Professional Documents
Culture Documents
Introduction
Recent empirical studies, both on service marketing and decision making, mostly focus on
developing quantitative tools for managing customers more effectively (Khajvand,
Zolfaghar, Ashoori, & Alizadeh, 2011; Pfeifer & Ovchinnikov, 2011; Yi Wen & Ku,
2010). This study is another attempt to accomplish that objective. The main aim of the
paper is to determine an optimal promotion strategy for the customers of a service
company that provides more than one specific type of product or service. Undoubtedly,
today corporate success depends on an organization’s ability to build and maintain relation-
ships with loyal and valued customers. Therefore it is necessary to build strategies for cus-
tomers based on their value (Kim, Jung, Suh, & Hwang, 2006). One tool for identifying
the value of customers is the measurement of customer lifetime value (CLV). CLV is
defined as the present value of all future profits obtained from a customer over the time of
that customer’s relationship with a firm (Bechwati & Eshghi, 2005; Gupta et al., 2006;
Pfeifer, Haskins, & Conroy, 2005). In other words, the value that a customer brings to the
firm is not limited to the profit obtained from each of his/her transactions, but is, rather,
the total profit s/he may provide over the duration of his/her relationship with the firm
(Kumar & George, 2007).
∗
Corresponding author. Email: ekinciyeliz@yahoo.com
Since the emergence of relationship marketing and the growing importance of custo-
mer relationship management (CRM), there has been an increase of interest in CLV (Dibb
& Meadows, 2001; Ryals & Knox, 2005). The literature on CLV consists of two groups of
studies. The first group focuses on different tools to measure CLV, while the second aims
to maximize it (Kumar & George, 2007; Kumar & Rajan, 2009). Although former research
in the literature has especially concentrated on the first group, recent research aiming to
find out methods to maximize CLV is rather limited. Thus, one contribution of this
study is to enhance the boundaries of the knowledge derived from the second group of
studies. On the other hand, CLV applications developed so far (see Table 1 given in the
second section) focus on offering mainly one type of product or service, while this
study provides the possibility of investigating the impact of promotion campaigns in a
company having a wide range of product portfolios. Furthermore, previous studies gener-
ally analysed the impact of a narrow range of promotion campaigns such as sending mail,
providing price discount or offering loyalty cards. However, this study considers the pro-
motion campaigns in a wider perspective, taking into account different types of campaigns
based on the frequency of their usage in practice.
One very important point about CLV analysis is that customers may differ sub-
stantially across industries, and such differences should be reflected in the models
used to evaluate them (Haenlein, Kaplan, & Beeser, 2007). Therefore, the calculation
of CLV has to be industry-specific rather than general (Gurau & Ranchhod, 2002; Jain
& Singh, 2002). On the other hand, not all customers are equally valuable. Therefore,
it may be desirable to allocate different resources to different groups of customers
(Gupta et al., 2006). The CLV measurement provides valuable information for estab-
lishing appropriate strategies for different customer segments, such as those who are
key customers, more costly, purchase on a regular basis, and those whom the firm
would be better off not serving (Gummesson, 2004; Ryals & Knox, 2005). Using the
CLV to guide marketing decisions not only encourages companies to recognize
differences between customers but also to create value through differential treatment
(Pfeifer & Ovchinnikov, 2011).
CLV measurement is possible in companies that maintain databases of customer/
end-user information on a substantial percentage of customers and that can customize
marketing ‘investments’, at least to some extent, across customers. Such companies
include hotels, airlines, credit-card companies, banks and financial service providers,
companies that sell over the Internet, telecommunications companies, cataloguers,
retail stores with ‘loyalty’ or ‘frequent shopper’ programmes, publishers, computer com-
panies that sell directly to consumers, and many more (Malthouse & Blattberg, 2005). In
fact nowadays, even smaller shops/restaurants are implementing various kinds of CRM
initiatives using the information about their customers. Companies that maintain data-
bases of customer/end-user information are able to keep extensive data on the personal
characteristics and purchasing behaviour of their customers. The advantage of this situ-
ation is that the company can derive information about consumer behaviour and plan its
marketing strategies accordingly (Yi Wen & Ku, 2010). By using customer information
contained in databases, companies can invest in customers who are (potentially) valu-
able to the company while trying to minimize their investment in non-valuable custo-
mers (Mulhern, 1999).
This study proposes a methodology to select customers according to the value that they
generate, and to specify which type of promotions should be allocated to which customers
in order to maximize the company’s overall value. The first section of this paper presents a
literature review on the CLV applications to marketing decisions in general, and to pro-
motion campaigns in particular. In the second section, the proposed model is explained.
The third section gives information about the application of the model and presents empiri-
cal results. The last section closes the paper with conclusions and suggestions for further
research.
profit gained through the customer or by extending the lifetime of the relationship
(Guillén, Nielsen, Scheike, and Pérez-Marı́n, 2012; Ryals & Knox, 2005). For this
reason, different relationship-marketing strategies are suggested for customers at different
stages of their lifetime.
Most studies about the use of CLV for marketing decisions have been focused on the
question of allocation of resources for marketing mix elements, together with its impact on
the CLVs of customers (Berger et al., 2002; Donkers & Verhoef, 2001; Kumar, Ramani, &
Bohling, 2004; Ryals & Knox, 2005).
related to different products, namely investment products, credits, and credit cards for the
banking industry.
Proposed model
The approach of finding an optimal promotion plan with the objective of maximizing CLV
is still rare, since previous studies have generally determined states via recency –
frequency – monetary variables, and only one type of service has been considered in the
application of these studies. A flowchart of the proposed procedure is shown in Figure 1.
Step 1. Classification and regression tree (CART) is a statistical procedure that was
introduced by Breiman, Friedman, Stone, and Olshen (1984). It is primarily used as a
classification tool, where the objective is to classify objects into two or more populations.
It is also a useful technique when the aim is to partition cases into homogeneous groups
according to a target value (in this study, customer value). CART is a non-parametric
decision-tree-learning technique that produces either classification or regression trees,
based on whether the dependent variable is categorical or numeric, respectively (Lee,
Chiu, Chou, & Li, 2006). The CART procedure partitions data into terminal nodes by a
sequence of binary splits, starting at a parent node. It starts by executing a partition of
an initial training data set into two subsets, so that the cases in each subset are more
homogeneous than they are in the single set of the whole group (Bevilacqua, Braglia, &
Montanari, 2003).
One noticeable advantage of decision tree-based models such as CART is that these
models are scalable to large problems (Razi & Athappilly, 2005). Thus, in this study,
we chose to define the states on the basis of the nodes resulting from CART, similarly to
the study by Haenlein et al. (2007). The states were the homogenous groups generated by
the tree. For the purpose of this study, several customer value drivers were used as predic-
tor variables, with the customer value being the target variable. The customer value was
defined as the revenue resulting from the transactions of customers minus the related
costs. The mean values of the states were then found, in order to be used in the other
steps of the procedure.
Step 2. The customers move between states according to a first-order Markov
process. A first-order Markov process is a stochastic process in which the probability
of a transition between two states depends only on the properties of the immediately
preceding state, independently of the path by which this state was reached. This con-
dition is referred to as the Markov property. Markov decision processes are widely
used in CLV studies (Morrison et al., 1982; Pfeifer & Carraway, 2000; Rust et al.,
2004). To estimate the transition probabilities, the states that the customers were in
at the beginning and at the end of each period were determined by using decision
rules obtained from the aforementioned CART analysis. Using this information and
also information about the promotional campaigns that were offered to customers in
each period, the transition probabilities between different states were found. Transition
probabilities are elements of Markov chains. A Markov chain model is characterized by
an N × N transition matrix P, where Pij (i, j ¼ 0, 1, 2,. . ., N 2 1) is the transition prob-
ability that a customer will move to state j in the next period given that currently the
customer is in state i (the retention probability of a customer in state i (i ¼ 0, 1, . . . ,
N 2 1) is given by Pii) (Ching et al., 2004). The transition probabilities in this step
were calculated from the formula:
Pij(k) ¼ (number of customers transiting from state i to state j under promotion cam-
paign k) / (number of customers in state i that are offered promotion campaign k).
Step 3. In this step, first the discount factors were selected, then the costs of the market-
ing promotion campaigns were derived, and finally the problem was modelled as a sto-
chastic dynamic programming model (i.e. a Markov decision process) that maximized
the present value for each state.
In this study, the problem was solved with the assumption of an infinite horizon. In a
Markov decision process, the optimal values vi satisfy the following (Ching et al., 2004;
Ross, 1983) for each state i and for j ¼ 1, . . . , M:
N −1
vi ≥ max c(j)
i +a p(j)
ik vk ,
j=1,...,M
k=0
where vi is the total expected value for a customer in state i, M is the total number of
marketing promotion campaigns ( j ¼ 1, . . ., M), dj is the cost of marketing promotion
campaign j in each period, pik( j) is the transition probability for the customer to move
from state i to state k with the jth marketing promotion campaign in each period, ci is
the mean customer value of state i, and a is the discount rate.
The solution to the problem gives the maximum expected present value for customers
in each state, together with the optimal promotion plan.
Step 4. The state values and the total value of the promotion campaigns were compared
with the status quo. The comparison was made according to the present values obtained for
each state, and appropriate promotion strategies were developed for each state.
The Service Industries Journal 111
Y. Ekinci et al.
Node Mean SD Number of customers in the cluster % Predicted mean Parent node Variable Improvement Split values
0 10.457 17.860 231,343 1 10.457
1 5.738 11.483 201,320 0.9 5.738 0 V1 149.327 ≤23.105
2 42.100 20.567 30,023 0.1 42.100 0 V1 149.327 .23.105
3 4.201 7.127 194,597 0.8 4.201 1 V2 59.517 ≤24.535
4 50.231 20.750 6723 0 50.231 1 V2 59.517 .24.535
5 33.340 12.795 21,708 0.1 33.340 2 V1 26.004 ≤50.595
6 64.972 19.403 8315 0 64.972 2 V1 26.004 .50.595
Table 3. CART table: final nodes.
Primary independent variable
Node Mean SD Number of customers in the cluster % Predicted mean Parent node Variable Improvement Split values
1 47.052 8.897 1659 0 47.052 9 V2 0.851 .37.505
2 62.891 9.377 1177 0 62.891 10 V2 0.907 ≤69.505
3 43.501 24.819 58 0 43.501 11 V2 0.368 .24.645
4 1.416 2.565 156,500 0.7 1.416 15 V3 4.350 ≤22.905
5 50.370 21.788 421 0 50.370 15 V3 4.350 .22.905
6 11.636 6.495 6700 0 11.636 16 V2 0.868 ≤13.365
113
29 91.221 5.894 1134 0 91.221 30 V1 0.420 .84.465
114 Y. Ekinci et al.
for state 3. The transition probabilities from state 3 to state 1 and to state 2 are 0.192 and
0.084, respectively, in the status quo case. However, these probabilities decrease with the
application of any type of promotion campaign.
Step 3.
(1) In order to examine the robustness of the results, the model was solved using 10
different discount factors (ranging from 0.90 to 0.99).
(2) The cost of promotion campaign 1 (status quo) was taken to be 0, while the costs of
the other three promotion campaigns were taken as $3, $2, and $1, respectively.
(3) Using Markov decision processes, the optimal marketing strategies were calcu-
lated with discount rates from 0.90 to 0.99 and an infinite horizon. The total
present values for these 10 discount rates are shown in Figure 2.
As can be seen in Figure 2, there is a breakpoint at 0.95, where the present value starts to
increase rapidly. Therefore, detailed empirical results are shown in Table 5 for discount rates
of 0.90, 0.95, and 0.99. This table includes the maximum values for each state with four
possible promotion campaigns, together with the optimal promotion campaign decision.
The expected values in Table 5 decrease when the discount factor decreases, but gen-
erally the optimal promotion campaign decisions do not change. Therefore the solution is
robust to different a values, but the present values obtained are changed. However, for
state 25 (the mean value of the state is $67.112, one of the most profitable states), when
the discount factor decreases to 0.90, the optimal promotion campaign decision changes
from promotion campaign 2 to promotion campaign 4, which is, in fact, a cheaper one.
Additionally, for state 29 (the most profitable state with mean value of $91.221), the
optimal decision is promotion campaign 2 for a discount factor of 0.99, whereas it is pro-
motion campaign 4 for smaller discount factors. Promotion campaign 2 includes offering
higher interest rates for investment products, so that customers will earn more by investing
in those products.
The values of the states increase from state 1 to state 29. The behaviour of customers in
each state can be evaluated based on this information and on the decision about the optimal
marketing promotion campaign. State 1 consists of those customers with the lowest cus-
tomer value, and it seems that whatever promotion campaign is directed towards them,
they do not move to more valuable states. Hence, the optimal marketing promotion
Table 5. Optimal promotion campaigns and present values for states under different discount rates.
Discount rates
a ¼ 0.99 a ¼ 0.95 a ¼ 0.90
Present Present Present
Optimal value of a Optimal value of a Optimal value of a
promotion customer in promotion customer in promotion customer in
State campaign state i ($) campaign state i ($) campaign state i ($)
1 1 2357.024 1 362.270 1 144.294
2 4 2462.178 4 438.455 4 201.131
3 3 2601.270 3 537.287 3 273.651
4 4 2671.699 4 592.588 4 318.112
5 3 2467.899 3 448.634 3 211.621
6 3 2494.538 3 460.540 3 217.218
7 2 2752.789 2 654.518 2 366.703
8 3 2578.123 3 522.350 3 264.707
9 3 2555.345 3 520.001 3 269.967
10 2 2699.852 2 585.008 2 298.798
11 4 2783.003 4 683.085 4 393.710
12 3 2703.026 3 619.847 3 342.832
13 3 2684.118 3 602.964 3 327.988
14 3 2644.896 3 594.571 3 332.647
15 3 2766.487 3 670.218 3 384.270
16 4 2742.538 4 669.381 4 392.565
17 3 2712.351 3 654.557 3 385.823
18 4 2487.679 4 471.436 4 238.909
19 2 2879.259 2 760.903 2 459.323
20 4 2854.691 4 742.577 4 445.241
21 3 2775.661 3 681.565 3 397.145
22 2 2588.484 2 539.886 2 287.092
23 3 2763.606 3 700.666 3 427.134
24 1 2601.532 1 567.458 1 318.558
25 2 2917.168 2 791.867 4 486.539
26 3 2864.446 3 760.865 3 466.658
27 3 3001.774 3 862.103 3 545.446
28 4 2809.301 4 742.232 4 465.521
29 2 2964.373 4 838.383 4 532.518
Total 78185.110 18076.210 10196.120
present
value of
the states
campaign is ‘status quo’. Another state of customers for which the optimal promotion
campaign is found to be campaign 1 is state 24. State 24 includes valuable customers,
and they seem to stay there even when no promotion campaign is offered to them. The
company can hold them in that state by doing nothing. Furthermore, the company may
decrease its profits by directing promotion campaigns to them, since promotion campaigns
are costly. The most costly promotion campaign is campaign 2, with a cost of $3. Custo-
mers in states 7, 10, 19, 22, 25, and 29 respond to this promotion campaign, as moving to
more valuable states increases the overall present value of the company. This finding indi-
cates that the promotion campaign for investment products (promotion campaign 2)
directly increases the value of the most valuable customers, who are in the higher
states, with a few exceptions.
The Service Industries Journal 117
One important point here is that promotion campaign 2 is not found to be attractive for
any of the customers in the first few states, because of its high cost. Another important
point is that the company does not need to direct any promotion campaign at the customers
in the 24th state in order to keep them as valuable customers of the organization. Undoubt-
edly, the major result of the analysis is that a promotion campaign for credit (promotion
campaign 3) is the optimal promotion campaign for a wide range of customers. Although
we expected to find that loans were generally used by customers who need more financial
support to make ends meet, the analysis in this study indicates that from the least valuable
to the most valuable, customers are influenced by promotion campaigns about credit. The
statistics of the Banks Association of Turkey also support this result, since the loans used
most often in 2009 in Turkey were personal loans (48%) and mortgage loans (46%) (Banks
Association of Turkey, 2009). Personal loans may be used mostly by customers in the less
valuable states, who are trying to meet their basic physiological and social needs and are
not interested in satisfying needs at the higher levels of the Maslow hierarchy. However, a
mortgage loan, which is a high-cost banking instrument, may be preferred by customers in
more valuable states. Promotion campaign 4, which is about credit cards, is also effective
for customers both in less valuable states such as 2, 4, and 11 and in higher states such as
16, 18, 20, and 28. Although this result has not been derived from the study, it might have
occurred because customers in lower states need to have more financial support for their
living expenses and use credit cards because they are short of cash, while customers in
higher states, who are expected to spend a lot and travel more, use credit cards because
of their convenience, validity around the world, availability, and some additional benefits
such as saving air-miles for flight tickets. As a result, credit cards are commonly held and
used by many customers.
Step 4. A comparison with the status quo was performed. Table 6 shows the
expected present values of the customers in each state for different discount factors
and the status quo case (promotion campaign 1). To demonstrate the findings, we
compare some values in Tables 5 and 6. The optimal marketing promotion campaign
for state 2 is found to be campaign 4 and, in this case, the expected value is found to
be $2462.178. However, in the status quo case, the corresponding value is $942.472,
meaning the former value is 2.6 times the latter. Similar findings are obtained for
other states.
The results in Table 6 show that when the problem is solved with an infinite horizon
using stochastic dynamic programming, the total value gained from the various states for
the optimal marketing promotion campaign decision is more than twice the value obtained
for the status quo, but the difference diminishes for smaller discount factors. The total
values in Table 5 for discount factors of 0.99, 0.95, and 0.90 are $78185.110,
$18076.210, and $10196.120, respectively. However, the corresponding values are
$30271.070, $8328.624, and $5391.514 for the status quo. This significant difference
shows us the importance of keeping previous data on customers, mining it, and using
optimization techniques for deciding on future marketing plans.
The results suggest a marketing promotion campaign strategy where campaign 1 is
offered to the customers who are in the lowest state and in one of the higher states
(namely states 1 and 24); campaign 2 is offered to states 7, 10, 19, 22, 25, and 29; cam-
paign 3 is offered to states 3, 5, 6, 8, 9, 12, 13, 14, 15, 17, 21, 23, 26, and 27; and,
finally, campaign 4 is offered to states 2, 4, 11, 16, 18, 20, and 28 if the discount factor
is expected to be 0.99. For a discount factor of 0.95, the optimal decision for state 29 is
changed to promotion campaign 4. Additionally, for a discount factor of 0.90, the
optimal decisions for states 25 and 29 are found to be campaign 4.
118 Y. Ekinci et al.
Table 6. Present values for states under different discount rates for the status quo.
Present value of a Present value of a Present value of a
customer in state i for a: customer in state i for a: customer in state i for a:
State i 0.99 ($) 0.95 ($) 0.90 ($)
1 915.942 169.975 78.901
2 942.472 193.202 99.016
3 970.741 217.924 120.416
4 990.866 235.915 136.339
5 952.775 203.431 109.171
6 962.562 211.790 116.211
7 1008.576 252.881 152.550
8 996.414 242.465 143.862
9 992.313 239.005 141.005
10 1030.088 273.108 171.466
11 1027.111 269.525 167.461
12 1025.178 268.933 168.055
13 1014.734 260.265 161.083
14 1041.021 283.186 180.870
15 1051.766 294.760 193.085
16 1080.652 320.050 214.977
17 1084.853 323.736 218.238
18 974.536 227.858 136.064
19 1106.049 344.269 237.962
20 1097.273 336.613 231.388
21 1074.774 316.715 214.075
22 1008.167 258.033 163.007
23 1129.461 365.360 256.934
24 1064.464 308.082 207.093
25 1128.526 365.809 258.546
26 1108.503 348.185 243.349
27 1150.054 386.474 278.338
28 1167.661 402.173 292.337
29 1173.539 408.902 299.714
Total present 30271.070 8328.624 5391.514
value of the
states
Conclusions
The study described in this paper aimed to determine the states of customers in terms of
their value by using the CART technique, which classifies customers according to their
CLV. Since previous studies have concentrated on optimal marketing strategies for com-
panies offering only one type of product or service; those studies used recency – fre-
quency – monetary values to define the states. However, this may not be efficient for
companies offering more than one type of product or service as in the case of the
finance company analysed in this study. While data on credit-card transactions includes
all of the recency – frequency – monetary values, frequency values do not exist for some
investment products such as forward accounts. Thus, this makes the use of recency –fre-
quency – monetary values impossible, since a customer may have or use many products or
services, such as investment products, and routine services such as electronic fund trans-
fers, cash withdrawal, credit cards, and loans in the case of banking services. In fact, the
reason for using recency – frequency – monetary values in the majority of previous banking
The Service Industries Journal 119
studies is that these studies were based solely on the segmentation of users of specific bank
service – mostly credit-card customers.
In addition, Tirenni et al. (2007) emphasize that regression-tree-based techniques out-
performed other clustering techniques in analysing the optimal promotion decisions
based on CLV. As a result, the present study used the CART technique which eliminates
the limitations of the previous studies and represents a more effective technique for the
specific service industry in order to determine the states of customers. This allows partition-
ing customers according to a target variable, similar to the study by Haenlein et al. (2007),
who suggest performing a segmentation for retail bank customers using the same technique.
The proposed methodology also includes a solution to the problem of discount factors.
The discount factor in this study is the yearly rate appropriate to investment in marketing.
Businesses need to consider the discount rate when deciding whether to spend some of
their profits on investment or whether to return profits to their shareholders. Therefore
companies can formulate their marketing action plans according to their forecast of the
discount factor, based on the conditions of the market, technology, competition, and
other environmental factors.
The methodology used in this study revealed important findings to guide managers in
deciding an optimal promotion campaign to be directed towards different customer seg-
ments in order to maximize profit.
The methodology shows that an optimal promotion campaign can be determined for
each of the different customer segments based on the CLV. The results underline that
among the three types of promotion campaigns, the promotion campaign for investments
(campaign 2) has a more positive impact on the lifetime value of customers in higher
states than on those in lower states. Thus the banks have to develop effective campaigns
based on investment products including different types of funds, bonds, stocks, repos,
and time deposit accounts for their customer segments with higher CLV in order to increase
their profitability. The other two promotion campaigns, namely campaigns for credit and for
credit cards, are effective in increasing the value of customers in a wide range of states. Thus
these findings indicate that the majority of segments of bank customers are more sensitive
and responsive to the credit-based promotion campaigns. However, compared with the
promotion campaign for credit cards, the analysis of the transition probabilities and of
the impact of discount factors shows that the implementation of credit-based promotion
has a more widespread and ongoing effect on the level of customer value, as well as on
moving customers from their current state to the next higher state in a continuous
manner. Thus, a promotion campaign for credit (promotion campaign 3) as the optimal pro-
motion campaign seems to be most attractive type of campaign for a wide range of custo-
mers to continue their relations with the bank by increasing their value for the institution.
However, promotion campaigns about credit consist of different types of promotions,
including those on personal loans, mortgage loans, automobile loans, etc. Each of these
different types of loans might be used by the customers in order to satisfy different needs.
For instance, a promotion campaign on personal loans might be more effective to increase
the value of customers at lower states compared with those on automobile loans. Thus,
determination of more specific promotion strategies requires the collection and analysis
of data on promotion campaigns on the basis of different types of credit. This allows us
to determine more specific optimal promotion campaigns for each segment of customers.
The promotion campaign for credit cards, on the other hand, has a leveraged impact on
the two ends of the continuum of states. The data on different types of credit-card-based
promotion campaigns have also been collected and analysed here to understand how and
when two extreme states are influenced by the nature of such promotion campaigns.
120 Y. Ekinci et al.
Importantly, all the empirical findings show that while some customers may be indif-
ferent to promotion campaigns, some care about the cost of promotion campaigns and
others really consider the content of the promotion campaigns offered to them. Further-
more, this behaviour is not really in parallel with the value generated by the customers;
in other words, the company cannot be sure that the valuable segments will respond to
the most costly promotion campaign or the less valuable segments will not respond to
them, etc. Therefore, companies should use their data warehouses as effectively as poss-
ible. The past data on customer behaviour collected by companies help them to understand
the various types of customer behaviour.
Additionally, the results show that the total value gained from the states when the
optimal promotion campaign is applied is more than twice the total value in the situation
of the status quo, when the problem is solved with an infinite horizon using stochastic
dynamic programming. As expected, the difference becomes smaller for smaller discount
factors. This result shows us the importance of keeping past data on customers, mining it,
and using optimization techniques to decide on future marketing plans. Undoubtedly, the
most important result derived from this study is that CLV-based segmentation is very
effective for determining optimal promotion strategies.
Acknowledgements
The authors would like to thank the anonymous financial services company that supplied
the data to perform this research study. This research was supported by Republic of
Turkey, Ministry of Industry and Trade (Project No: 00432.STZ.2009-2) and Istanbul
The Service Industries Journal 121
Technical University, Institute of Science and Technology (Project No: 34228). The
interpretation and conclusions revealed in this study do not represent the official
perspectives of the institutes stated above. The authors would also like to express their
deep gratitude to the anonymous referees for their invaluable comments.
References
Banks Association of Turkey. (2009). Consumer loans report. Ankara.
Bechwati, N.N., & Eshghi, A. (2005). Customer lifetime value analysis: Challenges and words of
caution. Marketing Management Journal, 15(2), 87 –97.
Berger, P.D., Bolton, R.N., Bowman, D., Briggs, E., Kumar, V., Parasuraman, A., & Creed, T.
(2002). Marketing actions and the value of customer assets: A framework for customer
asset management. Journal of Service Research, 5(1), 39 –54.
Bevilacqua, M., Braglia, M., & Montanari, R. (2003). The classification and regression tree approach
to pump failure rate analysis. Reliability Engineering and System Safety, 79, 59– 67.
Borle, S., Singh, S.S., & Jain, D.C. (2008). Customer lifetime value measurement. Management
Science, 54(1), 100 –112.
Breiman, L., Friedman, J., Stone, C.J., & Olshen, R.A. (1984). Classification and regression trees.
Pacific Grove, CA: Wadsworth.
Ching, W.-K., Ng, M.K., Wong, K.-K., & Altman, E. (2004). Customer lifetime value: Stochastic
optimization approach. Journal of the Operational Research Society, 55, 860–868.
Dibb, S., & Meadows, M. (2001). The application of a relationship marketing perspective in retail
banking. The Service Industries Journal, 21(1), 265– 280.
Donkers, B., & Verhoef, P.C. (2001). Predicting customer potential value: An application in the
insurance industry (working paper). Rotterdam: Erasmus University.
Guillén, M., Nielsen, J.P., Scheike, T.H., & Pérez-Marı́n, A.M. (2012). Time-varying effects in the
analysis of customer loyalty: A case study in insurance. Expert Systems with Applications, 39,
3551 –3558.
Gummesson, E. (2004). Return on relationships (ROR): The value of RM and CRM in business-to-
business contexts. Journal of Business and Industrial Marketing, 19(2), 136– 148.
Gupta, S., Hanssens, D., Hardie, B., Kahn, W., Kumar, V., Lin, N., . . ., Sriram, S. (2006). Modeling
customer lifetime value. Journal of Service Research, 9(2), 139–155.
Gurau, C., & Ranchhod, A. (2002). How to calculate the value of a customer. Measuring customer
satisfaction: A platform for calculating, predicting and increasing customer profitability.
Journal of Targeting, Measurement and Analysis for Marketing, 10(3), 203–220.
Haenlein, M., Kaplan, A.M., & Beeser, A.J. (2007). A model to determine customer lifetime value in
a retail banking context. European Management Journal, 25(3), 221–234.
Ho, T.H., Park, Y.H., & Zhou, Y.P. (2006). Incorporating satisfaction into customer value analysis:
Optimal investment in lifetime. Marketing Science, 25(3), 260– 273.
Jain, D., & Singh, S.S. (2002). Customer lifetime value research in marketing: A review and future
directions. Journal of Interactive Marketing, 16(2), 34–45.
Jonker, J.-J., Piersma, N., & Van den Poel, D. (2004). Joint optimization of customer segmentation
and marketing policy to maximize long-term profitability. Expert Systems with Applications,
27, 159 –168.
Khajvand, M., Zolfaghar, K., Ashoori, S., & Alizadeh, S. (2011). Estimating customer lifetime value
based on RFM analysis of customer purchase behavior: Case study. Procedia Computer
Science, 3, 57–63.
Kim, S.-Y., Jung, T.-S., Suh, E.-H., & Hwang, H.-S. (2006). Customer segmentation and strategy devel-
opment based on customer lifetime value: A case study. Expert Systems with Applications, 31,
101–107.
Kumar, V. (2006). CLV: The databased approach. Journal of Relationship Marketing, 5(2/3),
7 –35.
Kumar, V. (2010). A customer lifetime value-based approach to marketing in the multichannel, mul-
timedia retailing environment. Journal of Interactive Marketing, 24(2), 71– 85.
Kumar, V., & George, M. (2007). Measuring and maximizing customer equity: A critical analysis.
Journal of the Academy of Marketing Science, 35(2), 157–171.
Kumar, V., & Rajan, B. (2009). Profitable management: Measuring and maximizing customer life-
time value. Management Accounting Quarterly, 10(3), 1– 18.
122 Y. Ekinci et al.
Kumar, V., Ramani, G., & Bohling, T. (2004). Customer lifetime value approaches and best practice
applications. Journal of Interactive Marketing, 18(3), 60–72.
Kumar, V., & Reinartz, W.J. (2006). Customer relationship management: A databased approach
(1st ed.). New York, NY: John Wiley.
Kumar, V., Venkatesan, R., Bohling, T., & Beckmann, D. (2008). The power of CLV: Managing
customer lifetime value. Marketing Science, 27(4), 585– 599.
Labbi, A., & Berrospi, C. (2007). Optimizing marketing planning and budgeting using Markov
decision processes: An airline case study. IBM Journal of Research and Development,
51(3/4), 421 –431.
Lee, T.-S., Chiu, C.-C., Chou, Y.-C., & Li, C.-J. (2006). Mining the customer credit using classifi-
cation and regression tree and multivariate adaptive regression splines. Computational
Statistics & Data Analysis, 50, 1113–1130.
Malthouse, E.C., & Blattberg, R.C. (2005). Can we predict customer lifetime value? Journal of
Interactive Marketing, 19(1), 2 –16.
Morrison, D.G., Chen, R.D.H., Karpis, S.L., & Britney, K.E.A. (1982). Modeling retail customer be-
haviour at Merrill Lynch. Marketing Science, 1(2), 123–141.
Mulhern, F.J. (1999). Customer profitability analysis: Measurement, concentration, and research
directions. Journal of Interactive Marketing, 13(1), 25–40.
Pfeifer, P.E., & Carraway, R.L. (2000). Modeling customer relationships using Markov chains.
Journal of Interactive Marketing, 14(2), 43–55.
Pfeifer, P.E., Haskins, M.E., & Conroy, R.M. (2005). Customer lifetime value, customer profitability
and the treatment of acquisition spending. Journal of Managerial Issues, XVII(1), 11–25.
Pfeifer, P.E., & Ovchinnikov, A. (2011). A note on willingness to spend and customer lifetime value
for firms with limited capacity. Journal of Interactive Marketing, 25, 178–189.
Piersma, N., & Jonker, J.-J. (2004). Determining the optimal direct mailing frequency. European
Journal of Operational Research, 158, 173 –182.
Razi, M.A., & Athappilly, K. (2005). A comparative predictive analysis of neural networks (NNs),
nonlinear regression and classification and regression tree (CART) models. Expert Systems
with Applications, 29, 65 –74.
Ross, S. (1983). Introduction to stochastic dynamic programming. San Diego, CA: Academic Press.
Rust, R.T., Lemon, K.N., & Zeithaml, V.A. (2004). Return on marketing: Using customer equity to
focus marketing strategy. Journal of Marketing, 68(1), 109–127.
Ryals, L.J., & Knox, S. (2005). Measuring risk-adjusted customer lifetime value and its impact on
relationship marketing strategies and shareholder value. European Journal of Marketing,
39(5/6), 456 –472.
Tirenni, G., Labbi, A., Berrospi, C., Elisseeff, A., Bhose, T., Pauro, K., & Pöyhönen, S. (2007).
Customer equity and lifetime management (CELM) Finnair case study. Marketing Science,
26(4), 553 –577.
Venkatesan, R., Kumar, V., & Bohling, T. (2007). Optimal relationship management using Bayesian
decision theory: An application for customer selection. Journal of Marketing Research, XLIV,
579 –594.
Yi Wen, F., & Ku, E. (2010). Customer focus, service process fit and customer relationship manage-
ment profitability: The effect of knowledge sharing. The Service Industries Journal, 30(1/2),
203 –223.
Copyright of Service Industries Journal is the property of Routledge and its content may not
be copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.