utdallas.

edu/~metin
Page 1
Network Revenue Management

Outline
 Network Management Problem
 Greedy Heuristic
 LP Approach
 Virtual Nesting
 Bid Prices

Based on Phillips (2005) Chapter 8

 utdallas.edu/~metin
Page 2
Demand for Hotel Rooms Vary over a Week
Occupancy

100%

Monday Tuesday Wednesday Thursday Friday Saturday Sunday

Since the demand varies, so should the booking limits.
Many customers book for several days in a row
 Which booking limits apply to a booking request for Tuesday night check-in and 3 night stay? Tue, Wed, or Thu?
Since requests are coupled over days, so should the booking limits.
Compute the booking limits not separately day by day but by “jointly” for seven days!
 utdallas.edu/~metin
Page 3
Hotel Booking Network is a Chain but Extensive
Sat -1 Fri -1 Thu -1 Wed -1 Tue -1 Mon -1 Sun -2

Sun
-1 Time flow chronologically

Mon Tue Wed Thu Fri Sat Sun

Mo Tu We Th Fr Sa Su Mo+1
night night night night night night night

O O D O O D D D
Origin-Destination (check-in,check-out) day pairs with 7 nights:

Mo-Tu, Mo-We, Mo-Th, Mo-Fr, Mo-Sa, Mo-Su, Mo-Mo+1 all check in on Monday but stay 1, 2, …, 7 nights
Tu-We, Tu-Th, Tu-Fr, Tu-Sa, Tu-Su, Tu-Mo+1 all check in on Tuesday but stay 1, 2, …, 6 nights
We-Th, We-Fr, We-Sa, We-Su, We-Mo+1 all check in on Wednesday but stay 1, 2, …, 5 nights M T W Th F S Su
M
Th-Fr, Th-Sa, Th-Su, Th-Mo+1 all check in on Thursday but stay 1, 2, 3, 4 nights T

Check-in
W
Fr-Sa, Fr-Su, Fr-Mo+1 all check in on Friday but stay 1, 2, 3 nights Th
Sa-Su, Sa-Mo+1 all check in on Saturday but stay 1, 2 nights F
S
Su-Mo+1 checks in on Sunday and stays 1 night Su
A total of 7+6+5+4+3+2+1=7*8/2=28 O-D pairs. Check-out

Over 365 nights 365*366/2=66795 O-D pairs. Standard, deluxe, deluxe with a view, suit rooms, and suppose each
has a fare class of its own. 66795*4=267180 ODFs.
 utdallas.edu/~metin
Page 4
Demand for Flights Vary over Legs
Occupancy

100%

College Station  Dallas Dallas  Los Angeles Los Angeles  Honolulu

Many customers book for College Station  Dallas  Los Angeles  Honolulu trip
 Which booking limits apply to this booking request?
Since requests are coupled over legs, so should the booking limits.
 utdallas.edu/~metin
Page 5
Airline Network: Hub-and-spoke
Little
Las Vegas Rock
Honolulu Fresno

Dallas
LA

Palm Springs College

Station
Austin

Origin-Destination pairs with 10 cities: Houston

College Station  {Dallas, Houston, Austin, Little Rock, Las Vegas, LA, Fresno, Palm Springs, Honolulu}
Fresno  {LA, Honolulu, Palm Springs, Las Vegas, Dallas, Houston, Austin, Little Rock, College Station}
Dallas  {College Station, Houston, Austin, Little Rock, Las Vegas, LA, Fresno, Palm Springs, Honolulu}
Number of pairs is 10*9=90, where Dallas  Austin and Austin  Dallas treated as different legs.
Origin-Destination-Fare (ODF) class Combinations with 10 cities and 6 fares on every flight
Number of ODFs is 90*6=540
Too many ODFs, so joint optimization is a big challenge!
 utdallas.edu/~metin
Page 6
1. Greedy (leg-by-leg decomposition) Heuristic
Since joint optimization is difficult, let us focus on each leg to develop a greedy heuristic.
This is decomposition of the itineraries by force.

Consider the origin destination pair of College Station  Dallas  LA.

 College Station  Dallas leg: Full fare is $200 and discount fare is $120.
 Dallas  LA leg: Full fare is $400 and discount fare is $350.
 Because of the high passenger traffic out of College Station, the discount fare
class on College Station  Dallas leg may be closed.
– Once this class is closed, we cannot accept a discount booking from College
Station to LA which brings in a revenue of $470.
– This revenue of $470 is more than the full fare of $200 for a customer flying
on College Station  Dallas leg.
– Rather displace a full fare customer on College Station  Dallas leg than
reject a discount fare customer on College Station  LA OD pair.

 By considering legs one by one, we never entertain the possibility of protecting seats for
a discount fare passenger that travels multiple legs from a full fare customer that travels
a single leg although multiple leg customer pays (much) more than single leg customer.
 Greedy heuristic does not work; it failed miserably when applied to hotel booking.
 utdallas.edu/~metin
Page 7
2. Linear Programming Approach
 Suppose that the demands are known in advance. Think of assembling resources (legs, night-stays) to
make up products (itineraries, hotel-stays).
 Index products by j=1,2,…,n and resources by i=1,2,…,m. n=540 and m=20 in the simple network.
 pj: price of product j; dj: demand of product j; ci: capacity of resource i;
 All resources
1,11: College Station ,  Dallas; 2,12: Little Rock , Dallas; 3,13: Austin , Dallas;
4,14: Houston , Dallas; 5,15: Dallas , LA; 6,16: Dallas, Las Vegas; 7,17: LA, Las Vegas;
8,18: LA, Fresno; 9,19: LA, Honolulu; 10,20: LA , Palm Springs;
 Some products
– Product 1: College Station  LA discount fare, uses resources 1 and 5 or
resource 1 and 5 are used to make up product 1: a1,1=1, a5,1=1.
– Product 2: Little Rock  Honolulu full fare, uses resources 2, 5 and 9 or
resources 2, 5, and 9 make up product 2: a2,2=1, a5,2=1, a9,2=1.
– Product 3: Las Vegas  Dallas discount fare, uses only resource n
16 or resource 16 is product 3: a16,3=1. max  p j x j
xj
– Product 4: Las Vegas  Little Rock discount fare, uses resources j 1
12 and 16 or resources 12 and 16 make up product 4: a12,4=1, a16,4=1. st (subject to)
– Product 5: Las Vegas  Little Rock full fare, uses resources
n


12 and 16 or resources 12 and 16 make up product 5: a12,5=1, a16,45=1.
aij=1 if resource i is used in product j; 0 otherwise.
a xj 1
ij j  ci for all resources i
 xj=Total seats allocated to product (ODF) j. 0 xj  dj for all products j
 utdallas.edu/~metin
Page 8
Linear Programming Discussion
 Example Solution:
– x4=5 for ODF Las Vegas  Little Rock discount fare. 5 discount fare seats are allocated to this ODF
on resource 12 and 16: Dallas  Little Rock and Las Vegas  Dallas legs.
– x5=10 for ODF Las Vegas  Little Rock full fare. 10 full fare seats are allocated to this ODF on
resource 12 and 16: Dallas  Little Rock and Las Vegas  Dallas legs.
 Observations:
– Different fare classes on the same OD pair consume identical resources.
– Consider two fare classes on the same pair: high-fare class and low-fare class. Optimization implies
» If the high-fare class has no allocation, then the low-fare class will have no allocation.
» If the low-fare class is allocated a capacity equal to all of its demand, then the high-fare class is
allocated a capacity equal to all of its demand.
 In summary, a high-fare class OD pair has higher priority than the low-fare class OD
pair when receiving resource capacities provided that both classes consume the same
capacity.

 Linear programming is an efficient methodology; it is fast to find a solution.

 Linear programming may not be effective
– It assumes that the demands are known.
– It is an allotment (partitioned limits) method as it does not allow for nested classes.
 Question: Can we achieve some nesting, if not all?
 utdallas.edu/~metin
Page 9
3. Virtual Nesting
ODFs ODFs ODFs
Bucket 1 Bucket 1 Bucket 1
Y-class Y-class Y-class

M-class, M-W M-class, M-W M-class, M-W

M-class, M-T M-class, M-T M-class, M-T
Bucket 2 M-class, M-M M-class, M-M Bucket 2 M-class, M-M
Bucket 2
B-class B-class B-class

Monday Tuesday Wednesday

 Virtual Nesting starts by assigning every ODF to a bucket in a resource that is used in the ODF.
– A bucket works like a fare class.
– A bucket is a collection of ODF’s.
– Think of higher buckets as higher customer fare classes where we price the resources higher.
 ODF-to-bucket assignment is called indexing.
 Given this assignment, find the booking limits of an ODF by using EMSR heuristics or another method.
» Virtual Nesting is attributed to American Airlines and developed in 1983.
 The key issue here is indexing.
 utdallas.edu/~metin
Page 10
Indexing: Which ODF into which bucket?
 First attempt: Order ODF by prices and put highest priced ODFs into the highest bucket.
– Problem: High priced ODFs consume more resources
» You pay more if you travel further or if you stay longer
– Classification by ODF price alone ignores the consumption of other sources.
– Discourage ourselves from giving priority to high priced ODFs that consume a lot of resources.

 Inspired from profit margin=price-cost, consider net leg fare of an ODF based on opportunity cost
of capacity:

Net leg fare sum of opportunity costs

of ODF j pj on all resources
on resource i other than resource i.

 Example: An airline estimates the opportunity costs for

a seat on College Station  Dallas to be $80
a seat on Dallas LA to be $210
What are the net leg fares for M-fare class whose College Station LA itinerary is priced at $400?
On College Station  Dallas leg, net leg fare of this itinerary is 400-210=190, use this number when bucketing
College Station  LA itinerary M-fare class (product) on College Station  Dallas leg (resource).
On Dallas LA, net leg fare of this itinerary is 400-80=320, use this number when bucketing College Station 
LA itinerary M-fare class (product) on Dallas  LA leg (resource).
 utdallas.edu/~metin
Page 11
Net Leg Fare Computation for Leg (1,2)
With Two Opportunity Costs
Itinerary Price Opportunity Net leg Bucket
cost fare

Resource Opportu- 3,2,1,2 420 100+100 220 1

Itinerary Price
nity cost 3,2,1,2,3 510 100+100+100 210 1
1,2 100 > 1,2 100 2,1,2 300 100+100 200 1
1,2,1 210 > 2,1 100 1,2,3,2 360 100+100 160 2
1 1,2,3 190 >
2,3 100 1,2,3,2,1 450 100+100+100 150 2
1,2,3,2 360 > 3,2 100 1,2,1 210 100 110 3
1,2,3,2,1 450 >
1,2 100 0 100 3
2,1 180
2 1,2,3 190 100 90 3
2,1,2 300 >
2,3 220 Itinerary Price Opportunity Net leg Bucket
cost fare
2,3,2 300
Resource Opportu- 100+150
3,2,1,2 420 170 1
3,2 100 nity cost
3 3,2,3 210 3,2,1,2,3 510 100+150+100 160 1
1,2 50
1,2,3,2 360 100+100 160 1
3,2,1 190 2,1 150
2,1,2 300 150 150 1
3,2,1,2 420 > 2,3 100
1,2,3,2,1 450 100+100+150 100 2
3,2,1,2,3 510 > 3,2 100
1,2 100 0 100 2
Bottom tables with altered
opportunity costs, to see 1,2,3 190 100 90 2
the effect of these alterations
on the net leg fare 1,2,1 210 150 60 3
 utdallas.edu/~metin
Page 12
Indexing with
Net leg fares on Houston Dallas leg (Resource)

Honolulu Fresno Little Rock

Dallas
Palm Springs LA
Austin Houston

Itinerary: Origin-Destination Net leg fare Bucket

Houston Fresno 180 1
Houston Little Rock 120 1
Houston Austin 110 1
Houston Dallas 100 2
Houston Palm Springs 85 2
Houston Honolulu 40 3
Houston LA 30 3
 utdallas.edu/~metin
Page 13
Indexing with
Net leg fares on Dallas  LA leg
Fresno
Little Rock
Honolulu Houston
Palm Springs LA Dallas
Austin

Itinerary: Origin-Destination Net leg fare Bucket

 Houston-Honolulu ODF is
Houston Fresno 160 1
in bucket 1 on Dallas LA flight, but
Austin Honolulu 160 1 in bucket 3 on Houston  Dallas flight.
Houston Honolulu 140 1
Austin LA 110 1  This is possible when
DallasLA flight has a high opportunity cost.
Austin Fresno 90 2
Little Rock Fresno 85 2 • Recall that DallasLA opportunity cost is
Houston Palm Springs 80 2 not deducted from the leg fare of DallasLA but is
Austin Palm Springs 60 2 deducted from the leg fare of HoustonDallas.
Little Rock Honolulu 40 3
Inconsistency or not:
Dallas LA 30 3 As a result of these, if you are going from Houston to
Little Rock Palm Springs 25 3 Honolulu, the system can find the DallasLA ticket
but not the HoustonDallas ticket.
Little Rock LA 25 3
The funny thing is that your friend can buy a
Houston LA 20 3 HoustonDallas ticket after you!!
 utdallas.edu/~metin
Page 14
Booking with Virtual Nesting: Single Class Illustration
1 Check availability on corresponding
2 3
virtual classes on each involved resource
After virtual indexing is finished, Booking limit vector 𝑏 =
List the O-D pairs on each resource from the highest indexed to the lowest [280, 120; 280, 210; 170, 60; 170, 140]
Booking vector 𝐵 =
Resource 1 → 2 Resource 3 → 2 [192, 118; 228, 123; 142, 59; 30, 16]
O→D Booking Limit O→D Booking Limit A booking request for going from 3 to 1
1→2 C(1→2) 3→1 C(3→2) 𝑥 = [0, 0; 1, 1; 1, 0; 0, 0], 𝑏 ≥ 𝐵 + 𝑥 accept
𝐵 = [192, 118; 229, 124; 143, 59; 30, 16]
1→3 𝑏1→3 3→2 𝑏3→2
A booking request for going from 3 to 2
Resource 2 → 1 Resource 2 → 3
𝑥 = [0, 0; 0, 0; 1, 1; 0, 0], 𝑏 ≥ 𝐵 + 𝑥 accept
O→D Booking Limit O→D Booking Limit 𝐵 = [192, 118; 229, 124; 144, 60; 30, 16]

2→1 C(2→1) 2→3 C(2→3) A booking request for going from 3 to 2

3→1 𝑏3→1 1→3 𝑏′1→3 𝑥 = [0, 0; 0, 0; 1, 1; 0, 0], 𝑏 < 𝐵 + 𝑥 reject
𝐵 = [192, 118; 229, 124; 144, 60; 30, 16]
1→2 2→1 3→2 2→3 A booking request for going from 3 to 1
1→2 1→3 2→1 3→1 3→1 3→2 2→3 1→3 𝑥 = [0, 0; 1, 1; 1, 0; 0, 0], 𝑏 ≥ 𝐵 + 𝑥 accept
𝐵 = [192, 118; 230, 125; 145, 60; 30, 16]
C(1→2) 𝑏1→3 C(2→1) 𝑏3→1 C(3→2) 𝑏3→2 C(2→3) 𝑏′1→3
Puzzling for outsiders: (3→2) is unavailable
280 120 280 210 170 60 170 140
but (3→1) is, despite (3→1)=(3→2)+ (2→1)
 utdallas.edu/~metin
Page 15
Dynamic Opportunity Cost of Capacity
 If we have a network with 8 legs in a network and bi denotes the capacity on leg (O-D) i while
П(b) : Maximum Profit to be made from the network with capacity vector b=[b1,b2,b3,b4,b5,b6,b7,b8].
 The marginal cost of accepting an ODF j on leg i is d
aij  (b, t )  aij ( (b, t )   (b  1, t ))
dbi
Recall aij=1 if leg (resource) i is used in ODF (product) j.
d d d
 Before accepting an ODF 2 that uses legs 3,4 and 7, we consider (b, t ), (b, t ), (b, t ).
db3 db4 db7
Because only a32=a42=a72=1 for this ODF and others are 0. These computations are conceptually same as in single-leg.
 However, there are complexity and numerical issues when this type of computation is done in a network
 that includes hundreds and sometimes thousands of legs (resources),
 whose remaining capacity b changes dynamically over time t.
 If computation is not a concern, directly compute marginal revenue for accepting product 𝑗.
 Make up vector 𝑎𝑗 of the same size as 𝑏, the number of legs. The ith element in the vector 𝑎𝑗 is 1 if 𝑎𝑖𝑗 = 1.
 Ex: 𝑎2 = [0, 0, 1, 1, 0, 0, 1, 0] for product 2 that uses resources 3,4 and 7
 If 𝑏 − 𝑎𝑗 ≥ 0, exact marginal opportunity cost of capacity Π 𝑏, 𝑡 − Π(𝑏 − 𝑎𝑗 , 𝑡), there are 𝑛 of these; 𝑛 ≫ 𝑚.
 If concerned about computations, for each resource 𝑖 used in product 𝑗, make up vector 1𝑖𝑗 of the same size as 𝑏, the
number of legs. Only the ith element in the vector 1𝑖𝑗 has 1 all others are zero.
 Ex: 132 = [0, 0, 1, 0, 0, 0, 0, 0], 142 = [0, 0, 0, 1, 0, 0, 0, 0], 172 = [0, 0, 0, 0, 0, 0, 1, 0] so 𝑎2 = 132 + 142 + 172
 If 𝑏 − 𝑎𝑗 ≥ 0, approximate marginal opportunity cost of capacity 𝑎𝑖𝑗 =1 Π 𝑏, 𝑡 − Π(𝑏 − 1𝑖𝑗 , 𝑡)
For approximation, compute each Π 𝑏, 𝑡 − Π(𝑏 − 1𝑖𝑗 , 𝑡) in advance and keep in the memory, which requires a
space of number of resources m. In real time, perform only the sum.
 utdallas.edu/~metin
Page 16
4. Additive Bid Prices
Little
Las Vegas Rock
Honolulu Fresno
$180 $170
$210
$90
$290
$280 Dallas
$150
LA
$120
Palm Springs $100
$110 College
Station
Austin
Houston
E.g., Dallas – Houston = 100 = Π … , 𝑏𝐷𝐻 , … − Π … , 𝑏𝐷𝐻 − 1, …
Would you close a fare class on
Las Vegas  Fresno itinerary if the price is $350?
Little Rock  Honolulu itinerary if the price is $800?
Palm Springs  Austin itinerary if the price is $400?

Skip the section on calculation of the bid prices in the book. Recall that bid prices are about opportunity costs.
Just like opportunity costs, bid prices must be dynamic. Applications in Airlines (Scandinavian Airline System)
and Rental Cars (Hertz).
 utdallas.edu/~metin
Page 17
Virtual Nesting or Bid Prices
 Virtual nesting and bid prices both depend on the opportunity cost of the capacity.
 The opportunity cost needs to be updated dynamically as time passes and remaining
capacity drops.
– Both virtual nesting and bid pricing methods perform similarly when opportunity costs are updated
frequently.
 If the bid prices are not updated and
demand is stronger than expected, bid price becomes lower than what it should be (underpricing),
demand is weaker than expected, bid price become higher than what it should be (overpricing).
– Ignorance of dynamism is more costly when bid pricing methods are used.

 Historically, airlines had RM before hotels for single sources. Airlines used fare class
based booking controls.
 When network RM management came out, the legacy systems at Airlines are modified
to take legs of an OD into account. This led to virtual nesting.
 On the other hand, Hotels did not have any (legacy) RM systems so they started with bid
pricing systems.
 Over time, more airlines adopted bid pricing.
 utdallas.edu/~metin
Page 18
Practice as of 2011
Dynamic Virtual Nesting or Bid Prices
Dynamic Virtual Nesting Bid Prices
– United Airlines – American Airlines
– Delta Airlines – Lufthansa/Swiss
– LAN
– Iberia
– British Airways
– Air France/KLM
– SAS
– Cathay Pacific
– Qatar
– Etihad
– Royal Jordanian
– Thai
“Some of these airlines are using what we refer to as ‘hybrid’ controls, whereby leg authorization limits
are used to manage local services, and bid prices are used for connecting ones. That particular
variation of bid price controls retains aspects of leg authorizations levels (for RM analysts who are
more familiar with these types of controls), but it enables bid prices for connecting services to
avoid the huge increase in the total volume of controls required in an O&D implementation.”
Richard Ratliff , Senior Research Scientist, Sabre, Nov 2011
Lists above are based on his Nov 21 Demreman guest lecture.
 utdallas.edu/~metin
Page 19
Summary

 Network Management Problem

 Greedy Heuristic
 LP Approach
 Virtual Nesting
 Bid Prices
 utdallas.edu/~metin
Page 20
Net Leg Fare Computation for Product (1,2)
Deducting the opportunity cost of (1,2) as well
Itinerary Price Resource Original
opportu- Itinerary Price Original Net leg Bucket
1,2 100 > Opportunity cost
nity cost fare
1,2,1 210 > 1,2 100 100+100+100
3,2,1,2 420 120 1
1 1,2,3 190 > 2,1 100 3,2,1,2,3 510 100+100+100+100 110 1
1,2,3,2 360 > 2,3 100 2,1,2 300 100+100 100 1
1,2,3,2,1 450 > 3,2 100 100+100+100
1,2,3,2 360 60 2
2,1 180
1,2,3,2,1 450 100+100+100+100 50 2
2 2,1,2 300 > 100+100
1,2,1 210 10 3
2,3 220
Resource Altered 1,2 100 100 0 3
2,3,2 300 opportu-
1,2,3 190 100+100 -10 *
3,2 100 nity cost
Itinerary Price Altered Net leg Bucket
3,2,3 210 1,2 50 Opportunity cost
3 fare
3,2,1 190 2,1 150
3,2,1,2 420 100+150+50 120 1
3,2,1,2 420 > 2,3 100
3,2,1,2,3 510 100+150+50+100 110 1
3,2,1,2,3 510 > 3,2 100
1,2,3,2 360 50+100+100 110 1
Deduction of the opportunity cost of (1,2) does 2,1,2 300 150+50 100 1
not change prioritization of itineraries or bucketing. 50+100+100+150
1,2,3,2,1 450 50 2
Buckets with original Buckets with altered
opportunity costs opportunity costs 1,2 100 50 50 2

3,2,1,2; 3,2,1,2,3; 2,1,2 3,2,1,2; 3,2,1,2,3; 2,1,2; 1,2,3,2 1,2,3 190 50+100 40 2

1,2,3,2; 1,2,3,2,1 1,2,3,2,1; 1,2; 1,2,3 1,2,1 210 150+50 10 3

1,2,1; 1,2; 1,2,3. 1,2,1.

 utdallas.edu/~metin
Page 21
Net Leg Fare Computation for Product (3,2)
Deducting the opportunity cost of (3,2) as well
Itinerary Price Itinerary Price Net leg Bucket
Resource Original
1,2 100 opportu- fare
nity cost 3,2,1,2 420 120 1
1,2,1 210
1,2 100 3,2,1,2,3 510 110 1
1 1,2,3 190
2,1 100 2,3,2 300 100 1
1,2,3,2 360 >
2,3 100 1,2,3,2 360 60 2
1,2,3,2,1 450 >
3,2 100 1,2,3,2,1 450 50 2
2,1 180
3,2,3 210 10 3
2 2,1,2 300
3,2 100 0 3
2,3 220 Resource Altered
opportu- 3,2,1 190 -10 *
2,3,2 300 > nity cost
3,2 100 > Itinerary Price Net leg Bucket
1,2 50
3 3,2,3 210 > fare
2,1 150
3,2,1 190 > 3,2,1,2 420 120 1
2,3 100
3,2,1,2 420 > 3,2,1,2,3 510 110 1
3,2 100
3,2,1,2,3 510 > 1,2,3,2 360 110 1
2,3,2 300 100 1
Buckets with original Buckets with altered
1,2,3,2,1 450 50 2
opportunity costs opportunity costs
3,2,1,2; 3,2,1,2,3; 2,3,2 3,2,1,2; 3,2,1,2,3; 1,2,3,2; 2,3,2 3,2,3 210 10 3

1,2,3,2; 1,2,3,2,1 1,2,3,2,1; 3,2 100 0 3

3,2,3; 3,2; 3,2,1. 3,2,3; 3,2; 3,2,1. 3,2,1 190 -60 *

 utdallas.edu/~metin
Page 22
Net Leg Fare Computation for Product (2,1)
Deducting the opportunity cost of (2,1) as well
Itinerary Price Itinerary Price Net leg Bucket
Resource Original fare
1,2 100 opportu-
nity cost 3,2,1,2 420 120 1
1,2,1 210 >
1 1,2 100 3,2,1,2,3 510 110 1
1,2,3 190
2,1 100 2,1,2 300 100 1
1,2,3,2 360
2,3 100 2,1 180 80 1
1,2,3,2,1 450 >
3,2 100 1,2,3,2,1 450 50 2
2,1 180 >
2 1,2,1 210 10 3
2,1,2 300 >
3,2,1 190 -10 *
2,3 220 Resource Altered
opportu- Itinerary Price Net leg Bucket
2,3,2 300
nity cost fare
3,2 100
1,2 50 3,2,1,2 420 120 1
3 3,2,3 210
2,1 150 3,2,1,2,3 510 110 1
3,2,1 190 > 2,3 100 2,1,2 300 100 1
3,2,1,2 420 >
3,2 100 1,2,3,2,1 450 50 2
3,2,1,2,3 510 >
2,1 180 30 2

Buckets with original Buckets with altered 1,2,1 210 10 3

opportunity costs opportunity costs
3,2,1 190 -60 *
3,2,1,2; 3,2,1,2,3; 2,1,2; 2,1 3,2,1,2; 3,2,1,2,3; 2,1,2
1,2,3,2,1 1,2,3,2,1; 2,1
1,2,1; 3,2,1. 1,2,1; 3,2,1.
 utdallas.edu/~metin
Page 23
Net Leg Fare Computation for Product (2,3)
Deducting the opportunity cost of (2,3) as well
Itinerary Price Itinerary Price Net leg Bucket
Resource Original fare
1,2 100 opportu-
nity cost 2,3 220 120 1
1,2,1 210
1 1,2 100 3,2,1,2,3 510 110 1
1,2,3 190 >
2,1 100 2,3,2 300 100 1
1,2,3,2 360 >
2,3 100 1,2,3,2 360 60 2
1,2,3,2,1 450 >
3,2 100 1,2,3,2,1 450 50 2
2,1 180
2 3,2,3 210 10 3
2,1,2 300
1,2,3 190 -10 *
2,3 220 > Resource Altered
2,3,2 300 > opportu-
nity cost Itinerary Price Net leg Bucket
3,2 100 fare
3 1,2 50
3,2,3 210 > 2,3 220 120 1
2,1 150
3,2,1 190 3,2,1,2,3 510 110 1
2,3 100
3,2,1,2 420 1,2,3,2 360 110 1
3,2 100
3,2,1,2,3 510 > 2,3,2 300 100 1
1,2,3,2,1 450 50 2
Buckets with original Buckets with altered 1,2,3 190 40 2
opportunity costs opportunity costs
3,2,3 210 10 3
2,3; 3,2,1,2,3; 2,3,2; 2,3; 3,2,1,2,3; 1,2,3,2; 2,3,2;
1,2,3,2; 1,2,3,2,1; 1,2,3,2,1; 1,2,3;
3,2,3; 3,2,1. 3,2,3
 utdallas.edu/~metin
Page 24
Net Leg Fare Computation Results

1 2 3 2 1

Buckets with original Buckets with original Buckets with original Buckets with original
opportunity costs opportunity costs opportunity costs opportunity costs
3,2,1,2; 3,2,1,2,3; 2,1,2 2,3; 3,2,1,2,3; 2,3,2; 3,2,1,2; 3,2,1,2,3; 2,3,2 3,2,1,2; 3,2,1,2,3; 2,1,2; 2,1
1,2,3,2; 1,2,3,2,1 1,2,3,2; 1,2,3,2,1; 1,2,3,2; 1,2,3,2,1 1,2,3,2,1
1,2,1; 1,2; 1,2,3. 3,2,3; 3,2,1. 3,2,3; 3,2; 3,2,1. 1,2,1; 3,2,1.

Buckets with altered opportunity Buckets with altered Buckets with altered opportunity Buckets with altered
costs opportunity costs costs opportunity costs
3,2,1,2; 3,2,1,2,3; 2,1,2; 1,2,3,2 2,3; 3,2,1,2,3; 1,2,3,2; 2,3,2; 3,2,1,2; 3,2,1,2,3; 1,2,3,2; 2,3,2 3,2,1,2; 3,2,1,2,3; 2,1,2
1,2,3,2,1; 1,2; 1,2,3 1,2,3,2,1; 1,2,3; 1,2,3,2,1; 1,2,3,2,1; 2,1
1,2,1. 3,2,3 3,2,3; 3,2; 3,2,1. 1,2,1; 3,2,1.