
Class 5
58 58 55 50 25 0
58338300
Class 1
Class 4
10 for Class 3
Class 5
38 38 35 30 25 0
38138300
Class 2Request Rejected
Page39utdallas.edu/~
metin
Example: Booking Limits and Protection Levels5 Classes
Class 1
Class 4
6 for Class 3
Class 5
38 38 35 30 25 0
38138300
Class 2
Class 1
Class 4
Class 3
Class 5
32 32 30 30 25 0
3278-6200
Class 2 max{35-6,30}
min{3,8-6}
Order preservingoperations
Class 1
Class 4
3 for Class 3
Class 5
32 32 30 30 25 0
3272200
Class 2Request Rejected
6 for Class 2
Class 1
Class 4
Class 3
Class 5
32 32 30 30 25 0
3272200
Page40utdallas.edu/~
metin
Example: Booking Limits and Protection Levels5 Classes
Class 1
Class 4
1 for Class 3
Class 5
26 26 25 25 25 0
2611100
Class 2
Class 1
Class 4
Class 3
Class 5
26 26 25 25 25 0
267-61100
Class 2max{30-6,25}
min{2,7-6}min{2,1}
max{30-6,25}
Class 1
Class 4
Class 3
Class 5
25 25 25 25 25 0
2500000
Class 2
All low-fare classes are closed.From now on accept only the first class.Recover bookings by accumulating accepted bookings.
We must make order preserving operations to preservenestedness. These are cumbersome. Why not to keep booking limits constant and increase bookings? See the algorithm in the main body.