Industrial Management & Data Systems

Universal structure modeling approach to customer satisfaction index

Ali Turkyilmaz Asil Oztekin Selim Zaim Omer Fahrettin Demirel
Article information:
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

To cite this document:

Ali Turkyilmaz Asil Oztekin Selim Zaim Omer Fahrettin Demirel, (2013),"Universal structure modeling
approach to customer satisfaction index", Industrial Management & Data Systems, Vol. 113 Iss 7 pp. 932 -
949
Permanent link to this document:
http://dx.doi.org/10.1108/IMDS-12-2012-0444
Downloaded on: 12 March 2016, At: 00:40 (PT)
References: this document contains references to 41 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 1147 times since 2013*
Users who downloaded this article also downloaded:
Mohammad Suleiman Awwad, (2012),"An application of the American Customer Satisfaction Index
(ACSI) in the Jordanian mobile phone sector", The TQM Journal, Vol. 24 Iss 6 pp. 529-541 http://
dx.doi.org/10.1108/17542731211270098
Ali Trky#lmaz, Co#kun zkan, (2007),"Development of a customer satisfaction index model: An
application to the Turkish mobile phone sector", Industrial Management & Data Systems, Vol. 107 Iss
5 pp. 672-687 http://dx.doi.org/10.1108/02635570710750426
Manuel Jos Vilares, Pedro Simes Coelho, (2003),"The employee-customer satisfaction chain
in the ECSI model", European Journal of Marketing, Vol. 37 Iss 11/12 pp. 1703-1722 http://
dx.doi.org/10.1108/03090560310495429

Access to this document was granted through an Emerald subscription provided by emerald-srm:543736 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.

*Related content and download information correct at time of download.


IMDS
113,7
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)
932
Received 24 December 2012
Revised 8 March 2013
10 March 2013
Accepted 11 March 2013
Industrial Management & Data
Systems
Vol. 113 No. 7, 2013
pp. 932-949
q Emerald Group Publishing Limited
0263-5577
DOI 10.1108/IMDS-12-2012-0444
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0263-5577.htm
Universal structure modeling
approach to customer
satisfaction index
Ali Turkyilmaz
Department of Industrial Engineering, Fatih University, Istanbul, Turkey
Asil Oztekin
Manning School of Business, University of Massachusetts Lowell, Lowell,
Massachusetts, USA
Selim Zaim
Department of Mechanical Engineering, Marmara University,
Istanbul, Turkey, and
Omer Fahrettin Demirel
Department of Industrial Engineering, Fatih University, Istanbul, Turkey
Abstract
Purpose Previous researches have proven that customer satisfaction and loyalty are affected by
complicated relationships and are challenging to European customer satisfaction index (ECSI) model.
Existing approaches mostly limit their hypotheses to linear relationships, which hinder much information
that would lead to better modeling and understanding the relationship between customer satisfaction and
loyalty. The purpose of this paper is to reveal potential nonlinear and interaction effects that might be
embedded in antecedents of ECSI by exemplifying it in Turkish telecommunications sector.
Design/methodology/approach This papar has justified the validity and reliability of the ECSI
model implementation in Turk Telekom Company. The path models are tested via conventional
structural equation modeling (SEM) and using a novel method, i.e. universal structure modeling with
Bayesian neural networks.
Findings The findings of this study reveal that quality has the most important impact on customer
satisfaction. The next important construct was found to be the company image. The relationship
between customer expectation and customer satisfaction was revealed to be insignificant. This study
reveals the fact that while using the ECSI model more attention must be paid to the consideration of
potential nonlinear relationships that might be available among model constructs.
Originality/value This research presents uniqueness in that it reveals significant nonlinear
relationships between the model constructs of the ECSI model. Previous studies have identified purely
linear relationships, which may not hold true in reality. However, in this study it is revealed that
improving one determinant of customer satisfaction may not be as worthy as it is assumed to be in
theory, which refers to a nonlinear relationship.
Keywords Customer satisfaction index, Telecommunication, SEM, Universal structure modeling,
Bayesian neural networks, Customer satisfaction
Paper type Research paper
1. Introduction
Recently, the concept of customer satisfaction has attracted much attention. It
has become a central concern for companies and organizations in order to improve
product and service quality, and to maintain customer loyalty (CL) within a highly
 competitive marketplace. A key motivation for the growing emphasis on customer USM approach
satisfaction is that high customer satisfaction leads to a stronger competitive position to CSI
resulting in higher market share and profit (Fornell, 1992). Customer satisfaction is also
generally assumed to be a significant determinant of repeat sales, positive
word-of-mouth, and CL. Satisfied customers return and buy more, and they tell other
people about their experiences (Fornell et al., 1996; Turkyilmaz and Ozkan, 2007). It is
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

also a widely accepted concept that customer satisfaction improvement attempts via 933
total quality management (TQM) has been amongst the most critical factors that derive
an operations strategy of an organization in order to improve its operations performance
(Kannan and Tan, 2005). According to Chien et al. (2002), the success of customer
satisfaction is not only closely related to a firms TQM performance but it also has
permanent effects on the firms future.
Until recently, in most countries, telecommunications service providers were
state-owned, state-operated, and often monopolistic. The monopoly-based system of
service provision, which dominated the worlds telecommunications markets for over
three-quarters of the last century, gave way to more competitive supply in many
markets (Li and Xu, 2002).
The Turkish electronic communications market has developed rapidly since its
liberalization on January 1, 2004. The number of players in the market tended to
increase continuously after the liberalization. By the beginning of 2013 there exist
369 operators in the electronic communications market (ICTA, 2012). Turk Telekom
group (the leader of the telecommunication sector in Turkey) provides integrated
telecommunication services from PSTN (fixed line telephone service), GSM (Global
System for Mobile Communications) to broadband internet. Turk Telekom group
companies have 15.8 millions of PSTN customers, 6.7 millions of broadband internet
customers and 11.8 million GSM customers as of March 31, 2011 (Turk Telekom, 2012).
Fixed (incumbent and long distance telephone service providers) and mobile
operators total sales revenues constitute the most important part in total electronic
communications sector revenues. From the beginning of the liberalization process in
2004-2007, total revenues continuously increased. However, in 2008 sales revenues
decreased by 5 percent to 19.3 billion Turkish Liras (TL) which was 20.3 billion TL in
2007. After a continuous increase, in 2008, the share of mobile operators in total revenues
became 70 percent and fixed operators share decreased to 30 percent (ICTA, 2012).
The purpose of this study is to provide information about the customer satisfaction
indices (CSIs) and show the results of a CSI study carried out in Turkish
telecommunications sector. The main focus of this research is to reveal non-trivial,
implicit, previously unknown, and potentially useful nonlinear relationships (if any)
via the deployment of a novel method, i.e. the universal structure modeling (USM). This
integrated method along with the conceptual European customer satisfaction index
(ECSI) model would hypothetically provide a better in-depth understanding of the
potential complex relationships of the determinants of CL and customer satisfaction.
The rest of this paper is structured as follows: Section 2 briefly presents an overview of
the CSIs and their developments in different countries. Section 3 presents an ECSI
model implemented in Turkish telecommunication sector. The methodology and data
analysis are discussed in Section 4, followed by the analysis of the main results.
Discussions and concluding remarks are provided in Sections 5 and 6.
 IMDS 2. Background
113,7 The original interest in customer satisfaction research has focused on customers
experience with a product episode or service encounter (Anderson et al., 1994). More
recent studies have focused on cumulative satisfaction. Cumulative satisfaction defines
satisfaction as customers overall experience to date with a product or service provider.
This approach to satisfaction provides a more direct and comprehensive measure of a
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

934 customers consumption utility, subsequent behaviors and economic performance

(Fornell et al., 1996). CSIs have been built upon a cumulative view of satisfaction.
The CSI model is a structural model-based on the assumptions that customer
satisfaction is caused by some factors such as perceived quality (PQ), perceived value
(PV), expectations of customers, and image of a firm. These factors are the antecedents of
overall customer satisfaction. The model also estimates the results when a customer is
satisfied or not. These results of customer satisfaction are consequence factors such as
complaints or loyalty of customer ( Johnson et al., 2001). Each factor in the CSI model is a
latent construct which is operationalized by multiple indicators (Fornell, 1992;
Bayraktar et al., 2012).
Since the importance of customer satisfaction is widely recognized, each country has
provided its own national customer satisfaction index (national CSI) to analyze the level
of customers satisfaction through the companies efforts (Chien et al., 2003). Swedish
customer satisfaction barometer (SCSB), reported in 1989, was the first national CSI
(Fornell, 1992). The American customer satisfaction index (ACSI) was developed in 1993
by Claes Fornell, the founder of SCSB. The ACSI survey is conducted for seven main
economic sectors, 35 industries, and more than 200 companies with revenues totaling
nearly 40 percent of the US GNP (Fornell et al., 1996). The ECSI, developed by European
organization for quality and European foundation for quality management, was first
introduced in 1999 across 11 European countries (Eklof and Westlund, 2002). Other
efforts for establishing national CSIs can be seen in Denmark, Austria, France, The
Netherlands, Switzerland, Taiwan, New Zealand, South Korea, Malaysia, Hong Kong
and Russia (Grigoroudis and Siskos, 2003).
The original SCSB model contains two primary antecedents of satisfaction:
perceived performance and customer expectations (Fornell, 1992; Anderson et al.,
1994). The consequences of satisfaction in the SCSB model are derived from
Hirschmans (1970) exit-voice theory which describes the results of dissatisfaction.
The ACSI model builds upon the original SCSB model specifications adapted in the
distinct characteristics of the US economy. The main differences between the original
SCSB model and the ACSI model are the addition of a PQ component, as distinct from
PV, and the addition of measures for customer expectations (Anderson et al., 1994;
Fornell et al., 1996).
The ECSI, a modified adaptation of the ACSI model, considers the European
economy as a whole, and thus CSI scores of the countries can be compared with each
other and with the European average (Eklof and Westlund, 2002). In the ECSI model,
customer expectations, PQ, PV, customer satisfaction, and CL constructs are modeled
as in the ACSI. There are two fundamental differences between the ACSI and ECSI
models. First, the ECSI model does not include the complaint behavior construct as a
consequence of satisfaction. Second, the ECSI model incorporates company image as a
latent variable in the model (Grigoroudis and Siskos, 2003).
 The pioneer studies on the CSI model in Turkey are the research studies conducted USM approach
by Turkyilmaz and Ozkan (2007) and Aydin and Ozer (2005). The first national model, to CSI
Turkish customer satisfaction index (TCSI), and adaptation of ACSI model to Turkish
market, was reported as a pilot study in the fourth quarter of 2005 by Turkish Quality
Association (Kal-Der) and KA Research Limited (TMME, 2006).
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

3. The ECSI model 935

The model studied in this paper has been derived specifically for the ECSI which is an
economic indicator that measures customer satisfaction based on well-established
theories and approaches in customer behavior and is applicable for a number of
different industries (Turkyilmaz and Ozkan, 2007; Grigoroudis and Siskos, 2003). As
shown in Figure 1, the CSI model variables are structured in the following categories:
causes of satisfaction (e.g. image, expectations, PQ, PV), satisfaction and results of
satisfaction (e.g. loyalty). Given these relations between the latent variables, the CSI
model presents a system of cause and effect relationships. Then, a set of manifest
variables is associated with each latent variable. The use of multi-item (instead of
single-item) scales in the CSI model provides more reliable measurement of overall
satisfaction (Johnson and Fornell, 1991).

Customer satisfaction
CSI construct, located at the center, is the main interest of CSI study. It generates
index scores and provides comparison between companies and industries over time
periods. The index indicates how much customers are satisfied with the products
or services of a company, and how well their expectations are met. Similar to
other constructs, CSI is measured using multiple indicators suggested by the literature.
These indicators are: overall satisfaction level of customers (cumulative nature
of consumer satisfaction), fulfillment of customer expectations (i.e. the degree to
which performance exceeds or falls short of expectations (Oliver, 1980), and
comparison with an ideal provider (i.e. company performance versus the
consumers hypothetical ideal provider) (Johnson and Fornell, 1991; Fornell et al.,
1996; Chan et al., 2003).

Image

Expectaion Loyalty

Perceived
Satisfaction
Value

Perceived Figure 1.
Quality ECSI model
 IMDS Image
113,7 The image construct evaluates the underlying image of the company. Image refers to
the brand name and the kind of associations customers get from the product/company
(Andreassen and Lindestad, 1998). Martensen et al. (2000) indicate that image is an
important component of the customer satisfaction model. For the companies, image is a
result of being reliable, professional, and innovative, having contributions to society,
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

936 and adding prestige to its user. It is expected that image has a positive effect on
customer expectations, customer satisfaction, and loyalty (Martensen et al., 2000).

Expectations
The customer expectations represent both the prior consumption experience with the
firms offering including non-experiential information available through sources such
as advertising, promotion, pricing and word-of-mouth, and a forecast of the suppliers
ability to deliver quality in the future. Evaluation of expectations is important because
the nature of the ongoing relationship between a firm and its customer base is such
that expected future quality is critical to overall customer satisfaction. Evaluations of
expectations represents either positive disconfirmation (when performance is better
than expected), negative disconfirmation (when performance is worse than expected),
or confirmation that performance is as good as expected. Positive disconfirmation and
confirmation result in satisfaction outcomes, while negative disconfirmation results in
dissatisfaction outcomes (Fornell et al., 1996; Johnson et al., 2001).
This construct measures customer expectations for overall quality, for product, and
service quality, and for fulfillment of personal needs. The customer expectations
construct is expected to have a direct and positive relationship with customer
satisfaction (Anderson and Fornell, 2000).

Perceived quality
PQ is the served markets evaluation of recent consumption experience. In the literature,
this construct evaluates customization and reliability of a given product or service.
Customization is the degree to which a product or service meets customers requirements
and reliability is the degree to which firms offering is reliable, standardized, and free
from deficiencies. In this study the customization and reliability of the following
measures are evaluated: overall quality, products technical quality, service quality,
provided customer services quality related to product and appropriateness to intent of
use. PQ is not generally price or value related. PQ is expected to have a positive effect on
PV and customer satisfaction and to be positively affected by image (Andreassen and
Lindestad, 1998; Fornell et al., 1996).

Perceived value
PV is the perceived level of product quality relative to the price paid by customers.
Although PQ is not value or price related, it has been shown that consumer satisfaction
depends on the value which in turn depends on price. Thus, price information is
incorporated into the customer satisfaction measurement model to increase the
comparability of the index scores across products or services with different prices
(Fornell et al., 1996; Chan et al., 2003).
PV is the rating of the price paid for the quality perceived and a rating of the quality
perceived for the price paid. The PV structure provides an opportunity for comparison
 of the firms according their price/value ratio. In the ECSI model, PV is expected to be USM approach
positively affected by PQ, and it has a positive impact on satisfaction (Turkyilmaz and to CSI
Ozkan, 2007).

Customer loyalty
CL has been defined as a long-term commitment to repurchase involving both repeated
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

patronage and a favorable attitude. CL is the ultimate factor in the ECSI model. It is 937
measured by repurchase intention, price tolerance and intention to recommend
products or services to others. It is expected that better image and higher customer
satisfaction should increase CL (Anderson and Fornell, 2000).

4. Research methodology
4.1 Survey setting
A case study method was used to collect the required data on the underlying
dimensions of the research model. The main rationale behind selecting Turk Telekom
in the telecommunication industry can be attributed to the changing focus of the
company on customer satisfaction and loyalty after the privatization. Data was
gathered from a face-to-face survey from 280 Turk Telekom customers who live in
Istanbul, the largest metropolitan area in Turkey. The eligible respondents were those
who qualified by having recent experience as the purchaser and user of fixed line
PSTN. Among all collected data set, 266 were found appropriate for further data
analysis.

4.2 Survey instrument

The ECSI model consists of the aforementioned constructs in Section 3 which is based
on well-established theories and approaches in customer behavior. The constructs of
the ECSI model are unobservable (latent) variables indirectly described by a block
of observable variables which are called manifest variables or indicators.
The constructs and their observable items are presented in Table I. The use of
multiple questions for each construct increases the precision of the estimate as opposed
to an approach using a single question for each. A survey, developed to measure the
manifest variables, was prepared in Turkish, and the first draft was issued to 30
individuals to ensure that the wording, format, and sequencing of questionnaire were
appropriate. Feedback from this pilot study indicated that some questions were
ambiguous, difficult to understand, or irrelevant for the telecommunications sector.
The final questionnaire contained 23 questions pertaining to the CSI. Besides the model
questions, some demographic questions (e.g. age, gender, education level, etc.) were
also included in the questionnaire. A five-point measurement scale was used where a
value of 1 expresses a very negative point of view and a value of 5 expresses a very
positive one.

5. Data analysis and results

The analysis used in this study was undertaken in two stages. In the first stage, the
overall reliability and validity of ECSI Model in Turk Telekom Company was
measured. In the second stage, the relationships among constructs were assessed using
the USM. These steps are discussed in detail in the following subsections.
 IMDS
Latent variables Observable (manifest) variables
113,7
Image (IM) IM1: being reliable
IM2: being professional
IM3: social contributions to society
IM4: customer relations
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

938 IM5: innovative and forward looking

IM6: adding value to user (prestige)
Expectations (EXP) EXP1: expectations for fulfillment of personal need
EXP2: expectations for overall quality
EXP3: expectations for product quality
EXP4: expectations for service quality
PQ PQ1: overall quality
PQ2: product quality (technical)
PQ3: service quality
PQ4: customer services
PQ5: appropriateness to intent of use
PV PV1: price/performance ratio
PV2: performance/price ratio
CSI CSI1: overall satisfaction
Table I. CSI2: fulfillment of expectations
The latent variables and CSI3: compare with ideal
their observable CL CL1: repurchase intention
indicators in the ECSI CL2: recommendation to others
model CL3: price tolerance

5.1 Unidimensionality and convergent validity

CFA was used on the collected data to evaluate the unidimensionality of each of the six
dimensions of ECSI Model. Table II summarizes the results of this analysis.
Results provide the following model statistics for the assessment of goodness-of-fit
(GoF): x 2 statistics, its associated degrees of freedom, p-value of significance, GFI, AGFI,
CFI, and Tucker-Lewis index (TLI). One can conclude that each of the three dimensions
achieves unidimensionality at monomethod levels of analysis. To exemplify, six
items constituting the image dimension were subjected to CFA. The x 2 statistic was
7.331 (degrees of freedom 6, p . 0.05), with the x 2/df ratio having a value of
1.222, which is supposed to be between 0 and 3 with lower values indicating a better
fit. The goodness-of-fit index (GFI) was 0.990 and the adjusted goodness-of-fit

Number of
Construct indicators x2 df p-value GFI AGFI CFI TLI AVE a

IM 6 7.331 5 0.291 0.99 0.96 0.99 0.99 0.51 0.860

EXP 4 0.608 3 0.436 0.99 0.99 1.00 1.00 0.55 0.838
PQ 5 0.049 4 0.976 1.00 0.99 1.00 1.01 0.67 0.905
PV 2 0.916
CSI 3 0.826
CL 3 0.851
Table II.
Initial confirmatory Note: CFA results cannot be obtained due to inadequate number of variables constituting this
factor analysis results dimension
 index (AGFI) was 0.966. These scores are very close to 1.0 (a value of 1.0 indicates a USM approach
perfect fit). The comparative fit index (CFI) was 0.998, while the Tucker-Lewis to CSI
coefficient (TLI) was 0.994. All indices are close to a value of 1.0 in CFA indicating that
the constructs provide good support for the factor structure determined through the
EFA (Schumacker and Lomax, 1996; Hu and Bentler, 1996; Kline, 2005). As is evidenced
by the x 2 statistic with the associated degrees of freedom, the p-value of significance,
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

GFI, AGFI, CFI, and TLI provided in Table II that measure GoF, all of the analyzed 939
dimensions of service quality are unidimensional.
Convergent validity is the extent to which indicators of an ECSI construct converge or
share a high proportion of variance in common. All the individual factor loadings were
found to be highly significant, giving support to the convergent validity.
Average variance extracted (AVE) should exceed the 0.5 threshold value as a
rule-of-thumb (Yi and Davis, 2003). These values are summarized in Table II. The values
of AVE were higher than the recommended value of 0.50, providing further support to
convergent validity of constructs. The standardized regression weights for all variables
constituting each dimension were also found to be significant ( p , 0.01).

5.2 Reliability
The reliability of the scales is related to the homogeneity of their items. Cronbachs a is
commonly used to measure internal consistency of the scales. Basically, it measures the
extent to which the observable variables can explain their corresponding latent
variable and is also supportive reliability measurement criterion (Pedhazur and
Schmelkin, 1991). Table II lists Cronbachs a values of the scales developed in this
study. All the construct measures are over 0.80, thus exhibiting a satisfactory level of
construct reliability. This establishes the internal consistency of the dimensions being
studied in this research.

5.3 The path model

SEM is a comprehensive statistical approach for testing relations between observed and
latent variables. It has been widely utilized in TQM, service quality performance
measurement, and customer satisfaction analyses due to its capability of explaining
casual relationships among the determinative factors/variables (Li, 1997; Prajogo and
Amrik, 2006; Lee et al., 2011). There are two common statistical approaches for structural
model estimation. The most prominent SEM technique is the ML-based covariance
structure analysis method. The second approach is partial least squares (PLS)-based
variance analysis method (Turkyilmaz et al., 2011). Both approaches remain limited to
linear relations among model constructs. To overcome limitations of the traditional SEM
techniques, the USM approach has been developed by Buckler and Hennig-Thurau
(2008) and has been employed in various research domains from international business
(Garbe and Richter, 2009) to medical informatics (Oztekin et al., 2011).
USM builds on the iterative PLS approach for testing structural models but
substitutes its linear least squares regression element with a universal regression
method, namely, a Bayesian neural network. Thus, USM solves the black box problem
inherent to universal regression through its combined use of methods that measure the
strength of model paths and procedures that quantify and visualize nonlinear and
interactive effects among model constructs. Whereas PLS and CVSEM both limit
model estimation to a priori hypothesized paths, USM represents a more
 IMDS exploratory approach that also tests for hidden model structures, namely, theoretically
113,7 unproposed paths, nonlinearity, and interaction effects. USM limits the nonlinear
relations only to the structural model and it assumes that the measurement model part
is linear. The estimator j^ of the latent variable j is defined as the output of multilayer
perceptron (MLP) architecture and shown as in equation (1):
! !
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

940 X H X I
b
jj f w f j
w S j bi
b 1
Act2 h Act1 ih i 1h 2
h1 i1

where fAct1 is the activation function of the hidden neural units and fAct2 is the output
neural unit. H is the number of hidden neural units, I is the number of latent input
variables j, ws are the weights and bs are the biases for the neural network. S ji is the a
priori likelihood that a variable i influences another variable j. To prevent the over
fitting in the neural network model, USM minimizes the error function E for each latent
variable i of the structural model. E refers to the overall error of the respective
variables neural network and shown as in equation (2):
N

X 2 XH X
P
Ei b jd
i
t21;n 2 d
j i
t;n a t;h w2ph 2
n1 h1 p1

where n refers to the individual cases, N is the total number of cases, and p is the index
for the weights, w. On the other hand, jti is the conditional estimate of the latent
variable i in the current estimation step, t, calculated from the structural model by
i
the Bayesian neural network, and jt21;n is the estimate of the previous iteration for the
same latent variable. If the case is the first step for this estimation, jit21;n would then
refer to the initial composite score received from the measurement model. The hyper
parameters a and b prevent overfitting of the neural network model. They are updated
in every iteration of the learning process and are given by equations (3) and (4):
g
ah P N 3
2 n1 w2nh

N 2g
b 4
PN  i b i  2
2 n1 jn 2 jn
P
where N is the total number of records and g pp1 lp =lp aLI 21 . lp are the
eigenvalues of the Hessian matrix of the error function in equation (2) and aLI2 1 is the
hyperparameter a from the previous learning iteration.
USM was employed to test the causal relationships specified in the model as shown
in Figure 1. GoF is an overall coefficient to measure how well the data can explain each
other. It can be calculated as in equation (5) in the context of USM (Buckler and
Hennig-Thurau, 2008):
v
!
u
u 1 X I
GoF t M i communalityi R 2 5
M i1
 where I is the number of latent constructs in the model, M is the total number of USM approach
measurement variables in the model, and Mi is the number of measurement variables
for the construct i. Communality refers to the regression coefficient between an item
to CSI
and its latent variable, and R 2 is the mean explained variance of all endogenous latent
variables of the structural model (Buckler and Hennig-Thurau, 2008).
GoF value was calculated to be 0.678 after three iterations. Considering the degree of
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

freedom, the relationship between each of the constructs is described in three different 941
ways. Table III shows the relationship between each construct.
A value of 0 refers to all excluded paths, and a 1 refers to all the linear
relationships between constructs. On the other hand, a value of 2 means a quadratic
and 3 means a cubic effect between each construct. For every linear path
(cause-and-effect relation with DF 1), Matlabw-based NEUSREL (Buckler, 2001)
software computes a standardized linear path coefficient. This can be compared to those
computed by other causal analysis methods such as PLS. In fact, results of NEUSREL
are observed to be quite similar to PLS when the relations are linear and additive
(i.e. non-interactive).
On the other hand, for all nonlinear additive plots the polynomial coefficients are
included in the model. Polynomial regression has the structure as in equation (6):
y a0 a1 *x a2 *x 2 a3 *x 3 an *x n 6
The relationship between each of the model constructs, as shown in Table III, indicate
that there exist a cubic relationship in company image-CL, PQ-customer satisfaction,
customer satisfaction-CL; and a quadratic relationship in expectation-PQ, PV-customer
satisfaction whereas in linear relationship in company image-customer expectation,
company image-customer satisfaction, customer expectation-PV, customer
expectation-customer satisfaction, PQ-PV.

5.4 Results
Since the main target of this research is to develop an exploratory model that captures
the both linear and nonlinear impacts of various constructs on customer satisfaction
and CL, in this section such relationships are analysed. In analysing nonlinearities,
either progressive or degressive functions can be observed. Degressive means a
change in the endogenous variable takes higher values for low values of the exogenous
variable than for high values of the latter. The opposite holds true for progressive
functions. Namely, endogenous variable takes higher values for high values of the
exogenous variable than for low values of it. The leverage factor is a mean of
quantifying this rationale. Following the above explanations, it is calculated to be the
quotient of a path coefficient calculated only for values higher than average divided by

Construct EXP PQ PV CS CL

IM 1 0 0 1 3
EXP 0 2 1 1 0
PQ 0 0 1 3 0
PV 0 0 0 2 0 Table III.
CS 0 0 0 0 3 Relationships between
CL 0 0 0 0 0 each construct
 IMDS the path coefficient calculated only for values lower than average. Consequently,
113,7 a leverage factor higher than 1 refers to a progressive relationship and lower than 1 refers
to a degressive relationship (Buckler, 2001; Buckler and Hennig-Thurau, 2008). On the
other hand, for human-related studies, it is recommended that the correlation should be
at least 0.3 in behavioral sciences, which translates into a minimum acceptable level of
0.09 for the R 2-value (Cohen et al., 2003). Considering these rule-of-thumb threshold
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

942 values of leverage factor and R 2 simultaneously, the following non-linear relationships
(related to satisfaction and loyalty constructs) were determined to be worth
analyzing further.
Relationship between loyalty and satisfaction. Table IV shows the relationship
between CL and customer satisfaction, which is provided in equation (7):

Loyalty 20:11868 1:5405*Satisfaction 2 3:9882*Satisfaction 2

7
2 2:8067*Satisfaction 3
The first model has one endogenous variable (dependent variable), labeled as CL, and
one exogenous variable (independent variable), which is labeled as customer
satisfaction. There is a positive degressive relationship between satisfaction and
loyalty which is shown in Table IV. It means that increasing lower satisfaction values
have higher positive effect on loyalty than higher satisfaction values. This effect is also
measured by the leverage factor, which was found as a score of 0.632. Figure 2 further
helps visually inspect that nonlinear relationship as shown in equation (7).
Relationship between value and satisfaction. The relationship between PV and
customer satisfaction is provided in equation (8):

Satisfaction 0:023942 0:0076878*Value 2 0:22205*Value 2 8

The endogenous variable (dependent variable) of this equation is named as customer
satisfaction and the exogenous variable of the equation (independent variable) is
labeled as PV. There is a positive progressive relationship between PV and satisfaction

Constructs Model Leverage level Cont. R 2 R 2

Quality-expectation Quality 2 0.094 Expectation*1.482

Expectation2* 2 4.251 0.634 0.66 0.66
Value-quality Value 0.57*Quality 0.35 0.42
Value-expectation Value 0.11*Expectation 0.07
Satisfaction-value Satisfaction 0.023 Value*0.007
Value2* 2 0.222 1.610 0.12 0.72
Satisfaction- Satisfaction 0.07*Expectation
expectation 0.05
Satisfaction-quality Satisfaction 2 0.04 Quality*0.719
Quality2* 2 2.136 Quality3* 2 1.418 0.513 0.35
Satisfaction-image Satisfaction 0.27*Image 0.20
Expectation-image Expectation 0.70*Image 0.49 0.49
Loyalty-image Loyalty 20.019 Image*0.172
Image2* 2 0.239 Image3* 2 0.493 0.652 0.10 0.57
Table IV. Loyalty-satisfaction Loyalty 20.118 Satisfaction*1.54
USM path model results Satisfaction2* 2 3.98 Satisfaction3* 2 2.80 0.632 0.47
 1 USM approach
to CSI
0.5

0
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

943
0.5
Loyalty

1.5

2.5 Figure 2.
Representation of
nonlinear relationship
3 between loyalty and
1 1.5 2 2.5 3 3.5 4 4.5 5 satisfaction
Satisfaction

which is shown in Table IV. It means that increasing lower PVs have lower positive
effect on satisfaction than higher PVs. This effect is also measured by the leverage
factor, which was found as a score of 1.6104. Figure 3 further helps visually inspect
that nonlinear relationship as shown in equation (7). Table IV summarizes the results
of the whole path model and all relationships in the conceptual models.

0.5

0.4

0.3

0.2
Satisfaction

0.1

0.1

0.2
Figure 3.
0.3 Representation of
nonlinear relationship
0.4 between satisfaction
1 1.5 2 2.5 3 3.5 4 4.5 5 and value
Value
 IMDS Relationship between image and CL. The relationship between company image and CL
113,7 is shown in equation (9) and tabulated in Table IV:

Loyalty 20:019 0:172* Image 2 0:239* Image 2 2 0:493* Image 3 9

Following the results in Table IV, a leverage factor level of 0.652 that means the
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

944 relationship between image and CL is degressive. In other words, increasing lower
image values would have a higher positive impact on loyalty than higher image values.
Relationship between PQ and satisfaction. The relationship between customer
satisfaction and PQ as provided in equation (10) is also shown in Table IV:

Satisfaction 20:04 0:719*Quality 2 2:136*Quality 2 2 1:418*Quality 3 10

This relationship has a leverage factor of 0.513, which is less than 1 referring to a
degressive relationship between the two constructs. This translates into the fact that
increasing lower PQ values would have a higher positive impact on satisfaction
compared to higher PQ values.
Relationship between PQ and expectation. Table IV shows the relationship between
PQ and expectation, which is provided in equation (11):

Quality 20:094 1:482*Expectation 2 4:251*Expectation 2 11

This model has a leverage factor level of 0.634. This concludes that the relationship
between PQ and customer expectation is degressive. It indicates that increasing lower
expectation values have higher positive effect on PQ than higher expectation values.

5.5 Comparison of USM results vs conventional PLS results

A conventional PLS algorithm was computed using the same constructs along with
their corresponding indicators for comparison purposes. It was run via the path
weighting scheme with a maximum of 100 iterations and the results presented in
Table V are received after four iterations. The results of this model is also acceptable
but when compared to the USM model results as in Table V it is observed that USM
has a slight superiority in terms of both individual R 2-values of the constructs as well
as the overall GoF score. As seen in Table V, not only all constructs have a higher
R 2-value in the USM model over the PLS but also the overall model has a better GoF
value in the USM model although the difference is marginal.

Model construct USM R 2 PLS R 2

Expectation 0.49 0.48

Image 0 0
Loyalty 0.57 0.53
Table V. Perceived quality 0.66 0.63
Overall comparison Perceived value 0.42 0.35
of USM vs PLS Satisfaction 0.72 0.70
model results GoF of models 0.6784 0.6577
 6. Conclusion and managerial implications USM approach
A number of national CSIs have been developed in the past. Among other national to CSI
CSIs, ECSI has been implemented by many of the EU member countries (Eklof and
Westlund, 2002; Eklof and Westlund, 1998). There is a large body of research in the
literature, which considers the CSI model using either covariance-based or
variance-based SEM technique. The novelty in this research lies in its inclusion of a
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

nonlinear SEM approach via USM while investigating the relationship among 945
constructs in the ECSI model.
Based on theoretical considerations, a model was proposed to link the six constructs
of CSI model. Exploratory and confirmatory factor analyses were employed to produce
empirically-verified and -validated underlying dimensions of expectation, quality, PV,
image, satisfaction, and loyalty. USM was then applied to test the model in the
telecommunications industry with Turk Telekom company as a case study.
The main difference between the USM and PLS model is not attributed to the model fit
results. Instead, the interpretation of the relationships between model constructs is
where the fundamental rationale lies as to why USM should be preferred over the regular
PLS modelling. The findings of this study can be put in practice in various ways. First of
all, the nonlinear relationships revealed via the USM model (Section 5.4) signal out that
the managers should not have a plain logic such as increasing the satisfaction of
customers would definitively increase the loyalty of them towards our product/service.
For the case study in this research, it should be noted that increasing lower values of
satisfaction would have a higher impact on the improvement of loyalty. In other words,
in order to increase the loyalty with a better return-on-investment (ROI) value, managers
should rather focus on unsatisfied customers. If the results in Section 5.4 are taken into
consideration, the same holds true for the impact of image on CL and the impact of PQ on
satisfaction. In contrast, the relationship between value and satisfaction result reveals
that the opposite action should be taken into account in order to enhance the satisfaction
of customers via the PV. It is noteworthy that this study presents uniqueness in terms of
what kind of a strategy should be developed so as to improve determinants of customer
satisfaction and/or CL. In other words, to be able to make use of limited resources of time
and energy; managers would prefer to adopt a strategic move to focus on a specific
portion of the targeted customers in the lights of such this methods findings. This
information is valuable which is extracted via the help of the USM technique and would
not be possible to be learnt through the regular linear SEM approaches.
The findings of this study reveal that quality has the most important impact on
customer satisfaction. The next important construct was found to be image. The
relationship between customer expectation and customer satisfaction was not found to
be significant. This study reveals the fact that while using the ECSI model more
attention must be paid to the consideration of potential nonlinear relationships that
might be available among model constructs. It ways the path for this consideration via
the employment of USM-based ECSI model as exemplified in Turkish communications
sector. Such integration would hypothetically be applicable to other research domains
where customer satisfaction and loyalty are to be modeled.
The results from this USM-based ECSI model provide essential information for
managers in developing marketing strategies as well as a tool for measuring performance
and benchmarking. In addition, USM-based ECSI model delivers important information
for consumers decision-making process with regard to purchasing, which leads to better
 IMDS quality of products and services that they consume. Without doubt, the competition for
113,7 greater market share is intensifying within the telecommunication industry in Turkey.
A more focused approach to building up a companys strengths and unique competitive
edge is vital for survival in the market. This can obviously be achieved by a sound
marketing and customer retention strategy. It is more costly to attract customers than to
retain them. So, the key focus in managing customer satisfaction is to identify the core
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

946 satisfaction determinants from the users perspective and then to assess the companys
performance in addressing each of these determinants. One of the challenging tasks that
Turk Telekom faces is how to improve its image and satisfaction level of their customers.
They must strive to improve features and enhance service quality so that they can
improve customers experiences with communication tools and in particular to increase
overall reliability. Brand managers should also focus on enhancing their companys image
and develop advertising and promotional messages that could encourage customers to
think about their experiences with the actual services. This needs to be accomplished
simultaneously as reducing costs, if the existing telecommunication company desires to
maximize customer satisfaction and gain high market penetration through brand image.
The study, however, is subject to some limitations. While the findings of this study
confirm the direct and positive relationships among the constructs of the CSI model, they
cannot be generalized to all telecommunication companies in Turkey due to the nature of
the case study approach. The sensitive nature of the subject and the availability of
personal connections, however, have made the selection of case study methodology
mandatory over other large-scale quantitative surveys. Second, due to relatively small
sample size, a caution should be exercised when interpreting the results. Yet our findings
provide a base on which future research with larger sample sizes can build. Third, we
relied on the subjective evaluations of Turk Telekoms clients. Given the limitations
associated with subjective measures, findings should be interpreted with some degree of
caution. It would therefore be useful to replicate the results with objective measures of
service quality indicators. As the data used in this study were mainly collected from
Istanbul, future research based on a sample of various cities in Turkey, and a
comparison between Turk Telekom and other telecommunication companies may
provide important insights to practitioners and researchers. Finally, other estimation
methods, such as machine learning tools can also provide further insights into the
interrelationships among ECSI model constructs.

References
Anderson, E.W. and Fornell, C. (2000), Foundations of the American customer satisfaction
index, Journal of Total Quality Measurement, Vol. 11 No. 7, pp. 869-882.
Anderson, E.W., Fornell, C. and Lehmann, D.R. (1994), Customer satisfaction, market share, and
profitability: findings from Sweden, Journal of Marketing, No. 58, pp. 53-66.
Andreassen, T.W. and Lindestad, B. (1998), The effects of corporate image in the formation of
customer loyalty, Journal of Service Marketing, Vol. 9 No. 1, pp. 82-92.
Aydin, S. and Ozer, G. (2005), National customer satisfaction indices: an implementation in the
Turkish mobile telephone market, Marketing Intelligence and Planning, Vol. 23 No. 5,
pp. 486-504.
Bayraktar, E., Tatoglu, E., Turkyilmaz, A., Delen, D. and Zaim, D. (2012), Measuring the
efficiency of customer satisfaction and loyalty for mobile phone brands with DEA, Expert
Systems with Applications, Vol. 39 No. 1, pp. 99-106.
 Buckler, F. (2001), Neusrel Neuer Kausalanalyseansatz auf Basis Neuronaler Netze als USM approach
Instrument der Marketingforschung, Cuvillier, Gottingen, available at: http://neusrel.com/
to CSI
Buckler, F. and Hennig-Thurau, T. (2008), Identifying hidden structures in marketings
structural models through universal structure modeling: an explorative Bayesian neural
network complement to LISREL and PLS, Marketing Journal of Research and
Management, Vol. 4 No. 2, pp. 47-66.
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

Chan, L.K., Hui, Y.V., Lo, H.P., Tse, S.K., Tso, G.K.F. and Wu, M.L. (2003), Consumer satisfaction 947
index: new practice and findings, European Journal of Marketing, Vol. 37 Nos 5/6,
pp. 872-909.
Chien, T.-K., Chang, T.-H. and Su, C.-T. (2003), Did your efforts really win customers
satisfaction?, Industrial Management & Data Systems, Vol. 103 No. 4, pp. 253-262.
Chien, T.-K., Su, C.-H. and Su, C.-T. (2002), Implementation of a customer satisfaction program:
a case study, Industrial Management & Data Systems, Vol. 102 No. 5, pp. 252-259.
Cohen, J., Cohen, P., West, S.G. and Aiken, L.S. (2003), Applied Multiple Regression/Correlation
Analysis for the Behavioral Sciences, Erlbaum, Mahwah, NJ.
Eklof, J. and Westlund, A.H. (2002), The pan European customer satisfaction index program,
Total Quality Management, Vol. 13 No. 8, pp. 1099-1106.
Eklof, J.A. and Westlund, A.H (1998), Customer satisfaction index and its role in quality
management, Total Quality Management, Vol. 9 Nos 4/5, pp. 80-85.
Fornell, C. (1992), A national satisfaction barometer: the Swedish experience, Journal of
Marketing, No. 56, pp. 6-21.
Fornell, C., Johnson, M.D., Anderson, E.W., Cha, J. and Bryant, B.E. (1996), The American
customer satisfaction index: nature, purpose and findings, Journal of Marketing, No. 60,
pp. 7-18.
Garbe, J.N. and Richter, N.F. (2009), Causal analysis of the internationalization and performance
relationship based on neural networks advocating the transnational structure, Journal
of International Management, Vol. 15 No. 4, pp. 413-431.
Grigoroudis, E. and Siskos, Y. (2003), A survey of customer satisfaction barometers: some
results from the transportation-communications sector, European Journal of Operational
Research, Vol. 152 No. 2, pp. 334-353.
Hirschman, A.O. (1970), Exit, Voice, and Loyalty Responses to Declines in Firms, Organizations
and States, Harvard University Press, Cambridge, MA.
Hu, L.T. and Bentler, P.M. (1996), Cutoff criteria for fit indexes in covariance structure analysis:
conventional criteria versus new alternative, Structural Equation Modeling, Vol. 6 No. 1,
pp. 1-55.
ICTA (2012), available at: www.tk.gov.tr/kutuphane_ve_veribankasi/pazar_verileri/ucaylik12_
1.pdf
Johnson, M.D. and Fornell, C. (1991), A framework for comparing customer satisfaction across
individuals and product categories, Journal of Economic Psychology, Vol. 12 No. 2,
pp. 267-286.
Johnson, M.D., Gustafsson, A., Andressen, T.W., Lervik, L. and Cha, J. (2001), The evolution and
future of national customer satisfaction index models, Journal of Economic Psychology,
Vol. 22 No. 2, pp. 217-245.
Kannan, V.R. and Tan, K.C. (2005), Just in time, total quality management, and supply chain
management: understanding their linkages and impact on business performance,
OMEGA, Vol. 33 No. 2, pp. 153-162.
 IMDS Kline, R.B. (2005), Principles and Practice of Structural Equation Modeling, 2nd ed., The Guilford
Press, New York, NY.
113,7
Lee, P.K.C., Cheng, T.C.E., Yeung, A.C.L. and Lai, K.H. (2011), An empirical study of
transformational leadership, team performance, and service quality in retail banks,
OMEGA, Vol. 39 No. 6, pp. 690-701.
Li, L.X. (1997), Relationships between determinants of hospital quality management and service
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

948 quality performance a path analytic model, OMEGA, Vol. 25 No. 5, pp. 535-545.
Li, W. and Xu, L.C. (2002), The political economy of privatization and competition: cross-country
evidence from the telecommunications sector, Journal of Comparative Economics, Vol. 30
No. 3, pp. 439-462.
Martensen, A., Kristensen, K. and Grnholdt, L. (2000), Customer satisfaction measurement at
post Denmark: results of application of the European customer satisfaction index
methodology, Total Quality Management, Vol. 11 No. 7, pp. 1007-1015.
Oliver, R.L. (1980), A cognitive model of the antecedents and consequences of satisfaction
decisions, Journal of Marketing Research, Vol. 17, pp. 460-469.
Oztekin, A., Kong, Z.J. and Delen, D. (2011), Development of a structural equation
modeling-based decision tree methodology for the analysis of lung transplantations,
Decision Support Systems, Vol. 51 No. 1, pp. 155-166.
Pedhazur, E.J. and Schmelkin, L.P. (1991), Measurement, Design, and Analysis: An Integrated
Approach, Lawrence Erlbaum Associates, Hillsdale, NJ.
Prajogo, D.I. and Amrik, S.S. (2006), The integration of TQM and technology/R&D
management in determining quality and innovation performance, OMEGA, Vol. 34
No. 3, pp. 296-312.
Schumacker, R.E. and Lomax, R.G. (1996), Beginners Guide to Structural Equation Modeling,
Erlbaum, Mahwah, NJ.
TMME (2006), Turkish customer satisfaction index, available at: www.tmme.org.tr
Turk Telekom (2012), available at: www.turktelekom.com.tr/tt/portal/About-TT/Company-
Profile/
Turkyilmaz, A. and Ozkan, C. (2007), Development of a customer satisfaction index model: an
application to the Turkish mobile phone sector, Industrial Management and Data
Systems, Vol. 5 No. 107, pp. 672-687.
Turkyilmaz, A., Akman, G., Ozkan, C. and Pastuszak, Z. (2011), Empirical study of public sector
employee loyalty and satisfaction, Industrial Management & Data Systems, Vol. 111 No. 5,
pp. 675-696.
Yi, M.Y. and Davis, F.D. (2003), Developing and validating an observational learning model of
computer software training and skill acquisition, Information Systems Research, Vol. 14
No. 2, pp. 146-169.

Further reading
Gerbing, D.W. and Anderson, J.C. (1988), An updated paradigm for scale development
incorporating unidimensionality and its assessment, Journal of Marketing Research,
Vol. 25 No. 2, pp. 186-192.
OECD (2009), OECD Annual Report 2009, available at: www.oecd.org/newsroom/43125523.pdf
Xu, Y., Yen, D.C., Lin, B. and Chou, D.C. (2002), Adopting customer relationship management
technology, Industrial Management & Data Systems, Vol. 102 No. 8, pp. 442-452.
 About the authors USM approach
Ali Turkyilmaz, Associate Professor, is the Head of Industrial Engineering Department at
Fatih University. He got his PhD degree in Industrial Engineering Department of Istanbul to CSI
Technical University. His main research interests are production and service sector management,
total quality management, applied statistics, and decision making. Dr Turkyilmaz reviews papers
for a variety of journals He is also a consultant and trainer for industrial companies, an
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)

independent expert for EU projects, and a member of Turkish Quality Organization and Turkish
Operations Research Foundation. Ali Turkyilmaz is the corresponding author and can be 949
contacted at: aturkyilmaz@fatih.edu.tr
Asil Oztekin received the BS degree from Yildiz Technical University in 2004, and the MS
degree from Fatih University in 2006, both in Industrial Engineering. He completed his PhD
degree in Industrial Engineering and Management at Oklahoma State University in 2010. He is
currently working as an Assistant Professor in the Department of Operations and Information
Systems at the University of Massachusetts Lowell. His research interests include
human-computer interaction, usability engineering, quality improvement, medical informatics,
decision analysis, and data mining techniques. Dr Oztekin has reviewed manuscripts for IIE
Transactions, Decision Support Systems, International Journal of Production Research,
Computers in Industry and Journal of Systems & Software.
Selim Zaim has received his BS degree in Mechanical Engineering from Istanbul Technical
University and his PhD degree in Production and Operations Management from Istanbul
University. Dr Zaim has been serving as a full Professor in the College of Technology at
Marmara University. Dr Zaim has published over 100 articles and papers in various journals and
congress proceedings. His current scholarly interests focus on multivariate data analysis,
human-computer interaction, and multi-criteria decision making. Dr Zaim reviews papers for a
variety of journals. He is a member of Industrial Management and Development Associations
(IMDA) and Quality Association in Turkey (KALDER).
Omer Fahrettin Demirel has received his BS degree in Business Administration from Marmara
University. He has completed the MS degree in Operations Research and Statistics and PhD degree
in Decision Sciences and Engineering Sciences from Rensselaer Polytechnic Institute. He is
currently working as an Assistant Professor in Industrial Engineering Department at
Fatih University. His research interests are forecasting, supply chain management and time
series analysis. He is a member of the Institute for Operations Research and the Management
Sciences (INFORMS).

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com

Or visit our web site for further details: www.emeraldinsight.com/reprints
 This article has been cited by:

1. Jia-Bei Yu, Yang Yu, Lin-Na Wang, Ze Yuan, Xu Ji. 2014. The knowledge modeling system of ready-
mixed concrete enterprise and artificial intelligence with ANN-GA formanufacturing production. Journal
of Intelligent Manufacturing . [CrossRef]
Downloaded by SHAHEED ZULFIKAR ALI BHUTTO INST OF SCI & TECH KARACHI At 00:40 12 March 2016 (PT)