Module # 4

Capacity Allocation

l
ia
er
at
M
4 se
-2 ur
22 Co
M RI
,B L
M of X
Course: Service Operations Management Shounak Basak
SO rt
Pa
Programme: BM 22-24 Room no. 6, Library Building,
XLRI – Xavier School of Management,
Term VI, 2024 Jamshedpur – 831001
Jamshed1
Phone: 0657 – 665 – 3424
Email: shounak@[Link]

Shounak Basak
PODS Area, XLRI – Xavier School of Management
 l
ia
er
at
M
4 se
Capacity allocation

-2 ur
22 Co
M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
2
Capacity allocation problem
Example# 4.1

Consider a flight with a capacity of C. If the number of discounted seats are 𝑏. 𝐹𝑑 𝑥 is the probability that discount demand is less
than or equal to 𝑥. 𝐹𝑓 𝑥 is the probability that full price demand is less than or equal to 𝑥. Derive the condition for changing the

l
ia
booking limit from 𝑏 to 𝑏 + 1.

er
at
M
4 se
-2 ur
22 Co
M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
3
Capacity allocation problem solution
𝐹𝑑 𝑏
𝑑𝑑 ≤ 𝑏 0

l
ia
er
1 − 𝐹𝑓 𝐶 − 𝑏

at
𝑝𝑑 − 𝑝𝑓

M
𝑑𝑓 > 𝐶 − 𝑏

4 se
-2 ur
1 − 𝐹𝑑 𝑏

22 Co
𝑏 →𝑏+1
𝑑𝑑 > 𝑏

M RI
,B L
M of X
𝐹𝑓 𝐶 − 𝑏 𝑝𝑑
SO rt
Pa
𝑑𝑓 ≤ 𝐶 − 𝑏

Hold 0
𝑏 constant
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
4
Price differentiation problem solution

Expected change in returns due to increase in discounted seat:

l
ia
𝐸ℎ 𝑏 = 𝐹𝑑 𝑏 . 0 + 1 − 𝐹𝑑 𝑏 1 − 𝐹𝑓 𝐶 − 𝑏 𝑝𝑑 − 𝑝𝑓 + 𝐹𝑓 𝐶 − 𝑏 𝑝𝑑 = 1 − 𝐹𝑑 𝑏 {𝑝𝑑 − 1 − 𝐹𝑓 𝐶 − 𝑏 𝑝𝑓 }

er
at
M
4 se
-2 ur
22 Co
M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
5
Capacity allocation problem
Problem# 4.2

Consider you are the revenue operations manager for Inbigo airlines. The airlines uses A320 airplanes in an all economy configuration
with 30 rows of 6 seats with 3 on each side of a single aisle. You are planning for the upcoming off-season ticket prices on the Kolkata

l
ia
er
(CCU) to Mumbai (BOM) route. Adding discounts would induce passengers to leave Duronto tickets and shift to air travel. However,

at
there is an available customer pool for business travel who would avail the facilities.

M
4 se
Currently, the full fare is priced at Rs. 8500 and you plan to offer a discount fare of Rs. 5500. How many seats do you plan to keep for

-2 ur
discount customers?

22 Co
M RI
From various surveys and past data, it is known that number of passengers availing full fare follow a normal probability distribution

,B L
with mean 150 and standard deviation 10 and number of passengers availing discount fare follow a normal probability distribution

M of X
with mean 50 and standard deviation 5.

SO rt
Pa
Discount customers book their tickets earlier than full fare customers

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
6
LittLewood’s rule

l
ia
er
Algorithm for finding optimal discount seats:

at
M
4 se
-2 ur
1. Set b=0

22 Co
If b=C, set 𝑏 ∗ = 𝐶 and stop, otherwise go to step 3

M RI
2.

,B L
M of X
3. Calculate 𝐸 ℎ 𝑏 = 𝑝𝑑 − 1 − 𝐹𝑓 𝐶 − 𝑏 𝑝𝑓
4. SO rt
If 𝐸 ℎ 𝑏 ≤ 0 𝑜𝑟 𝐹𝑑 𝑏 + 1 = 1, set 𝑏 ∗ = 𝑏 and stop
Pa
5. If 𝐸 ℎ 𝑏 > 0 𝑎𝑛𝑑 𝐹𝑑 𝑏 + 1 < 1, set 𝑏 = 𝑏 + 1 and go to step 2

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
7
Capacity allocation

l
ia
er
at
M
4 se
-2 ur
Excel Solution

22 Co
M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
8
LittLewood’s rule

l
ia
er
at
M
The optimal point can also be found by this formula:

4 se
-2 ur
22 Co
1 − 𝐹𝑓 𝐶 − 𝑏 ∗ = 𝑝𝑑 /𝑝𝑓

M RI
,B L
M of X
SO rt
Where, 𝑏 ∗ is the optimal number of discount seat
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
9
LittLewood’s rule

l
ia
The optimal point is:

er
at
M
4 se
𝑝𝑑
∗
𝐹𝑓−1

-2 ur
𝑏 = C − min 1− ,𝐶

22 Co
𝑝𝑓

M RI
+

,B L
𝑝𝑑

M of X
= 𝐶− 𝐹𝑓−1 1−
𝑝𝑓
SO rt
Pa

Where, 𝑏 ∗ is the optimal number of discount seat

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
10
Capacity allocation problem
Problem# 4.2

Consider you are the revenue operations manager for Inbigo airlines. The airlines uses A320 airplanes in an all economy configuration
with 30 rows of 6 seats with 3 on each side of a single aisle. You are planning for the upcoming off-season ticket prices on the Kolkata

l
ia
er
(CCU) to Mumbai (BOM) route. Adding discounts would induce passengers to leave Duronto tickets and shift to air travel. However,

at
there is an available customer pool for business travel who would avail the facilities.

M
4 se
Currently, the full fare is priced at Rs. 8500 and you plan to offer a discount fare of Rs. 5500. How many seats do you plan to keep for

-2 ur
discount customers? Solve it using Littlewood’s formula

22 Co
M RI
From various surveys and past data, it is known that number of passengers availing full fare follow a normal probability distribution

,B L
with mean 150 and standard deviation 10 and number of passengers availing discount fare follow a normal probability distribution

M of X
with mean 50 and standard deviation 5.

SO rt
Pa
Discount customers book their tickets earlier than full fare customers

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
11
LittLewood’s rule

l
ia
It can be inferred:

er
at
M
4 se
𝑝𝑑
When, ≤ 0.5, it is optimal to protect more seats or provide lesser

-2 ur
𝑝𝑓

22 Co
M RI
discount seats

,B L
M of X
𝑝𝑑
When, SO rt
≥ 0.5, it is optimal to protect lesser seats or provide more
Pa
𝑝𝑓
discount seats

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
12
LittLewood’s rule

It can be inferred:

l
ia
𝑝𝑑

er
When, < 0.5, 𝐶 − 𝑏 ∗ is greater than mean full fare demand 𝜇 and increases

at
𝑝𝑓

M
4 se
with standard deviation of full fare demand

-2 ur
22 Co
M RI
𝑝𝑑
When, > 0.5, 𝐶 − 𝑏 ∗ is lesser than mean full fare demand 𝜇 and decreases

,B L
M of X
𝑝𝑓

SO rt
with standard deviation of full fare demand
Pa

𝑝𝑑
When, = 0.5, 𝐶 − 𝑏 ∗ is equal to mean full fare demand 𝜇 and does not
𝑝𝑓
change with standard deviation of full fare demand
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
13
Multi-class Allocation problem

l
ia
er
at
M
Airlines would not limit themselves to two class discount allocation problem

4 se
-2 ur
22 Co
M RI
Multi-class allocation problem:

,B L
M of X
Multiple prices are offered at different time periods
SO rt
These multiple time-periods follow an order of discount
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
14
Multi-class Allocation problem

l
ia
Assumptions:

er
• There are 𝑛 fare classes, each with an associated fare. Fare classes are numbered in descending order,

at
M
that 𝑝1 > 𝑝2 > 𝑝3 > … 𝑝𝑛 . A fare class is called higher than another if it has a higher fare.

4 se
-2 ur
• Demand in each class is an independent random variable. We denote the demand in fare class 𝑖 by 𝑑𝑖 .

22 Co
Demand in class 𝑖 follows a probability distribution 𝑓𝑖 𝑥 , where 𝐹𝑖 𝑥 denotes the probability 𝑑𝑖 ≤ 𝑥

M RI
,B L
M of X
• Demand books in increasing order fare order. That is lowest fare 𝑝𝑛 first, followed by 𝑝𝑛−1 and so on.

SO rt
Pa
• There are no no-shows or cancellations

• The objective function is to maximize expected revenue (as variable cost is almost negligible)

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
15
Multi-class Allocation problem

Departure

l
ia
er
at
M
Period Period Period Period Period Period

4 se
n n-1 n-2 3 2 1

-2 ur
22 Co
M RI
Fare Fare
Fare 𝑝𝑛 Fare 𝑝3 Fare 𝑝2 Fare 𝑝1

,B L
𝑝𝑛−1 𝑝𝑛−2

M of X
SO rt
Pa
Bookings Bookings Bookings Bookings Bookings Bookings
𝑥𝑛 𝑥𝑛−1 𝑥𝑛−2 𝑥3 𝑥2 𝑥1

Low fare bookings High fare bookings

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
16
Multi-class Allocation problem

l
ia
er
at
M
4 se
-2 ur
Question:

22 Co
M RI
How many bookings it should accept in 𝑗𝑡ℎ class given it knows how many bookings it has accepted in

,B L
M of X
𝑗 + 1 𝑡ℎ to 𝑛𝑡ℎ class

SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
17
Multi-class Allocation problem

l
ia
er
Capacity available for the 𝑗𝑡ℎ 𝑐𝑙𝑎𝑠𝑠:

at
M
4 se
-2 ur
22 Co
𝑡ℎ 𝑛
to 𝑛𝑡ℎ class = Σ𝑖=𝑗+1

M RI
Total bookings it has accepted in 𝑗 + 1 𝑥𝑖

,B L
M of X
𝑛
Unbooked capacity 𝐶𝑗 = 𝐶 − Σ𝑖=𝑗+1 𝑥𝑖
SO rt
Pa
Booking limit is 𝑏𝑗 (maximum number of seats in that fare class that the airline plans to offer)

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
18
Solution approach

Decision tree Approach:

l
ia
• We start from the first period that is the period just before departure

er
at
M
• Since, there are no no-shows and cancellations, bookings 𝑏1 = 𝐶1

4 se
-2 ur
22 Co
• In the penultimate period (period 2), airline has 𝐶2 capacity left

M RI
,B L
• The problem at the beginning of Period 2 is a two fare class problem

M of X
SO rt
Pa 𝑝𝑑
• The optimal booking limit is 0 ≤ 𝑏2 ≤ 𝐶2 and is given by: 𝐶2 − 𝑏2 < 𝐹 −1 1 −
𝑝𝑓
𝑝2
⇒ 𝐹 −1 1−
𝑝1

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
19
 Solution approach
Decision tree Approach:

• In the third period, a similar approach is followed and a decision tree is made:

l
ia
er
𝐹3 𝑏3

at
0
𝑞1

M
𝑑3 ≤ 𝑏3

4 se
Displace Class 1

-2 ur
booking 𝑝3 − 𝑝1

22 Co
M RI
𝑞2

,B L
M of X
1 − 𝐹3 𝑏3 Displace Class 2 𝑝3 − 𝑝2
𝑏3 → 𝑏3 + 1
𝑑3 > 𝑏 booking
SO rt
Pa
1 − 𝑞1 − 𝑞2 𝑝3
No
displacement
Hold
0
𝑏3 constant
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
20
Solution approach

l
ia
er
at
M
4 se
-2 ur
• If 𝑞1 and 𝑞2 were known , then the solution could be found straightforward

22 Co
M RI
• But, as it is not known, dynamic programming is used to solve these kind of problems

,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
21
Dynamic programming approach

Dynamic programming is defined by 3 essential characteristics:

l
ia
er
at
State: 𝑠𝑖

M
Decision Variable: 𝑥𝑖

4 se
Objective function: 𝑓𝑖 (𝑠𝑖 , 𝑥𝑖 )

-2 ur
22 Co
At any stage 𝑖, 𝑣𝑖 𝑠𝑖 = max(𝑓𝑖 𝑠𝑖 , 𝑥𝑖 + 𝑓𝑖−1 𝑠𝑖−1 , 𝑥𝑖−1 + 𝑓𝑖−2 𝑠𝑖−2 , 𝑥𝑖−2 + ⋯ + 𝑓1 𝑠1 , 𝑥1 )

M RI
,B L
M of X
𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡: 𝑠𝑗−1 = 𝑡𝑗 𝑠𝑗 , 𝑥𝑗 ∀ 𝑗 ∈ {𝑖, 𝑖 − 1, 𝑖 − 2, … 1}
𝑥𝑗 ∈ 𝑋𝑗
SO rt ∀𝑗 ∈ {𝑖, 𝑖 − 1, 𝑖 − 2, … 1}
Pa

Where, 𝑋𝑗 is the set of all permissible decision variable values in stage 𝑗

And 𝑠𝑗 ∈ {𝑠𝑖 , 𝑠𝑖−1 , 𝑠𝑖−2 , … 𝑠1 , 𝑠0 }
𝑡𝑗 is the transition function

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
22
Dynamic programming approach

l
ia
Let there be 𝑛 periods of booking

er
at
M
Therefore, there are 𝑛 stages in the dynamic programming formulation

4 se
-2 ur
22 Co
𝑥𝑖 : amount of booking to be made for stage 𝑖

M RI
,B L
M of X
𝑠𝑖 : available capacity before stage 𝑖

SO rt
Pa
𝑓𝑖 : profit to be maximized at stage 𝑖

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
23
Multi-class Allocation problem

Departure

l
ia
er
at
M
Period Period Period Period Period Period

4 se
n n-1 n-2 3 2 1

-2 ur
22 Co
M RI
Fare Fare
Fare 𝑝𝑛 Fare 𝑝3 Fare 𝑝2 Fare 𝑝1

,B L
𝑝𝑛−1 𝑝𝑛−2

M of X
SO rt
Pa
Bookings Bookings Bookings Bookings Bookings Bookings
𝑥𝑛 𝑥𝑛−1 𝑥𝑛−2 𝑥3 𝑥2 𝑥1

Low fare bookings High fare bookings

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
24
Multi-class Allocation problem

Departure

l
ia
er
at
Stage Stage Stage Stage Stage Stage

M
4 se
n n-1 n-2 3 2 1

-2 ur
22 Co
M RI
Fare Fare
Fare 𝑝𝑛 Fare 𝑝3 Fare 𝑝2 Fare 𝑝1

,B L
𝑝𝑛−1 𝑝𝑛−2

M of X
SO rt
Pa
Bookings Bookings Bookings Bookings Bookings Bookings
𝑥𝑛 𝑥𝑛−1 𝑥𝑛−2 𝑥3 𝑥2 𝑥1

Low fare bookings High fare bookings

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
25
Dynamic programming approach

l
ia
er
at
At any stage 𝑖,

M
4 se
-2 ur
Initial State: 𝑠𝑖

22 Co
Decision Variable: 𝑥𝑖

M RI
Objective function: 𝑓𝑖 (𝑠𝑖 , 𝑥𝑖 )

,B L
M of X
Final state: 𝑠𝑖−1

SO rt
Pa
Transition equation: 𝑠𝑖−1 = 𝑡𝑖 𝑠𝑖 , 𝑥𝑖

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
26
Dynamic programming approach

At any stage 𝑖,

l
ia
Initial State: 𝑠𝑖 : available capacity before stage 𝑖

er
Decision Variable: 𝑥𝑖 : amount of booking to be made for stage 𝑖

at
M
Final state: 𝑠𝑖−1 : available capacity after stage 𝑖

4 se
-2 ur
Objective function: 𝑓𝑖 𝑠𝑖 , 𝑥𝑖 = 𝑝𝑖 𝐸෨𝑖

22 Co
Transition equation:𝑠𝑖−1 = 𝑡𝑖 𝑠𝑖 , 𝑥𝑖 = 𝑠𝑖 − 𝐸෨𝑖

M RI
,B L
At any stage 𝑖, 𝑣𝑖 𝑠𝑖 = max(𝑓𝑖 𝑠𝑖 , 𝑥𝑖 + 𝑓𝑖−1 𝑠𝑖−1 , 𝑥𝑖−1 + 𝑓𝑖−2 𝑠𝑖−2 , 𝑥𝑖−2 + ⋯ + 𝑓1 𝑠1 , 𝑥1 )

M of X
SO rt
𝑠𝑢𝑐ℎ 𝑡ℎ𝑎𝑡: 𝑠𝑗−1 = 𝑡𝑗 𝑠𝑗 , 𝑥𝑗 ∀ 𝑗 ∈ {𝑖, 𝑖 − 1, 𝑖 − 2, … 1}
Pa
𝑥𝑗 ≤ 𝑠𝑗 ∀ 𝑗 ∈ {𝑖, 𝑖 − 1, 𝑖 − 2, … 1}
𝑥𝑗 ∈ 𝑋𝑗 ∀𝑗 ∈ {𝑖, 𝑖 − 1, 𝑖 − 2, … 1}

Where, 𝑋𝑗 is the set of all permissible decision variable values in stage 𝑗

And 𝑠𝑗 ∈ {𝑠𝑖 , 𝑠𝑖−1 , 𝑠𝑖−2 , … 𝑠1 , 𝑠0 }
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
27
Dynamic programming approach

l
ia
er
at
M
4 se
-2 ur
22 Co
We solve dynamic programming problem by backward induction based recursive methodology

M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
28
Dynamic programming approach

𝑥𝑖
𝐸 𝑓𝑖 𝑠𝑖 , 𝑥𝑖 = 𝐸 𝑝𝑖 𝐸෨𝑖 𝑥𝑖 = 𝑝𝑖 𝐸 𝐸෨𝑖 = 𝑝𝑖 𝑥𝑖 1 − 𝐹𝑖 𝑥𝑖 ෩𝑖 𝑓𝑖 𝐷
+න 𝐷 ෩𝑖 𝑑 𝐷
෩𝑖

l
ia
0

er
at
M
𝑥1
෪1 = 𝑝1 𝑥1 1 − 𝐹1 𝑥1 ෩1 𝑓1 𝐷
෩1 𝑑 𝐷
෩1

4 se
𝐸 𝑓1 𝑠1 , 𝑥1 = 𝐸 𝑝1 𝐸 +න 𝐷

-2 ur
0

22 Co
M RI
𝑣1 𝑠1 = max 𝑓1 𝑠1 , 𝑥1

,B L
M of X
𝑥1
= max 𝑝1 𝑥1 1 − 𝐹1 𝑥1 ෩1 𝑓1 𝐷
+න 𝐷 ෩1 𝑑 𝐷
෩1

SO rt
𝑥1 0 Pa

Such that:
𝑥1 ≤ 𝑠1
𝑥1 ≥ 0

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
29
Dynamic programming approach

l
With first order conditions as:

ia
er
𝜕ℒ

at
= 𝑝1 1 − 𝐹1 𝑥1 − ℓ1 + ℓ2 = 0

M
𝜕𝑥1

4 se
ℓ1 𝑠1 − 𝑥1 = 0

-2 ur
ℓ2 (𝑥1 ) = 0

22 Co
ℓ1 ≥ 0, ℓ2 ≥ 0, 𝑠1 − 𝑥1 ≥ 0, 𝑥1 ≥ 0

M RI
,B L
M of X
Solution:

SO rt
𝑠1 … 𝑖𝑓 𝑝 ≥ 0
Pa
𝑥1 = ቊ
0 … 𝑖𝑓 𝑝 < 0

This satisfies all the cases mentioned above

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
30
Dynamic programming approach

𝑣𝑖 𝑠𝑖 = max(𝑓𝑖 𝑠𝑖 , 𝑥𝑖 + 𝑓𝑖−1 𝑠𝑖−1 , 𝑥𝑖−1 + 𝑓𝑖−2 𝑠𝑖−2 , 𝑥𝑖−2 + ⋯ + 𝑓1 𝑠1 , 𝑥1 )

l
ia
er
𝑣2 𝑠2 = max(𝑓2 𝑠2 , 𝑥2 + 𝑓1 (𝑠1 , 𝑥1 )))

at
M
4 se
𝑓𝑖 𝑠𝑖 , 𝑥𝑖 = 𝑝𝑖 𝐸෨𝑖 𝑥𝑖 = 𝑝𝑖 𝐸෨𝑖

-2 ur
22 Co
෪1
𝑓1 𝑠1 , 𝑥1 = 𝑝1 𝐸

M RI
,B L
M of X
෪2
𝑓2 𝑠2 , 𝑥2 = 𝑝2 𝐸

SO rt
Pa
Such that:
𝑥2 ≤ 𝑠2
𝑥2 ≥ 0

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
31
Dynamic programming approach

𝑥𝑖
𝐸 𝑓𝑖 𝑠𝑖 , 𝑥𝑖 = 𝐸 𝑝𝑖 𝐸෨𝑖 𝑥𝑖 = 𝑝𝑖 𝐸 𝐸෨𝑖 = 𝑝𝑖 𝑥𝑖 1 − 𝐹𝑖 𝑥𝑖 ෩𝑖 𝑓𝑖 𝐷
+න 𝐷 ෩𝑖 𝑑 𝐷
෩𝑖

l
0

ia
er
at
𝑥1

M
𝐸 𝑓1 𝑠1 , 𝑥1 ෪1 = 𝑝1 𝑥1 1 − 𝐹1 𝑥1
= 𝐸 𝑝1 𝐸 ෩1 𝑓1 𝐷
+න 𝐷 ෩1 𝑑 𝐷
෩1

4 se
0

-2 ur
𝑥2

22 Co
𝐸 𝑓2 𝑠2 , 𝑥2 ෪2 = 𝑝2 𝑥2 1 − 𝐹2 𝑥2
= 𝐸 𝑝2 𝐸 ෩2 𝑓2 𝐷
+න 𝐷 ෩2 𝑑 𝐷
෩2

M RI
0

,B L
M of X
𝑣2 𝑠2 = max(𝑓2 𝑠2 , 𝑥2 + 𝑓1 (𝑠1 , 𝑥1 )))

SO rt
𝑥1 Pa 𝑥2
= max 𝑝1 𝑥1 1 − 𝐹1 𝑥1 ෩1 𝑓1 𝐷
+න 𝐷 ෩1 𝑑 𝐷
෩1 + 𝑝2 𝑥2 1 − 𝐹2 𝑥2 ෩2 𝑓2 𝐷
+න 𝐷 ෩2 𝑑 𝐷
෩2
𝑥2 0 0
Such that:
𝑥2 ≤ 𝑠2
𝑥2 ≥ 0

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
32
Dynamic programming approach

With first order conditions as:

𝜕ℒ
= 𝑝2 − 𝑝1 1 − 𝐹1 𝑥1 1 − 𝐹2 𝑥2 − ℓ1 + ℓ2 = 0

l
ia
𝜕𝑥2

er
ℓ1 𝑠2 − 𝑥2 = 0

at
M
ℓ2 (𝑥2 ) = 0

4 se
ℓ1 ≥ 0, ℓ2 ≥ 0, 𝑠2 − 𝑥2 ≥ 0, 𝑥2 ≥ 0

-2 ur
22 Co
Solution:

M RI
,B L
𝑠2 … 𝑖𝑓 𝑎𝑛𝑑 𝑝2 = 𝑝1

M of X
𝑝2 𝑝2
𝑠2 − 𝐹1−1 1− … 𝑖𝑓 𝑠2 ≥ 𝐹1−1 1− 𝑎𝑛𝑑 𝑝2 < 𝑝1
𝑝1
SO rt 𝑝1
Pa
𝑥2 =
𝑝2
0 … 𝑖𝑓 𝑠2 ≤ 𝐹1−1 1− 𝑎𝑛𝑑 𝑝2 < 𝑝1
𝑝1

This satisfies all the cases mentioned above

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
33
Heuristics Solution approach

Expected marginal seat revenue heuristics:

l
EMSR-a

ia
er
at
• This heuristics proposes to calculate full fare bookings (protection levels) for the current class using

M
4 se
Littlewood’s rule

-2 ur
22 Co
• That is for a three fare class problem

M RI
,B L
M of X
• Determine 𝑦31 with class 3 compared to class 1
𝑝
SO rt
• 𝑦31 = 𝐹1−1 1 − 3 Pa
𝑝1
• Determine 𝑦32 with class 3 compared to class 2
𝑝
• 𝑦32 = 𝐹2−1 1 − 3
𝑝2
• Total protection level is 𝑦3 = 𝑦31 + 𝑦32 , booking limit is 𝑏3 = 𝐶3 − 𝑦3

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
34
Capacity allocation problem
Problem# 4.3
Consider you are the revenue operations manager for Inbigo airlines. The airlines uses bigger airplanes in an all economy
configuration with 50 rows of 6 seats with 3 on each side of a single aisle. You are planning for the upcoming off-season ticket prices
on the Kolkata (CCU) to Mumbai (BOM) route. Adding discounts would induce passengers to leave Duronto tickets and shift to air

l
ia
er
travel. However, there is an available customer pool for business travel who would avail the facilities. You plan to have three-fare-

at
class problem

M
4 se
Currently, the full fare is priced at Rs. 8500 and you plan to offer a discount fare of Rs. 6500 for late discount and Rs. 4500 for early

-2 ur
discount. How many seats do you plan to keep for discount customers? Solve it using EMSR-a

22 Co
M RI
From various surveys and past data, it is known that number of passengers availing full fare follow a normal probability distribution

,B L
with mean 150 and standard deviation 10 and number of passengers availing late discount fare follow a normal probability

M of X
distribution with mean 50 and standard deviation 5 and number of passengers availing early discount fare follow a normal probability
distribution with mean 30 and standard deviation 5
SO rt
Pa
Follow the rule of multi-fare-class allocation problem

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
35
EMSR-a solution

l
ia
er
at
M
4 se
-2 ur
22 Co
Solution in excel 4_3

M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
36
 Heuristics Solution approach
Expected marginal seat revenue heuristics:

EMSR-b

l
ia
• This heuristics also proposes to calculate full fare bookings (protection levels) for the current class using

er
at
Littlewood’s rule

M
4 se
-2 ur
• That is for a three fare class problem

22 Co
M RI
• Determine 𝑦𝑗 for an artificial class against class j

,B L
M of X
𝑗−1
• Artificial mean demand for any 𝑗𝑡ℎ class 𝜇 = Σ𝑖=1 𝜇𝑖
𝑗−1 𝑝𝑖 𝜇𝑖
SO rt
• Artificial price for any 𝑗𝑡ℎ class 𝑝 = Σ𝑖=1
Pa
𝜇
𝑗−1
• Artificial Standard deviation for any 𝑗𝑡ℎ class 𝛿 = Σ𝑖=1 𝛿𝑖2
• Artificial class is constructed from all the fare classes above class j
−1 𝑝𝑗
• Protection level 𝑦𝑗 = 𝐹𝑎𝑟𝑡𝑖𝑓𝑖𝑐𝑖𝑎𝑙 1−
𝑝

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
37
Capacity allocation problem
Problem# 4.4
Consider you are the revenue operations manager for Inbigo airlines. The airlines uses bigger airplanes in an all economy
configuration with 50 rows of 6 seats with 3 on each side of a single aisle. You are planning for the upcoming off-season ticket prices
on the Kolkata (CCU) to Mumbai (BOM) route. Adding discounts would induce passengers to leave Duronto tickets and shift to air

l
ia
er
travel. However, there is an available customer pool for business travel who would avail the facilities. You plan to have three-fare-

at
class problem

M
4 se
Currently, the full fare is priced at Rs. 8500 and you plan to offer a discount fare of Rs. 6500 for late discount and Rs. 4500 for early

-2 ur
discount. How many seats do you plan to keep for discount customers? Solve it using EMSR-a

22 Co
M RI
From various surveys and past data, it is known that number of passengers availing full fare follow a normal probability distribution

,B L
with mean 150 and standard deviation 10 and number of passengers availing late discount fare follow a normal probability

M of X
distribution with mean 50 and standard deviation 5 and number of passengers availing early discount fare follow a normal probability
distribution with mean 30 and standard deviation 5
SO rt
Pa
Follow the rule of multi-fare-class allocation problem

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
38
EMSR-B solution

l
ia
er
at
M
4 se
-2 ur
22 Co
Solution in excel 4_4

M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
39
Demand dependence two-fare-class

Demand of higher fare class declines with increase in demand for lower

l
ia
er
fare class

at
M
4 se
Why would you use demand independent fare-differentiation?

-2 ur
22 Co
Analysis is simple and quick

M RI
,B L
In airline industry, there is a class of customer who would buy late

M of X
tickets and so fall in the high-fare class
SO rt
Pa
In offline business, discrimination is easy and practiced
In hospitality sector, business travelers form the high-fare class
This allows the demand independent analysis feasible

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
40
Demand dependence two-fare-class

𝑑𝑑 = Total discount demand

l
ia
er
𝑑𝑓 = Total full-fare demand assuming all discount bookings are

at
M
accepted

4 se
-2 ur
22 Co
M RI
Logic: if passengers in higher-fare class shift to discount-fare class, it

,B L
M of X
can be restated as passengers in discount-fare class that is willing to
SO rt
shift to higher-fare class Pa

෢𝑓 = 𝑑𝑓 + 𝛼 𝑑𝑑 − 𝑏
𝑑 +

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
41
Demand dependent Capacity allocation
𝐹𝑑 𝑏
𝑑𝑑 ≤ 𝑏 0

l
ia
er
1 − 𝐹𝑓 𝐶 − 𝑏

at
M
𝑑𝑓 > 𝐶 − 𝑏 𝑝𝑑 − 𝑝𝑓

4 se
-2 ur
1 − 𝐹𝑑 𝑏

22 Co
𝑏 →𝑏+1
𝑑𝑑 > 𝑏

M RI
,B L
M of X
𝐹𝑓 𝐶 − 𝑏 𝑝𝑑 − 𝛼𝑝𝑓
SO rt
Pa
𝑑𝑓 ≤ 𝐶 − 𝑏

Hold
0
𝑏 constant
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
42
Price differentiation problem solution

Expected change in returns due to increase in discounted seat:

𝐸 ℎ 𝑏 = 𝐹𝑑 𝑏 . 0 + 1 − 𝐹𝑑 𝑏 1 − 𝐹𝑓 𝐶 − 𝑏 𝑝𝑑 − 𝑝𝑓 + 𝐹𝑓 𝐶 − 𝑏 (𝑝𝑑 −𝛼𝑝𝑓 )

l
ia
er
= 1 − 𝐹𝑑 𝑏 {𝑝𝑑 − 𝛼𝑝𝑓 − 1 − 𝐹𝑓 𝐶 − 𝑏 (1 − 𝛼)𝑝𝑓 }

at
M
4 se
-2 ur
22 Co
M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
43
LittLewood’s modified application rule

The optimal point can also be found by this formula:

l
ia
er
at
1 − 𝐹𝑑 𝑏 𝑝𝑑 − 𝛼𝑝𝑓 − 1 − 𝐹𝑓 𝐶 − 𝑏 1 − 𝛼 𝑝𝑓 = 0

M
𝑝𝑑 − 𝛼𝑝𝑓

4 se
-2 ur
⇒ 1 − 𝐹𝑓 𝐶 − 𝑏 =

22 Co
1 − 𝛼 𝑝𝑓

M RI
𝑝𝑓 − 𝑝𝑑

,B L
M of X
∗ −1
⇒ 𝐶 − 𝑏 = 𝐹𝑓
1 − 𝛼 𝑝𝑓
SO rt
Pa
1 𝑝𝑑
⇒𝐶− = 𝑏∗ 𝐹𝑓−1 1−
1−𝛼 𝑝𝑓
Where, 𝑏 ∗ is the optimal number of discount seat
Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
44
LittLewood’s modified application rule

l
ia
Modified fare ratio is given by

er
at
M
4 se
-2 ur
22 Co
𝑝𝑓 − 𝑝𝑑 𝑝𝑑 − 𝛼𝑝𝑓 1 𝑝𝑑

M RI
1− = = −𝛼

,B L
1 − 𝛼 𝑝𝑓 1 − 𝛼 𝑝𝑓 1 − 𝛼 𝑝𝑓

M of X
SO rt
Pa
Where, 𝑏 ∗ is the optimal number of discount seat

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
45
Capacity allocation problem
Problem# 4.5

Consider you are the revenue operations manager for Inbigo airlines. The airlines uses A320 airplanes in an all economy configuration
with 30 rows of 6 seats with 3 on each side of a single aisle. You are planning for the upcoming off-season ticket prices on the Kolkata

l
ia
(CCU) to Mumbai (BOM) route. Adding discounts would induce passengers to leave Duronto tickets and shift to air travel. However,

er
there is an available customer pool for business travel who would avail the facilities.

at
M
Currently, the full fare is priced at Rs. 8500 and you plan to offer a discount fare of Rs. 5500. How many seats do you plan to keep for

4 se
discount customers? Solve it using Littlewood’s formula

-2 ur
22 Co
From various surveys and past data, it is known that number of passengers availing full fare follow a normal probability distribution

M RI
with mean 150 and standard deviation 10 and number of passengers availing discount fare follow a normal probability distribution

,B L
M of X
with mean 50 and standard deviation 5.

SO rt
Discount customers book their tickets earlier than full fare customers but there are some discount customers who when faced with
Pa
constrained discount fare capacity buy the full fare tickets which is found to be around 20%.

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
46
EMSR approach
Modified Expected marginal seat revenue heuristics:

EMSR-a

l
ia
er
• This modified heuristics proposes to calculate full fare bookings (protection levels) for the current class

at
M
using Littlewood’s rule with demand dependence

4 se
-2 ur
22 Co
• That is for a three fare class problem

M RI
,B L
• Determine 𝑦31 with class 3 compared to class 1

M of X
1 𝑝3
• 𝑦31 = 𝐹1−1 1−
SO rt
1−𝛼 𝑝1
Pa
• Determine 𝑦32 with class 3 compared to class 2
1 𝑝2
• 𝑦21 = 𝐹1−1 1−
1−𝛼 𝑝1
• Total protection level is 𝑦3 = 𝑦31 + 𝑦32 , booking limit is 𝑏3 = 𝐶3 − 𝑦3

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
47
 Solution approach
Expected marginal seat revenue heuristics:

EMSR-b

l
ia
• This heuristics also proposes to calculate full fare bookings (protection levels) for the current class using

er
Littlewood’s rule

at
M
4 se
• That is for a three fare class problem

-2 ur
22 Co
• Determine 𝑦𝑗 for an artificial class against class j

M RI
,B L
𝑗−1
• Artificial mean demand for any 𝑗𝑡ℎ class 𝜇 = Σ𝑖=1 𝜇𝑖

M of X
𝑗−1 𝑝𝑖 𝜇𝑖
• Artificial price for any 𝑗𝑡ℎ class 𝑝 = Σ𝑖=1
SO rt
Pa 𝜇
𝑗−1
• Artificial Standard deviation for any 𝑗𝑡ℎ class 𝛿 = Σ𝑖=1 𝛿𝑖2
• Artificial class is constructed from all the fare classes above class j
−1 1 𝑝𝑗
• Protection level 𝑦𝑗 = 𝐹𝑎𝑟𝑡𝑖𝑓𝑖𝑐𝑖𝑎𝑙 1−
1−𝛼 𝑝

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
48
Capacity options

l
ia
er
at
M
4 se
Capacity options problem: A service provider and an asset provider enter into a contract for providing

-2 ur
capacity. The other buyers and sellers exist in the spot market to provide additional capacity.

22 Co
M RI
,B L
Example: Air freight capacity booking through contract and spot market is an appropriate example

M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
49
Capacity options

l
ia
Service provider’s problem:

er
at
M
• The service provider buys the capacity through options contract or spot market.

4 se
• The service provider sells the capacity as a service along with other combined services at price 𝑝.

-2 ur
22 Co
• The price of the service sold by the service provider is 𝑝 = 𝑝(𝑟, 𝑥, 𝜆) where 𝜆 is the mark up by the service

M RI
provider.

,B L
• The service provider faces an uncertain demand 𝐷 ෪𝑐 .

M of X
• Sale of a capacity unit in the spot market involves an uncertain spot price of 𝑠ǁ and the cost incurred is 𝑡.
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
50
Capacity options

l
ia
er
at
M
Service provider’s variable:

4 se
-2 ur
The service provider decides on the capacity to reserve 𝑁.

22 Co
M RI
The service provider decides to call on the amount of reserved capacity 𝐸෨ (𝐸෨ ≤ 𝑁).

,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
51
Capacity options

Service provider’s optimization:

l
ia
Given:

er
Option prices → reservation price 𝑟 and execution price 𝑥.

at
M
Spot prices 𝑠ǁ is a random variable with distribution function 𝑔(𝑠)ǁ and cumulative distribution function 𝐺(𝑠).
ǁ

4 se
Also, 𝐺 𝑠ǁ = 0 𝑓𝑜𝑟 𝑠ǁ < 0.

-2 ur
Also, 𝑡 ≥ 𝑐 + 𝑣 for the option contract to be more lucrative for the seller than spot market.

22 Co
M RI
Service provider’s price is given by:

,B L
M of X
𝑝=𝑟+𝑥+𝜆

SO rt
Pa
෨
Service provider decides on the number of reserved capacity to call on as 𝐸.
Service provider decides on the reserved capacity 𝑁 based on the number of reserved capacity to call on that
෨
is 𝐸.

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
52
Capacity options

Service provider’s optimization:

l
ia
෪𝑐 .

er
Service provider faces uncertain demand 𝐷

at
෪𝑐 > 𝐸,
If 𝐷 ෨ then spot market purchase is 𝐷
෪𝑐 − 𝐸෨ for the service provider.

M
4 se
-2 ur
Service provider’s profit function is given by:

22 Co
max 𝑃෩𝑠 = max 𝑝𝐷෪𝑐 − 𝑟𝑁 − 𝑥𝐸෨ − 𝑠ǁ 𝐷 ෪𝑐 − 𝐸෨ +

M RI
𝐸෨ 𝐸෨

,B L
Such that:

M of X
𝑁 ≥ 𝐸෨
𝐸෨ ≤ 𝐷
SO rt
෪𝑐 Pa
𝐸෨ ≥ 0

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
53
Capacity options

With first order conditions as:

l
𝜕ℒ

ia
= −𝑥 + 𝑠ǁ − ℓ1 − ℓ2 + ℓ3 = 0

er
𝜕𝐸෨

at
M
ℓ1 𝑁 − 𝐸෨ = 0

4 se
ℓ2 𝐷෪𝑐 − 𝐸෨ = 0

-2 ur
ℓ3 𝐸෨ = 0

22 Co
ℓ1 ≥ 0, ℓ2 ≥ 0, ℓ3 ≥ 0, 𝑁 − 𝐸෨ ≥ 0, 𝐷
෪𝑐 − 𝐸෨ ≥ 0, 𝐸෨ ≥ 0

M RI
,B L
M of X
Solution:
SO rt
Pa
෪𝑐 , 𝑁
min 𝐷 … 𝑖𝑓 0 < 𝑥 ≤ 𝑠ǁ
෨
𝐸=൝
0 … 𝑖𝑓 𝑥 > 𝑠ǁ

This satisfies all the cases mentioned above

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
54
Capacity options

Service provider’s optimization:

l
ia
Given:

er
Option prices → reservation price 𝑟 and execution price 𝑥.

at
M
Spot prices 𝑠ǁ is a random variable with distribution function 𝑔(𝑠)ǁ and cumulative distribution function 𝐺(𝑠).
ǁ

4 se
Also, 𝐺 𝑠ǁ = 0 𝑓𝑜𝑟 𝑠ǁ < 0.

-2 ur
Also, 𝑡 ≥ 𝑐 + 𝑣 for the option contract to be more lucrative for the seller than spot market.

22 Co
M RI
Service provider’s price is given by:

,B L
M of X
𝑝=𝑟+𝑥+𝜆

SO rt
Pa
෨
Service provider decides on the number of reserved capacity to call on as 𝐸.
Service provider decides on the reserved capacity 𝑵 based on the number of reserved capacity to call on
෩.
that is 𝑬

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
55
Capacity options

l
ia
er
Service provider’s optimization:

at
M
෪𝑐 .

4 se
Service provider faces uncertain demand 𝐷

-2 ur
෪𝑐 > 𝐸,
If 𝐷 ෨ then spot market purchase is 𝐷
෪𝑐 − 𝐸෨ for the service provider.

22 Co
M RI
Service provider’s expected profit function is given by:

,B L
M of X
+
max Π = max ℰ 𝑃෩𝑠 = max ℰ 𝑝𝐷 ෪𝑐 − 𝑟𝑁 − 𝑥𝐸෨ − 𝑠ǁ 𝐷 ෪𝑐 − 𝐸෨
𝑁 𝑁 𝑁

SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
56
Capacity options

Service provider’s optimization:

l
ia
er
at
෪𝑐 .

M
Service provider faces uncertain demand 𝐷

4 se
෪𝑐 > 𝐸,
If 𝐷 ෨ then spot market purchase is 𝐷
෪𝑐 − 𝐸෨ for the service provider.

-2 ur
22 Co
Service provider’s expected profit function is given by:

M RI
+

,B L
max Π = max ℰ 𝑃෩𝑠 = max ℰ 𝑝𝐷 ෪𝑐 − 𝑟𝑁 − 𝑥𝐸෨ − 𝑠ǁ 𝐷 ෪𝑐 − 𝐸෨

M of X
𝑁 𝑁 𝑁
෪𝑐 − 𝑟𝑁 + ℰ 𝑠ǁ − 𝑥
= max 𝑝 − ℰ 𝑠ǁ ℰ 𝐷 + ෪𝑐 , 𝑁
ℰ min 𝐷
SO rt
𝑁
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
57
Capacity options

l
ia
er
at
Service provider’s optimization:

M
4 se
-2 ur
ℰ 𝑠ǁ − 𝑥 + − 𝑟

22 Co
𝑁 ∗ = 𝐹 −1
ℰ 𝑠ǁ − 𝑥 +

M RI
,B L
M of X
The optimum capacity to be reserved by the service provider at the asset provider is given by 𝑁 ∗

SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
58
Capacity allocation problem
Problem# 4.6

l
ia
er
at
Consider you are the revenue operations manager for UPDS freight airlines. UPDS freighter contracts freight plane space from AXI

M
carriers. The contract price mentioned by AXI is Rs. 1000 per unit of space and the execution price is Rs. 2000.

4 se
-2 ur
From various surveys and past data, it is known that the spot market price of airline cargo space follow a normal probability

22 Co
distribution with mean 5000 and standard deviation 2000 and the demand faced by UPDS follows a normal probability distribution

M RI
with mean 100 and standard deviation 10.

,B L
Please, prescribe the capacity to be reserved at AXI for maximizing profit.

M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
59
EMSR-B solution

l
ia
er
at
M
4 se
-2 ur
22 Co
Solution in excel 4_6

M RI
,B L
M of X
SO rt
Pa

Shounak Basak
2/24/2024
PODS Area, XLRI – Xavier School of Management
60
References
Material for the Lecture has been sourced from: Selected References:

Russell, R. S., & Taylor-Iii, B. W. (2008). Operations management along the supply [Link]
chain. John Wiley & Sons. [Link]
Stevenson, W. (2014). Operations management. McGraw-Hill Higher Education. carrier-by-2018-348222

l
ia
Chase, R. B., Jacobs, F. R., & Shankar, R. (2010). Operations and supply chain [Link]

er
management. Mcgraw Hill completes-four-years-465256

at
Mahadevan, B. (2015). Operations management: Theory and practice. Pearson [Link]

M
Education India railways-opening-date-facts-features/1162677/
Krajewski, L. J., Ritzman, L. P., & Malhotra, M. K. (2010). Operations management: [Link]

4 se
processes and supply chains. Upper Saddle River, New Jersey: Pearson diesel-loco-to-make-in-india-10000-hp-electric-locomotive/1086123/

-2 ur
Simchi-Levi, D., Kaminsky, P., Simchi-Levi, E., & Shankar, R. (2008). Designing and

22 Co
managing the supply chain: concepts, strategies and case studies. Tata McGraw-Hill
Education
Chopra, S., & Meindl, P. (2007). Supply chain management. Strategy, planning &

M RI
operation. In Das summa summarum des management (pp. 265-275). Gabler

,B L
Shah, J. (2009). Supply chain management: text and cases. Pearson Education India

M of X
SO rt
Pa

Shounak Basak
24-02-2024
PODS Area, XLRI – Xavier School of Management
61