264 Int. J. Revenue Management, Vol. 11, No.

4, 2020

Pricing strategies under customer recapture in airline

revenue management

Prashant Kumar Garg and

Sri Vanamalla Venkataraman*
Department of Industrial and Management Engineering,
Indian Institute of Technology Kanpur, India
Email: pgarg.iitk@gmail.com
Email: vanamala@iitk.ac.in
*Corresponding author

Abstract: Airline revenue management (RM) focuses on techniques such as

dynamic variation of price and seat inventory control. There is a lack of
sufficient literature on pricing strategies in competitive environments. In this
research we attempt to address the problem of optimal seat allocation and
pricing in a duopoly where each of the competing airlines has two fare-classes.
We study Nash equilibrium (NE) by extending the work of Mazumdar and
Ramachandran (2014) by including the phenomena of recapturing of
passengers who overflow to the competitor’s airline. We develop optimisation
models which incorporate this condition and develop an algorithm to obtain
their optimal solutions. These solutions determine the pay-offs in the
formulation of the underlying strategic form game between the competing
airlines. The NE of this game under various scenarios are finally computed and
analysed.
Keywords: revenue management; pricing; duopoly; customer recapture; Nash
equilibrium.
Reference to this paper should be made as follows: Garg, P.K. and
Venkataraman, S.V. (2020) ‘Pricing strategies under customer recapture in
airline revenue management’, Int. J. Revenue Management, Vol. 11, No. 4,
pp.264–276.
Biographical notes: Prashant Kumar Garg received his Master’s in
Technology from the Indian Institute of Technology Kanpur. His research
interests are applied operations research, game theory and machine learning. He
is currently a data scientist in American Express.
Sri Vanamalla Venkataraman is currently an Associate Professor in the
Department of Industrial and Management Engineering, Indian Institute of
Technology Kanpur. Her research interests include applied operations research,
optimisation and game theory.

1 Introduction

Airline Deregulation Act of 1978 can be considered as the starting point of revenue
management (RM). With this act, the US Civil Aviation Board lost its control over
setting the airline prices and at the same time, new low-cost and charter airlines entered
the market (Talluri and van Ryzin, 2006). The deregulation of airline’s prices and the

Copyright © 2020 Inderscience Enterprises Ltd.

 Pricing strategies under customer recapture in airline revenue management 265

introduction of discounted fare-classes led to the study and introduction of the new
concept, called yield management and later known as RM. RM is the art of managing
perishable assets and inventories to maximise revenue generated. It involves controlling
the availability and/or price of perishable assets and/or inventories with the aim of
maximising the total revenue. RM is applied in several industries such as the hotel and
tourism industry, for example, Roper et al. (2018) presented an integer programming
formulation that maximises parking revenue over a system of garages. Researchers have
used various approaches to classify the RM decisions. Talluri and van Ryzin (2006)
define the three basic categories of demand-management decisions to be structural
decisions, price decisions and quantity decisions. Raza and Akgunduz (2008), classify
RM literature into the four branches of demand forecasting, overbooking, seat inventory
control and pricing. British Overseas Airways Corporation (BOAC) was the first to start
the flight booking with discounted price. Littlewood (1972) of BOAC proposed that
discount fare booking should be accepted as long as their revenue exceeded the expected
revenue of future full fare. This rule is known as Littlewood’s rule of yield management.
Belobaba (1989) extended Littlewood’s rule to more fare classes through the introduction
of the concept of expected marginal seat revenue (EMSR). EMSR does not give the
optimal booking limits if there are more than two fare classes (McGill and van Ryzin,
1999). A modified version of EMSR, called EMSRb gives a better approximation for
booking limits if the fare classes are more than two. Initial work in this area concentrated
more about the monopolistic market with single fare class. Belobaba and Weatherford
(1996) discussed RM of perishable assets incorporating customer diversion. In their
paper, they presented a heuristic approach and showed that the largest gains (in revenues)
come from situations where the mean demand-to-capacity ratio was equal to 1.5 (i.e.,
situations where potential demand is large relative to capacity). Lippman and McCardle
(1997) studied the classical version of newsboy problem and examined the relation
between equilibrium inventory levels and splitting rule and provided the conditions for
which there is a unique equilibrium. Further, they proved that these conditions are
equally applicable to an airline industry by using randomised rationing rule for splitting
the seat among two airlines. Zhang and Cooper (2005) studied customer-choice
behaviour and addressed the problem of simultaneous seat inventory control of a set of
parallel flights between a common origin and destination with dynamic customer choice
among the flights. They had assumed that the fares for each flight during a certain period
are the same and modelled the problem as an extension of the classic multiperiod,
single-flight block demand RM model with the help of Markov decision process.
To the best of our knowledge Netessine and Shumsky (2005) were the first to address
the problem of seat inventory in the duopolistic market with more than one fare class.
They assumed a probability distribution for the demand of the two airlines and examined
this problem under both horizontal and vertical competition. They developed sufficiency
conditions to determine the pure-strategy Nash equilibrium (PSNE) for these RM games.
They also compared the total number of seats available in each fare class with and
without competition. Li et al. (2007, 2008) also studied the fare class allocation with two
airlines in competition. Li et al. (2008) established the existence of a PSNE by using
proportion rationing rule for splitting the common market demand among the airlines.
Gao et al. (2010) examined how the booking limit decisions are affected by competition
in a duopolistic market when the two airlines operate on the same route and focus on the
regret criterion. They assumed that the demand of airlines is independent and a customer
266 P.K. Garg and S.V. Venkataraman

books a ticket in only one fare class, but does not buy-up or buy-down to another fare-
class if their first choice is not available. They showed that there is a unique PSNE when
the number of fare classes is limited to two. Raza and Akgunduz (2008) studied the
duopoly market RM game to determine the equilibrium fare-pricing strategies and
booking limits. With the help of computational experiments they showed the existence of
PSNE for pricing under pre-committed booking limit; however, they did not prove the
existence of PSNE in the non-cooperative game.
Mazumdar and Ramachandran (2014) simultaneously addressed the problem of seat
inventory allocation and pricing in a duopoly where each of the competing airline
operates two fare-classes with same fare structure. Their model gives a static realisation
of seat allocation and pricing problem. Through numerical analysis, they computed the
PSNE that optimise the revenue generated by the two airlines. They also determined the
booking limits and price structure at equilibrium. The authors indicate that a direction of
future work is to include the phenomena of recapture of customers who were refused a
ticket by the competitor. In this study we extend the work of Mazumdar and
Ramachandran (2014) by including recapturing of customers who had overflown into the
competitor’s airline due to competitive prices. Our contribution through this research is in
modelling such aspects of customer recapture and in the development of an algorithm to
solve the underlying optimisation model. We study NE under these settings and compare
them with those of Mazumdar and Ramachandran (2014).

2 Model

In this section, we formulate the mathematical model which incorporates recapturing of

overflown passengers into the competitor’s airline. The model is then applied to
determine the optimal booking limits which are used to compute the optimal revenue
generated by the airlines. These optimal values are then taken as the pay-offs for the
underlying strategic form game. The game is then finally analysed for NE.

2.1 Problem setting

We consider a duopoly with the airlines offering two fare-classes. We assume that the
two airlines have the same capacity; both provide exactly the same service in all the
fare-classes, i.e., ticket of any fare-class gives the access to the same facilities and
services; the fare structure is same for both the airlines and they operate in the same
market. We assume that passengers do not move across two fare-classes. Our model
permits recapturing of passengers who overflow to the competitor but were denied
booking. We first develop an optimisation model under this setting which maximises the
airline’s revenue. This model is a refinement of the model developed by Mazumdar and
Ramachandran (2014) to accommodate recapture of overflown customers. We develop a
novel algorithm to solve the optimisation model. Applying the outcomes of this model we
setup the non-cooperative game [as developed by Mazumdar and Ramachandran (2014)]
induced between the competing airlines and compute the NE under various scenarios.
 Pricing strategies under customer recapture in airline revenue management 267

2.1.1 Solution methodology

The payoff of each airline for various pricing strategies is first determined by solving an
optimisation problem to maximise its revenue. However, the optimisation model of an
airline will also consist some of the decision variables of its competitor and vice-verse.
To address this problem we develop an algorithm to solve the optimisation problems of
the two competing airlines. The model analysis is conducted in a competitive framework
(by setting up the strategic form game) and the optimal pricing strategies of the
competitors are determined as their NE payoffs.

2.1.2 Assumptions
• Only two airlines A and B operate in the market, i.e., duopoly market.
• The capacity of the airlines A and B denoted by CA and CB respectively are the same.
i.e., C A = C B = C ( say )

• Each airline has two fare-classes, i.e., class 1 is the high fare-class and class 2 is the
low fare-class.
• The airlines provide exactly the same service in all the fare-classes.
• Every passenger is allowed to book a ticket in only one fare-class.
• Buy-up and buy-down are not allowed.
• The two airlines have the same fare structure and operate in the same market.
• The two airlines have the equal market share in a particular fare-class, i.e., the
demand in any particular fare-class is the same for the two airlines.
i.e., D1A = D1B and D2A = D2B

where Dkn denotes demand of airline n(= A, B) for class k = (1, 2).

• Overbooking is not allowed in any of the fare-class.

• Passengers do not move across the fare-classes.
• Customers overflow from one airline to another if the price difference exceeds a
threshold limit.
• Recapture of overflown customers who were denied booking in the competitor’s
airline is permitted.
We next describe the parameters and variables required for model development.

2.1.3 Parameters and variables

2.1.3.1 Parameters
CA, CB Total capacity of airlines A and B respectively.
Pkn Price of ticket of airline n(= A, B) for class k = (1, 2).
268 P.K. Garg and S.V. Venkataraman

Dkn Demand of airline n(= A, B) for class k = (1, 2).

RA, RB Revenue generated by airlines A and B respectively.

M kn Market share for class k of airline n.

Pkth Threshold price (threshold price refers to the price above which customer
moves to the other airline) for class k (= 1, 2).
c1, c2 Proportionality constants used to model customer overflow
a−Ao1 , a−Bo1 Number of overflown passengers (in class 1) from airline B to A and airline A
to B respectively.
a−Ao 2 , a−Bo 2 Number of overflown passengers (in class 2) from airline B to A and airline A
to B respectively.

2.1.3.2 Variables
a1A , a1B Demand to be accepted in class 1 for airlines A and B respectively.

a2A , a2B Demand to be accepted in class 2 for airlines A and B respectively.

aoA1 , aoB1 Number of overflown requests to be accepted in class 1 by airlines A and B

respectively.
aoA2 , aoB2 Number of overflown requests to be accepted in class 2 by airlines A and B
respectively.
xA, xB Booking limit of class 2 for airlines A and B respectively.
arA1 , arB1 Number of passengers recaptured in class 1 by airlines A and B respectively.

arA2 , arB2 Number of passengers recaptured in class 2 by airlines A and B respectively.

The description of the parameters and variables used in this paper are based on
Mazumdar and Ramachandran (2014) and Mazumdar (2017).

2.2 Optimisation model

In this section, we extend the model developed by Mazumdar and Ramachandran (2014)
by incorporating recapture of the overflown passengers in their model. The total number
of passengers available for recapture will be restricted by the difference between the
number of overflown passengers and the actual number of overflown passengers accepted
by the competitor, i.e.,
arkA ≤ a−Bok − aok
B
; k = 1, 2

and
arkB ≤ a−Aok − aokA ; k = 1, 2
 Pricing strategies under customer recapture in airline revenue management 269

Next we develop the optimisation models of airline A and B.

The decision variables of airline A are a1A , aoA1 , arA1 , a2A , aoA2 , arA2 and xA and that of
airline B are a1B , aoB1 , arB1 , a2B , aoB2 , arB2 and xB.
Airline A’s optimisation problem is described below.

Max R A = P1A ( a1A + aoA1 + arA1 ) + P2A ( a2A + aoA2 + arA2 ) (1)

Subject to:
a1A ≤ D1A − a−Bo1 (2)

a2A ≤ D2A − a−Bo 2 (3)

a1A + aoA1 + arA1 ≤ C − x A (4)

a2A + aoA2 + arA2 ≤ x A (5)

arA1 ≤ C1 ( a−Bo1 − aoB1 ) (6)

arA2 ≤ C1 ( a−Bo 2 − aoB2 ) (7)

where a1A , aoA1 , arA1 , a2A , aoA2 , arA2 and xA ≥ 0 are decision variables with integer values and

a−Ao1 = c1M 1A + c2 ∗ max ( ( P1B − P1th ) , 0 ) (8)

a−Ao 2 = c1 M 2A + c2 ∗ max ( ( P2B − P2th ) , 0 ) (9)

The customers overflown from the high-fare and low-fare classes are shown in
equations (8) and (9) respectively; the number of overflown passengers in a fare class is a
function of the airline’s market share and competitor’s price.
Airline B’s optimisation problem is as follows.

Max R B = P1B ( a1B + aoB1 + arB1 ) + P2B ( a2B + aoB2 + arB2 ) (10)

Subject to:
a1B ≤ D1B − a−Ao1 (11)

a2B ≤ D2B − a−Ao 2 (12)

a1B + aoB1 + arB1 ≤ C − x B (13)

a2B + aoB2 + arB2 ≤ x B (14)

arB1 ≤ C1 ( a−Ao1 − aoA1 ) (15)

arB2 ≤ C1 ( a−Ao 2 − aoA2 ) (16)

where a1B , aoB1 , arB1 , a2B , aoB2 , arB2 and xB ≥ 0 are decision variables with integer value and
270 P.K. Garg and S.V. Venkataraman

a−Bo1 = c1 M 1B + c2 ∗ max ( ( P1A − P1 ) , 0 ) (17)

a−Bo 2 = c1 M 2B + c2 ∗ max ( ( P2A − P2 ) , 0 ) (18)

Equation (1) represents the total revenue from the two fare classes. Equations (2) and (3)
represent the constraint that the number of seats accepted in a fare-class does not exceed
the difference between the demand of the airline and overflown customers. Equations (4)
and (5) denote capacity constraints for classes 1 and 2 respectively. Equations (6) and (7)
represent the restriction on the number of passengers available for recapture in classes 1
and 2 respectively. Equations (8) and (9) give the number of passengers overflown from
airline B to A in classes 1 and 2 respectively; the number of overflown passengers in a
fare class is a function of the airline’s market share and competitor’s price. Equations
(10)–(18) can be explained on the similar lines for airline B.
Here we observe that the decision variables aoB1 and aoB2 of airline B appear in the
optimisation problem of airline A. Similarly the decision variables aoA1 and aoA2 of airline
A appear in the optimisation problem of airline B.
To obtain an optimal solution we adopt an algorithmic approach. We next outline the
steps used in this approach (we illustrate the particular case when customer overflow
happens only in class 1 from one airline to another).
1 A trial value of aoB1 is initially used to solve the optimisation problem for airline A.
2 The optimum solution of step-1 is then used to solve the optimisation problem for
airline B.
3 If the new value of aoB1 obtained from step-2 is the same as the trial value used in
step-1 then the solution obtained is feasible and we stop, otherwise we go back to
step-1 in which the value of aoB1 obtained in step-2 is used to solve airline A’s
problem.

3 Numerical experiments

We next formulate the strategic form game to compute the NE for the two competing
airlines. Our formulation of the game is the same as that of Mazumdar and
Ramachandran (2014).
We denote 〈N, (Si), (ui)〉 as the strategic form game where:
• N ≡ {airline A, airline B} is the player set.
• Si ≡ {s1, s2, s3, s4}, i = A, B is the strategy set for players A and B respectively, with
the four strategies defined as follows:
s1 Both classes are priced at the base price (the base price, refers to the minimum
possible value of an airline ticket).
s2 Class 1 is priced at a predetermined percentage higher than the base price and
class 2 at the base price.
s3 Class 1 is priced at the base price and class 2 at a predetermined percentage
higher than the base price.
s4 Both classes are priced at a predetermined percentage higher than the base price.
 Pricing strategies under customer recapture in airline revenue management 271

• ui: S1 × S2 → , i = A, B, are the revenues generated by the airlines A and B which

are computed from the respective airlines’ optimisation model.
We next describe the numerical experiments performed to obtain the NE for the
underlying strategic game. The numerical values chosen for these experiments and the
experimental design setup are the same as that of Mazumdar and Ramachandran (2014).
We consider the base price for the two airlines to be 500 and 300 for classes 1 and 2
respectively. We consider the maximum capacity of the two airlines to be 150. Airlines’
demands are taken as 100 and 150 for high and low fare-class respectively. The
experimental design is given in Table 1.
Table 1 Experimental design

Factors Levels No. of levels

Price (% of base price) 5, 8, 10, 12, 15, 16, 18, 20, 25, 26, 30,40, 50 13
Threshold (% above P ) k
n
3, 5, 8 3
Total configurations 13 × 3 = 39
Instances 16
Total instances 39 × 16 = 624

3.1 Computation of NE
Tables 2–8 represent the revenue obtained from the above optimisation models with price
variation over base prices as 5, 8, 10, 15, 18, 20 and 26% respectively and a threshold of
3% for passenger overflow to happen. Equilibrium is computed by evaluating the best
response of each player assuming that the other player also plays his best response. The
best responses of airline A are marked with an asterisk (i.e., *) while the best responses of
airline B are marked with a double asterisk (i.e., **). In Table 2 (threshold 3%, price
variation 5%) and Table 3 (threshold 3%, price variation 8%) we observe that s3 is a
strictly dominant strategy for the two airlines and (s3, s3) is the only NE. In Table 4
(threshold 3%, price variation 10%) and Table 5 (threshold 3%, price variation 15%), we
observe that (s3, s3) is a NE for the two airlines though not a dominant strategy
equilibrium. Further it is observed that these NE are not Pareto-optimal. In Table 6
(threshold 3%, price variation 18%) and Table 7 (threshold 3%, price variation 20%) we
observe that (s3, s3) and (s4, s4) are the two NE for the two airlines with (s4, s4) being
Pareto-optimal. From Table 8 (threshold 3%, price variation 26%) we observe that (s4, s4)
is the only NE which is both Pareto-optimal and a strictly dominant strategy NE.
Table 2 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 5%

Airline B
s1 (500, 300) s2 (525, 300) s3 (500, 315) s4 (525, 315)
Airline A
s1 (500, 300) 65, 65 68, 64.13 65, 65.75** 68, 65.1
s2 (525, 300) 64.13, 68 67.5, 67.5 64.13, 68.53** 67.5, 68.25
s3 (500, 315) 65.75*, 65 68.53*, 64.13 65.75*, 65.75** 68.53*, 65.1
s4 (525, 315) 65.1, 68 68.25, 67.5 65.1, 68.53** 68.25, 68.25
272 P.K. Garg and S.V. Venkataraman

Table 3 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 8%

Airline B
s1 (500, 300) s2 (540, 300) s3 (500, 324) s4 (540, 324)
Airline A
s1 (500, 300) 65, 65 72.6, 59.88 65, 66.2** 72.6, 61.99
s2 (540, 300) 59.88, 72.6 69, 69 59.88, 72.89** 69, 70.2
s3 (500, 324) 66.2*, 65 72.89*, 59.88 66.2*, 66.2** 72.89*, 61.99
s4 (540, 324) 61.99, 72.6 70.2, 69 61.99, 72.89** 70.2, 70.2

Table 4 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 10%

Airline B
s1 (500, 300) s2 (550, 300) s3 (500, 330) s4 (550, 330)
Airline A
s1 (500, 300) 65, 65 75*, 57.5 65, 66.5** 75*, 60.5
s2 (550, 300) 57.5, 75** 70, 70 57.5, 75** 70, 71.5
s3 (500, 330) 66.5*, 65 75*, 57.5 66.5*, 66.5** 75*, 60.5
s4 (550, 330) 60.5, 75** 71.5, 70 60.5, 75** 71.5, 71.5

Table 5 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 15%

Airline B
s1 (500, 300) s2 (575, 300) s3 (500, 345) s4 (575, 345)
Airline A
s1 (500, 300) 65, 65 75*, 58.75 65, 67.25** 75*, 63.25
s2 (575, 300) 58.75, 75** 72.5, 72.5 58.75, 75** 72.5, 74.75
s3 (500, 345) 67.25*, 65 75*, 58.75 67.25*, 66.25** 75*, 63.25
s4 (575, 345) 63.25, 75** 74.75, 72.5 63.25, 75** 74.75, 74.75

Table 6 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 18%

Airline B
s1 (500, 300) s2 (590, 300) s3 (500, 354) s4 (590, 354)
Airline A
s1 (500, 300) 65, 65 75, 59 65, 67.4** 75, 63.8
s2 (590, 300) 59, 75 73, 73 59, 75 73, 75.4**
s3 (500, 354) 67.4*, 65 75, 59 67.4*, 67.4** 75, 63.8
s4 (590, 354) 63.8, 75 75.4*, 73 63.8, 75 75.4*, 75.4**

Table 7 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 20%

Airline B
s1 (500, 300) s2 (600, 300) s3 (500, 360) s4 (600, 360)
Airline A
s1 (500, 300) 65, 65 75, 60 65, 68** 75, 66
s2 (600, 300) 60, 75 75, 75 60, 75 75, 78**
s3 (500, 360) 68*, 65 75, 60 68*, 68** 75, 66
s4 (600, 360) 66, 75 78*, 75 66, 75 78*, 78**
 Pricing strategies under customer recapture in airline revenue management 273

Table 8 Payoff matrix of the two airlines (in 000’s) with a threshold of 3% and price variation
of 26%

Airline B
s1 (500, 300) s2 (630, 300) s3 (500, 378) s4 (630, 378)
Airline A
s1 (500, 300) 65, 65 75, 61.5 65, 68.9 75, 69.3**
s2 (630, 300) 61.5, 75 78, 78 61.5, 75 78, 81.9**
s3 (500, 378) 68.9, 65 75, 61.5 68.9, 68.9 75, 69.3**
s4 (630, 378) 69.3*, 75 81.9*, 78 69.3*, 75 81.9*, 81.9**

3.2 Results and discussion

Experimental results for different configurations are summarised in Table 9.
From Table 9 we observe that NE is achieved when both the players play the same
strategy. We also observe that NE shifts from (s3, s3) to (s4, s4) with percentage increase
in base price. While Mazumdar and Ramachandran (2014) established that NE is reached
only when prices are identical, we observe that incorporating customer recapture into
their model leads to improved (higher revenue) NE strategy profiles of (s3, s3) and (s4, s4).
Table 9 Summary of computational experiments

Price (% of base Threshold (% Strictly dominant Pareto

above Pkn ) NE
price) strategy (if any) optimal
5 3 (s3, s3) s3 --
8 3 (s3, s3) s3 --
10 3 (s3, s3) -- --
12 3 (s3, s3) -- --
15 3 (s3, s3) -- --
16 3 (s3, s3) and (s4, s4) -- 
18 3 (s3, s3) and (s4, s4) -- 
20 3 (s3, s3) and (s4, s4) -- 
25 3 (s3, s3) and (s4, s4) s4 (weakly dominant) 
26 3 (s4, s4) s4 
40 3 (s4, s4) s4 

Thus we conclude that for a substantial increase in price over the base price, the NE shifts
towards the strategy when both airlines price high, thereby providing both the airlines
with their maximum achievable revenue under the presence of competition. Thus we
infer that modelling customer recapture enables the airlines’ to adopt a higher pricing
strategy.

3.3 Model analysis

We next derive the conditions under which the NE shifts from one strategy profile to
another.
We introduce the following parameters and variables required for this discussion.
274 P.K. Garg and S.V. Venkataraman

P1 The equal value of base price in class 1 for the two airlines.
P2 The equal value of base price in class 2 for the two airlines.
C Maximum capacity of the two airlines.
D1 The equal demand value of the two airlines in class 1.
x Price variation over base price (in percentage).
y Threshold price over which overflow occurs (as an excess percentage of base price).
From Tables 2–8 in the last section, one can observe that the revenue generated by airline
i with respect to the strategy profile (si = s3, s–i = s2) becomes a constant (i.e., 75,000 in
our case) beyond 8% price variation over the base price.
In other words,
ui ( si = s3 , s−i = s2 ) = constant; i = A, B (for a variation above 8% over base price)

On the other hand revenue generated with respect to the strategy profile (si = s4, s–i = s2)
is increasing.
We also observe that whenever ui (si = s4, s–i = s2) ≥ ui (si = s3, s–i = s2) then strategy s3
fails to be a dominant strategy. (s3, s3) and (s4, s4) are observed to be the NE in such
situation. This phenomenon is expressed through the following relation.
 x   x 
1 ≤ P1 D1  1 +
PC  + P2 ( C − D1 ) 1 +  (19)
 100   100 
For parameters values of P1 = 500, P2 = 300, C = 150, D1 = 100 the value of x computed
from equation (19) is 15.38% and this has also been observed through the experiments.
We also observe that the recapture of passengers will not be happen until the
competitor rejects the overflown passengers. This happens when the remaining capacity
of the competitor in class 1 is less than the number of overflown passengers. This relation
is derived using equations (8) and (17) and is given below.
  x   y 
C − D1 ≤ c1M 1 + c2  P1  1 +  − P1 1 + 
  100   100 
On rearranging, we get the following expression.
 x− y
C − D1 ≤ c1M 1 + c2 P1   (20)
 100 
We also observe that beyond certain values of threshold and price variation over base
price, strategy s4 becomes the strictly dominant strategy and (s4, s4) becomes the strictly
dominant strategy NE. This happens when the following relation holds.
ui ( si = s4 , s−i = s j ) > ui ( si = sk , s−i = s j ) ∀j = 1,  , 4 and k = 1, 2, 3

This condition is equivalent to

 x   x   x 
P1 D1 + P2 ( C − D1 ) 1 +  > P1 ( 2 D1 − C ) 1 +  + P2 ( 2C − 2 D1 ) 1 + 
 100   100   100 
 Pricing strategies under customer recapture in airline revenue management 275

On rearranging we obtain,
 x 
P1 D1 > { P1 ( 2 D1 − C ) + P2 ( C − D1 )} 1 +  (21)
 100 
We observe that the parameters P1 = 500, P2 = 300, C = 150, D1 = 100 considered in our
experiments also satisfy the above relation. Solving for x we see that for any price
variation above 25% over the base price the NE shifts to (s4, s4) and strategy s4 becomes
the strictly dominant strategy.

4 Summary

In this work we consider duopoly market in which the airlines offer two fare-classes. We
formulate optimisation models which incorporate customer recapture. We develop a
novel algorithm to solve the optimisation models and these solutions were used as
pay-offs for the development of strategic form game. Experiments were conducted to
study the NE under various scenarios which established that incorporating customer
recapture leads to improved NE. Analysis of phenomena leading to various NE was also
performed. Scope for future work is to model buy-up and buy-down behaviour of
customers and study its effect on the pricing strategies.

References
Belobaba, P.P. (1989) ‘OR practice – application of a probabilistic decision model to airline seat
inventory control’, Operations Research, Vol. 37, No. 2, pp.183–197.
Belobaba, P.P. and Weatherford, L.R. (1996) ‘Comparing decision rules that incorporate customer
diversion in perishable asset revenue management situations’, Decision Sciences, Vol. 27,
No. 2, pp.343–363.
Gao, H., Ball, M.O. and Karaesmen, I.Z. (2010) ‘Competitive seat inventory control decisions
under the regret criterion’, Journal of Revenue and Pricing Management, Vol. 9, Nos. 1–2,
pp.49–65.
Li, M.Z.F., Oum, T.H. and Anderson, C.K. (2007) ‘An airline seat allocation game’, Journal of
Revenue and Pricing Management, Vol. 6, No. 4, pp.321–330.
Li, M.Z.F., Zhang, A. and Zhang, Y. (2008) ‘Airline seat allocation competition’, International
Transactions in Operational Research, Vol. 15, No. 4, pp.439–459.
Lippman, S.A. and McCardle, K.F. (1997) ‘The competitive newsboy’, Operations Research,
Vol. 45, No. 1, pp.54–65.
Littlewood, K. (1972) ‘Forecasting and control of passenger bookings’, AGIFORS 12th Annul
Symposium Proceedings, Nathanya, Israel, October, pp.193–204.
Mazumdar, C.S. (2017) Seat Allocation and Pricing in a Duopoly in the Airline Industry, Doctoral
dissertation.
Mazumdar, C.S. and Ramachandran, P. (2014) ‘Seat allocation and pricing in a duopoly in the
airline industry’, International Journal of Revenue Management, Vol. 8, No. 1, pp.20–33.
McGill, J.I. and van Ryzin, G.J. (1999) ‘Revenue management: research overview and prospects’,
Transportation Science, Vol. 33, No. 2, pp.233–256.
Netessine, S. and Shumsky, R.A. (2005) ‘Revenue management games: horizontal and vertical
competition’, Management Science, Vol. 51, No. 5, pp.813–831.
276 P.K. Garg and S.V. Venkataraman

Raza, A.S. and Akgunduz, A. (2008) ‘An airline revenue management pricing game with seat
allocation’, International Journal of Revenue Management, Vol. 2, No. 1, pp.42–62.
Roper, M.A., Triantis, K., Taylor, G.D. and Teodorović, D. (2018) ‘Revenue management in the
parking industry: a multiple garage intelligent reservation model’, Transportation Planning
and Technology, Vol. 41, No. 3, pp.286–300.
Talluri, K.T. and van Ryzin, G.J. (2006) The Theory and Practice of Revenue Management,
Vol. 68, Springer Science & Business Media.
Zhang, D. and Cooper, W.L. (2005) ‘Revenue management for parallel flights with
customer-choice behavior’, Operations Research, Vol. 53, No. 3, pp.415–431.