ARTICLE IN PRESS

Journal of Retailing and Consumer Services 16 (2009) 352–359

Contents lists available at ScienceDirect

Journal of Retailing and Consumer Services
journal homepage: www.elsevier.com/locate/jretconser

Customer satisfaction study via a latent segment model
Jaime R.S. Fonseca Ã
Technical University of Lisbon, Polo Universitario do Alto da Ajuda, Institute of Social and Political Sciences, Rua Almerindo Lessa, Lisboa 1300-663, Portugal

a r t i c l e in f o

a b s t r a c t
The aim of this study is to apply a new conceptual model, and a new technique as an approach to the modelling of customers’ satisfaction, and to develop an overall satisfaction index (OSI). This study evaluates customers’ satisfaction of a certain public organization service, and argues that in order to estimate the global customers’ satisfaction measure we must appeal to methodologies recognizing that satisfaction must be understood as a latent variable, quantified through multiple indicators. Thus, it is natural that we consider the latent segment models (LSM) approach to proceed to the evaluation of customer’s service satisfaction. As a result of these models estimation, we selected a three latent segment model, that is to say, the latent variable customer satisfaction has three classes: segment 1, with 50.4 percent of the customers, that represents ‘‘The Very Satisfied’’, for those to whom everything is very well in the organization service; a segment 2, with 33.4 percent of the customers, representative of the ‘‘The Well Satisfied’’, not totally satisfied with the quality of the organization, and a segment 3, with 16.2 percent of the customers, ‘‘Satisfaction Demanders’’, thinking that organizational quality can be improved. Finally, we developed an overall satisfaction index which is important to show how the company as a whole is performing. & 2009 Elsevier Ltd. All rights reserved.

Keywords: Quantitative methods Service quality American customer satisfaction index Customer satisfaction Overall satisfaction index Latent segment models Information criteria

1. Introduction Customer satisfaction is central to the marketing concept, with evidence of strategic links between satisfaction and overall service performance (Truch, 2006), and is an important theoretical and practical issue for most marketeers and consumer researchers (Goode, 2001); it is a key issue for all those organizations that wish to create and keep a competitive advantage in this highly competitive world. Customer satisfaction which remains in the limelight (Bartikowski and Llosa, 2004), especially in the service field, is typically defined as an overall assessment of the performance of various attributes that constitute a service. The organization wants to know how satisfied their customers are in order to be translated into marketing strategy and organizational development. First, it was important to understand the ways that services can influence customer behaviour in terms of satisfaction, so that we may achieve a consistent customer satisfaction measure, knowing that satisfaction level increases as the congruence between the organization’s goals and the customers’ interest also increases (Garbarino and Johnson, 2001). This service is a non-profit professional service, social service, whose customers are the organization’s employees and organization’s retired employees. For the organization management,

customer satisfaction could be indirectly measured by means of several response determinants (e.g. performance, equity, expectation, disconfirmation, attribution, etc.), and these impacts on satisfaction are heterogeneous (Wu and DeSarbo, 2005).

2. Literature review and conceptualization of service quality Service quality has been studied for a long time. However, this literature also suggests that there is no consensus on how to ´ conceptualize perceived service quality (Caro and Garcıa, 2007), and two different approaches have been adopted regarding this issue, mainly because of the difficulties involved in delimiting and measuring the construct (Parasuraman et al., 1985). The first one suggests that perceived service quality is based on the disconfirmation paradigm (by a comparison between customers’ expectations and their perceptions of the received service) (Gronroos, 1984; Parasuraman et al., 1985). The second approach suggests that service quality should be measured considering only customer perceptions rather than ´ expectations minus perceptions (Caro and Garcıa, 2007). Nowadays we can see a movement away from using expectations, and the theoretical background of service quality is moving from expectation disconfirmation to the theory of reasoned action which states that the behaviour of individuals can be predicted from their intentions, which can be predicted from their attitudes about the behaviour and subjective norms (Collier and Bienstock, 2006).

à Tel.: +351 21 361 94 30.

E-mail address: jaimefonseca@iscsp.utl.pt 0969-6989/$ - see front matter & 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.jretconser.2009.04.001

ARTICLE IN PRESS
J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359 353

It is well known that service quality and customer satisfaction are distinct constructs (Dabholkar, 2000). Another important question was answered by Oliver (1993), which first suggests that service quality would be antecedent to customer satisfaction regardless of whether these constructs were measured for a given experience or over time. Spreng and Macoy (1996) find empirical support for this model, wherein customer satisfaction is a consequence of service quality, and Dabholkar (2000) proves that customer satisfaction is a consequence of service quality (mediator model of customer satisfaction). The results of Bodet (2006) suggest that the quality of human factors, such as staff behaviour, and non-tangible factors, such as image, are determinant in the formation of customer satisfaction. In this sense, by knowing customers’ perceptions about service quality we think that we can measure customer’s service satisfaction, using service quality as an indirect approach to customer satisfaction. Because of the difficulty in measuring the customers’ expectations about a service quality (can they have expectations about unknown services?), we think that quality is about conformance to a service design or service specification. Once the design is set, quality is about ensuring that the end to be delivered to the customer meets this specification or design. As a consequence, from a service point of view, customer satisfaction is about monitoring the quality of delivery of the service, thus measuring how well the organization is delivering the providing service. Services can only be experienced, and the production of a service takes place at the same time and in the same place as its consumption. The perception of service quality by customers during service delivery will be influenced mainly by three factors: technical quality (what the supplier delivers), result of know-how available to the organization, with objective evaluations; functional quality (how the supplier delivers), representing the way the service is provided (staff appear to be a key element in the service encounter and more precisely their capacity to answer or solve problems encountered by the customer on the premises, Bodet, 2006); the image (of the organization which is delivering the service, and the supplier’s corporate image). In order to provide insights for marketing managers to make better customer satisfaction measurement decisions, we think that service performance, with technical quality, functional quality and corporate image, is the best determinant of overall customer satisfaction in this particular service. Bearing this service in mind, we think that increasing service performance is the key to increasing customer satisfaction (all the coefficient correlations between technical quality and functional quality are significant at the 0.01 level, Table 11). An important theoretical advantage of this approach is that its results are derived from actually experienced services performances. Though building on and extending (see Harris and Goode, 2004; Oliver, 1997) forwards a framework of service that presents quality which leads to satisfaction (which in turn affects loyalty). Customers completely satisfied (dissatisfied) with service quality and corporate image will be completely satisfied (dissatisfied) with service. So, following Oliver (1997), stating that service ´ quality leads to satisfaction, and Caro and Garcıa (2007), suggesting that service quality should be measured considering only customers’ perceptions rather than expectations minus perceptions, we intend to present a simplified conceptualization model (Fig. 1).

Indicators TECHNICAL QUALITY →

Latent

FUNCTIONAL QUALITY

CORPORATE IMAGE

Fig. 1. Model conceptualization.

3. Measures and methodology The organization uses a survey tool to collect these data from their key customer base, that is to say the target population being

the visitors/customers to the service. All of the customers are post experience, because they can only be satisfied or not with the service, having experienced it. Because the number and nature of service quality dimensions is in direct relation to the service under analysis, the questionnaire used in this study was designed through a lot of discussions with the organization manager, after careful literature review. We use a careful questionnaire about different aspects of the service because the more detailed the information is, the more useful it is likely to be for improving the service. At a preliminary scale we use a set of 23 items representing all relevant sides of service quality as input to customer satisfaction. By using focus group interviews with students of the Technical University of Lisbon we simplified the scale, by eliminating some confusing items, and rewording others. Because of customers’ satisfaction was our main goal, we asked from customers’ service how satisfied they were with service quality and image; the final questionnaire had 18 items with a 10point Likert-type scale, ranging from 1 (indicates an extremely negative classification, not at all satisfied or dissatisfied with service quality and image) to 10 (indicates an extremely positive classification, completely satisfied with service quality and image). Because overall satisfaction depends on how the customers experience the quality of different aspects like, for example, service quality and expectations, reception and welcome, professional reliability, the orientation to the customer, we used the attributes we present in Table 1. These variables are the indicator variables or segmentation base variables, and the LSM is indicated because we have no response variable on global satisfaction in the questionnaire to indicate as dependent variable. For assessing content validity, the survey questionnaire was subjected to pre-test and refinement trough a pilot study of 70 randomly selected customers. The data for this study was collected from the service customers, using a face to face interviewing technique. An initial sample of 873 customers was obtained, but 17 questionnaires were considered non-valid. The final sample was representative of the individuals’ population heterogeneity with regard to demographic characteristics such as service, customer kind, gender and education. Reliability was examined through confirmatory factor analysis and each indicator loaded significantly on its designated factor (pvalueo0.01). Overall, the analysis produced chi-squared-degrees of freedom ratio well below the criterion of Marsh and Hocevar (1985) and adjusted goodness-of-fit significantly better than a one factor model. Reliability was also gauged via the Cronbach alpha coefficient, that Churchill (1979) suggested should be over 0.7 for a scale to be considered reliable; in this study, the Cronbach alpha coefficients range from 0.72 to 0.96. We ran factor analysis with these items, in order to see if they were structurally related. The value of Kaiser–Meyer–Olkin (KMO)

CUSTOMER SATISFACTION

ARTICLE IN PRESS
354 J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359

Table 1 The survey items. Technical quality Service quality and demand Service quality perception The ideal public service Global quality of services Reception and welcome Services area Services location Services identification Professional knowledge Information clarity Waiting time Liking, attention and professional interest Professional reliability Interest on customer Modernity Credibility The orientation to the customer The spreading of the services Service Customer kind Gender Education

Table 3 Rotated component matrixa. Component 1 Service location Service identification Reception and welcome area Services area Professional knowledge Information clarity Waiting time Liking, attention and interest Professional reliability Interest in customer problems Modernity of procedures and technological supports Credibility and inspired confidence Orientation to serve the customer Spreading of given services Service quality and demand Service quality perception The ideal public service Global quality of services Extraction method: principal component analysis. Rotation method: varimax with Kaiser normalization. a Rotation converged in five iterations. 0.03 0.11 0.26 0.36 0.86 0.88 0.82 0.90 0.88 0.87 0.19 0.64 0.45 0.11 0.38 0.34 0.67 0.60 2 0.58 0.84 0.87 0.81 0.23 0.20 0.15 0.10 0.15 0.19 0.64 0.40 0.28 0.53 0.28 0.27 0.27 0.22 3 0.37 0.18 0.06 0.15 0.26 0.25 0.17 0.19 0.20 0.22 0.18 0.32 0.71 0.44 0.80 0.82 0.40 0.58

Functional quality

Corporate image

Socio-demographic

Table 2 Items and factor loadings. Items Customer kind Gender Education (school attendance) Service location Service identification Reception and welcome area Services area Professional knowledge Information clarity Waiting time Liking, attention and interest Professional reliability Interest in customer problems Modernity of procedures and technological supports Credibility and inspired confidence Orientation to serve the customer Spreading of given services Service quality and demand Service quality perception The ideal public service Global quality of services Factor loadings 0.529 0.587 0.674 0.545 0.718 0.869 0.833 0.857 0.867 0.729 0.852 0.817 0.830 0.495 0.691 0.744 0.531 0.786 0.781 0.682 0.730

measure, 0.879, and Bartlett’s test of Sphericity (any p-value) indicate that the data is suitable for factor analysis application. As we can see, almost all of the factor loadings of the items are significant (Table 2) and so, Technical quality, Functional quality and Corporate image could be viewed as the core value items of the service in increasing overall customer satisfaction for the further analysis. The results for principal component analysis (Tables 2 and 3) showed that the items of Table 1 loaded on three factors (eigenvalues over 1) as displayed in Table 3, rotated component matrix. Those factors that explained 74 percent of the meaningful variation in the initial items, roughly represent Technical quality, Functional quality and Corporate image.

4. The model and model selection Many organizations have felt the critical need to use a tool for evaluating service quality in order to appropriately assess and

improve their service performance and consequently improve customer satisfaction. Even though customer satisfaction cannot be directly observed, it is possible to employ indicators to capture empirically the construct (Anderson and Fornell, 2000). This study measures customer satisfaction in a public administration service, based on a questionnaire inspired on the American customer satisfaction index (ACSI) model. This index, measuring overall customer satisfaction, is a customer evaluation tool for aspects that cannot be measured directly (Anderson and Fornell, 2000). So, the methodology used to analyse data must recognize this and must also be able to measure this latent variable (customer satisfaction) by multiple observable indicators. This article examines a customer satisfaction model for assessing the relationship of overall satisfaction with a service. Thus, we propose measuring customer’s satisfaction indirectly as a latent variable, by the used indicators, estimating latent segment models (LSM), assuming that there is heterogeneity, which is natural in services because a service supplier consists of different staff all working on the provision of the same services. That is, we suggest market segmentation for customer satisfaction study, via LSM, because it provides a probabilistic methodology for segmentation, based on the indicator (observable) variables. Homogeneity within the segment is critical to defining the target customer. In LSM, customers that exhibit the most similar attributes’ results would be grouped in one segment. Heterogeneity across the segments allows for the differentiation of segments and customers. The latent segment model was initially introduced by Lazarsfield and Henry (1968) as latent class model, assuming that the latent variable is categorical, in segmentation. LSM (e.g., Cohen and Ramaswamy, 1998; Fonseca and Cardoso, 2005), are used to identify the latent segments required to explain the associations among a set of observed variables (segmentation base variables) and to allocate observations to these segments. It represents a model-based approach to clustering, which connects clustering with classical statistical estimation methods, and assumes that the variables’ observations in a sample arise from different segments of unknown proportions. Customer heterogeneity in satisfaction requires marketing managers to

ARTICLE IN PRESS
J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359 355

segment the market and make segment-specific customer satisfaction measurement decisions. The use of LSM has become increasingly popular in the marketing literature, for instance Wedel and Kamakura (1998), Dillon and Kumar (1994), and Bhatnagar and Ghose (2004). This approach to segmentation offers some advantages when compared with other segmentation techniques: it identifies market segments, and provides unbiased market segment memberships estimates (Dillon and Kumar, 1994); it provides means to selecting the number of segments (McLachlan and Peel, 2000); it is able to deal with different measurement levels (Vermunt and Magidson, 2002); demographic and other covariates can be used for segment description (Magidson and Vermunt, 2003); it allocates cases into segments based upon membership probabilities estimated directly from the model, instead of using an ad-hoc definition of ‘‘distance’’ (e.g., Euclidian distance) (Bonilla and Huntington, 2005). Compared with variable precision rough set theory (VPRS) method, it offers a very useful methodology by estimating probabilistic rules which can predict customers’ behaviour (or to make prediction of corporate failure), based on a number of exogenous variables (Benyon et al., 2005); LSM or unsupervised learning, because of the absence of dependent variable, do not make prediction, but create a typology based on probabilities. VPRS methods only address discrete data, which limits its usefulness in certain types of analysis; on the contrary, LSM apply with only discrete data, with only continuous data, and with both discrete and continuous data. The primary focus and contribution of this manuscript is to present a simplified conceptualization model to access customers’ satisfaction; the second one consists in using latent segmentation methodology in order to provide a way for marketing managers to make segment-specific decisions in customer satisfaction measurement. This methodology accommodates multiple performance attributes (including mixed case), provides parsimonious models in order to account for the relationships between these multiple performance’s attributes, and derives latent segments for customers’ overall satisfaction, based on these attributes or indicators on technical and functional quality and corporate image. Let yi ¼ ðyip Þ denote the vector representing the scores of the ith case for the pth segmentation base attributes (i ¼ 1,y,n; p ¼ 1,y,P). We consider that the cases on which the attributes are measured arise from a population which we assume to be a mixture of S segments, in proportions ls (mixing proportions or relative segment sizes), s ¼ 1,y,S. The statistical probability density function of the vector yi, given that yi comes from segment s, is represented by f s ðyi jys Þ, with ys representing the vector of unknown parameters associated with the specific pdf chosen. We are dealing with mixture models of nominal variables, thus multinomial distributional models are used (Fonseca and Cardoso, 2007a), that is to say, f s ðyi jys Þ is multinomial. Then the population density can be represented as a finite mixture of the densities f s ðyi jys Þ of S distinct segments, i.e. f ðyi j cÞ ¼ where
S X s¼1

LSM naturally provides means for constituting a partition by means of assigning each case to the segment with the highest ^ segment-membership probability, Maxs¼1;...;S tis , where ^ ^ tis ¼ ts ðyi jcðkÞ Þ ¼ P ^ ^ ls f s ðyi jys Þ
ðkÞ ðkÞ

S ^ ðkÞ ^ ðkÞ j¼1 lj f j ðyi jyj Þ

(2)

In order to derive meaningful results from clustering, the mixture model must be identifiable, that is, a unique maximum likelihood solution should exist (Bozdogan, 1994). A goal of traditional LSM estimation is to determine the smallest number of latent segments S that is sufficient to explain the relationships observed among the variables of segmentation base variables. If the baseline model (S ¼ 1) provides a good fit to the data, no LSM is needed, since there is no relationship among the variables to be explained; otherwise, a model with S ¼ 2 segments is then fitted to the data. This process continues by fitting successive LSM to the data, every time adding another dimension by incrementing the number of segments by 1, until a parsimonious model is found that provides an adequate fit. In order to select the best number of segments and in an attempt to overcome most of the limitations of likelihood ratio tests (regularity conditions in finite mixtures do not hold) (Ramaswamy et al., 1996), theoretical information criteria can be used. They assist in determining the adequate value of S based on minimum criteria values. ^ The general form of information criteria is À log LðcÞ þ C. The first term is the negative logarithm of the maximum likelihood which decreases when model complexity increases; the second term, or penalty term, penalizes too complex models, and increases with the model number of parameters ðnc Þ.The selected LSM should evidence a good trade-off between good description of the data and the model number of parameters. The second focus and contribution of this manuscript consists in using an adequate information criterion for selecting the best LSM. In a recent work, Fonseca and Cardoso (2007a), based on an empirical analysis, evidences the good performance of AIC3 when dealing with only categorical segmentation base variables which ^ is defined by À2 log LðcÞ þ 3nc (Bozdogan, 1987).

5. Results and discussion As a result of LSM estimation, we selected, by means of the used information criterion AIC3, a three-segment solution, as displayed in Table 4, because AIC3 minimizes for S ¼ 3. Tables 5–7 display two kinds of probabilities, estimated by LSM, in order to characterize the customers’ typology; the probabilities ls (s ¼ 1,y,S) of belonging to segment s, and probabilities f s ðyi jys Þ, of being on a variable category, conditional on belonging to a segment s. This is a probabilistic model and it makes that LSM is easily interpretable. Thus, we can now understand the service customers, concerning the considered variables and so we can have a better understanding about customers’ satisfaction. We are going

ls

P Y p¼1

f s ðyi jys Þ
S P s¼1

(1)

i ¼ 1,y,n,

ls 40,

ls ¼ 1,

c ¼ fl; Yg,

with

fl1 ; . . . ; lsÀ1 g; Y ¼ fy1 ; . . . ; ys g, and c is the vector of all unknown parameters. The LSM estimation problem, simultaneously addresses the estimation of distributional parameters and classification of cases into segments, yielding mixing probabilities. The estimation process is typically directed to maximum likelihood using the expectation-maximization (EM) algorithm (Dempster et al., 1977; McLachlan and Peel, 2000; Figueiredo and Jain, 2002).

Table 4 Information criterion values. Model 1—Latent 2—Latent 3—Latent 4—Latent segment segment segment segment AIC3 8381.9 7278.8 7093.4 7152.2

ARTICLE IN PRESS
356 J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359

Table 5 Customers’ technical quality profile by model parameters’ estimates. Segment 1 50.4 percent Segment 2 33.4 percent Segment 3 16.2 percent

Table 6 Customers’ functional quality profile by model parameters’ estimates. Segment 1 50.4 percent Segment 2 33.4 percent Segment 3 16.2 percent

Relative segment size (ls) Variables Service quality and demand 6 7 8 9 10 Service quality perception 5 6 7 8 9 10 The ideal public service 4 5 6 7 8 9 10 Global quality of services 6 7 8 9 10 Reception and welcome 5 6 7 8 9 10 Service area 5 6 7 8 9 10 Service location 5 6 7 8 9 10 Service identification 5 6 7 8 9 10

Relative segment size (ls) Variables Professional knowledge 6 7 8 9 10 Information clarity 6 7 8 9 10 Waiting time 6 7 8 9 10 Liking, attention and interest 6 7 8 9 10 Professional reliability 6 7 8 9 10

0.01 0.00 0.04 0.25 0.70 0.01 0.00 0.01 0.04 0.29 0.65 0.00 0.00 0.00 0.00 0.05 0.29 0.66 0.01 0.00 0.01 0.24 0.74 0.01 0.01 0.04 0.25 0.24 0.45 0.00 0.01 0.02 0.25 0.24 0.47 0.04 0.05 0.12 0.26 0.26 0.24 0.01 0.08 0.08 0.26 0.21 0.33

0.00 0.04 0.20 0.67 0.08 0.02 0.02 0.08 0.32 0.48 0.08 0.00 0.02 0.00 0.08 0.26 0.60 0.04 0.02 0.00 0.16 0.60 0.22 0.02 0.10 0.18 0.38 0.28 0.04 0.02 0.10 0.12 0.30 0.44 0.02 0.07 0.10 0.16 0.41 0.22 0.04 0.10 0.14 0.16 0.22 0.28 0.06

0.08 0.37 0.45 0.09 0.01 0.04 0.17 0.29 0.37 0.13 0.00 0.04 0.12 0.21 0.29 0.25 0.09 0.00 0.12 0.41 0.37 0.09 0.01 0.17 0.10 0.25 0.33 0.09 0.04 0.16 0.16 0.21 0.37 0.00 0.04 0.09 0.16 0.21 0.54 0.00 0.00 0.17 0.17 0.27 0.39 0.08 0.00

0.00 0.00 0.00 0.01 0.98 0.00 0.00 0.00 0.07 0.93 0.00 0.00 0.05 0.11 0.84 0.00 0.00 0.00 0.03 0.97 0.00 0.00 0.00 0.00 1.00

0.00 0.00 0.20 0.72 0.09 0.00 0.00 0.14 0.77 0.09 0.00 0.00 0.12 0.65 0.21 0.00 0.00 0.04 0.65 0.31 0.00 0.02 0.04 0.71 0.23 0.00 0.00 0.20 0.65 0.15

0.12 0.33 0.37 0.09 0.09 0.16 0.25 0.49 0.05 0.05 0.08 0.29 0.33 0.05 0.17 0.08 0.12 0.41 0.21 0.17 0.04 0.25 0.45 0.09 0.17 0.04 0.37 0.41 0.13 0.05

Interest in customer problems 6 0.00 7 0.00 8 0.00 9 0.07 10 0.93

to initiate the results discussion, trying to extract knowledge about service customers’ typology. For instance, the probability 0.70 (bold, in Table 5) represents the probability of an individual answer 10 for service quality and demand, given that he/she belongs to segment 1. In the same line, 0.08 represents the probability of an individual answering 10 for service quality and demand, given that he/she belongs to segment 2, and 0.01 represents the probability of an individual answering 10 for service quality and demand, given that he/she belongs to segment 3.

Because the greater probability value is in segment 1, we associate the individual that provides this answer (10) to segment 1. Thus, the information of these tables (probabilities) is easily interpretable, and this gives us a way to describe the individual pattern about satisfaction. These estimates of conditional probabilities displayed in Tables 5–7 allow us to name the three segments as follows: a segment 1, with 50.4 percent of customers, which represents the very satisfied (top), and for them, everything is very well in the organization service; a segment 2, with 33.4 percent of customers, which represents the well satisfied; a segment 3, with 16.2 percent of customers, which represents the satisfaction demanders. As regards the very satisfied, by the customer’s technical quality profile (Table 5), we can see that they classify all the items with 10 (customers are completely satisfied with service technical quality). For them, the service quality is extremely positive in all items of technical quality, and they are naturally very satisfied. As for functional quality (Table 6), again, they consider the staff as extremely positive (because their answers are almost all about 10, customers are very well satisfied with service functional quality). As for corporate image (Table 7), they also consider the service image as extremely positive, because their answers are almost all about 10, customers are very well satisfied with corporate image. Thus, customers of segment 1 belong to a high level (top) of customer satisfaction.

ARTICLE IN PRESS
J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359 357

Table 7 Customers’ corporate image profile by model parameters’ estimates. Segment 1 50.4 percent Segment 2 33.4 percent Segment 3 16.2 percent

Table 8 Latent segment summary profile. Attributes Very satisfied 10 10 10 10 10 10 9, 10 10 10 10 10 10 10 10 9, 10 10 10 10 Well satisfied 9 9 8, 9 9 8, 9 9 8 9 9 9 9 9 9 9 8 9 9 8, 9 Satisfaction demanders 6–8 5–8 4–7 6–8 5–7 5–8 5–7 5–8 6–8 6–8 6–8 6–8 6–8 6–8 6, 7 6–8 6–8 5–7

Relative segment size (ls)

Variables Modernity of procedures and technological supports 6 0.07 0.14 7 0.03 0.16 8 0.25 0.34 9 0.29 0.28 10 0.34 0.04 Credibility and inspired confidence 6 0.00 7 0.00 8 0.07 9 0.30 10 0.63 Orientation to serve the customer 6 0.00 7 0.01 8 0.05 9 0.22 10 0.71 Spreading of given services 5 6 7 8 9 10 0.05 0.08 0.09 0.15 0.15 0.49 0.04 0.04 0.36 0.56 0.00 0.00 0.06 0.34 0.58 0.02 0.02 0.08 0.18 0.32 0.32 0.08

0.21 0.50 0.21 0.00 0.04 0.16 0.37 0.42 0.00 0.05 0.12 0.37 0.37 0.05 0.05 0.12 0.29 0.29 0.29 0.00 0.00

Service quality and demand Service quality perception The ideal public service Global quality of services Reception and welcome Service area Service location Service identification Professional knowledge Information clarity Waiting time Liking, attention and interest Professional reliability Interest in customer problems Modernity of procedures and technological Credibility and inspired confidence Orientation to serve the customer Spreading of given services

Table 9 Customers’ socio-demographic profile by model parameters’ estimates. Segments Relative segment size (ls) Variables Service Military staff Social share Customer kind Reserve Retired Widowers or relatives Pensioners Availability Gender Male Female Education (school attendance) Primary school Basic school Secondary school Higher education Very satisfied 50.4 percent Well satisfied 33.4 percent Satisfaction demanders 16.2 percent

Quite on the contrary, we have segment 3, with 16.2 percent of customers, the satisfaction demanders, thinking that service quality can be improved. In terms of technical quality (Table 5), they classify service quality and demand with 6–8, service quality and expectation with 5–8, the ideal public service with 4–7, and global quality of services with 6–8. Regarding reception and welcome, they classified them from 3 to 7, and services area from 4 to 8; as for services location, these customers classified it from 6 to 8, but they classify services identification from 3 to 8. Thus a few of them considered the performance of the organization negative on some items (the ideal public service, reception and welcome, services area and services identification). Concerning functional quality (Table 6) they classified professional knowledge from 6 to 8, information clarity from 6 to 8, waiting time from 5 to 8, liking, attention and professional interest from 6 to 8, professional reliability from 6 to 8, and interest in customer problems from 6 to 8. So, for everything else they classified as positive. As far as corporate image is concerned (Table 7) they classified modernity of the procedures and technological support from 5 to 7, credibility and inspired confidence from 5 to 8, the orientation to serve the customer from 5 to 7, and the spreading of the given services from 5 to 7. Thus, even for these customers almost everything is slightly positive. Between these two segments we have segment 2, with 33.4 percent of customers, the well satisfied. They are always less demanding customers than those of segment 1, and almost always more demanding customers than those of segment 3, except for services location. These summary results are displayed in Table 8. We have a segment 1, majority female, basic education, very satisfied, and a segment 3, majority male, retired customers, higher education, the most dissatisfied. Higher education tends to develop auto-motivation (Fonseca, 2007), and so they become more and more satisfaction demanders, that is to say, they are more critical persons. So, men seem to be more dissatisfied, as is the case of retired customers, but we think that the same reason

0.33 0.67 0.12 0.49 0.18 0.11 0.11 0.79 0.20 0.07 0.54 0.21 0.18

0.62 0.38 0.18 0.42 0.16 0.04 0.20 0.76 0.22 0.06 0.44 0.28 0.22

0.59 0.42 0.12 0.62 0.04 0.08 0.13 0.87 0.13 0.04 0.33 0.25 0.37

holds: Higher education more exigent customers, and so, more dissatisfaction. In order to achieve a better understanding of customers, in the three segments, we display customers’ socio-demographic profile (Table 9). Thus we have in segment 1 social share service customers, widowers or relatives and pensioners, and basic school customers; in segment 2, customers are mostly military customers, on reserve or on the availability, and also female, and customers with secondary school; finally, in segment 3, we have retired customers, most of all male customers, with higher education. Finally, we have developed an overall satisfaction index (OSI), in order to measure overall customer satisfaction, according to their responses. Let the mean satisfaction rating (MSR) (Table 10) be the mean of all responses to each item that sums 162 for all items. Then, the weighted factor (WF) is the result of MSR dividing by 162. Next, we have weighted scores (WS) that results from the

ARTICLE IN PRESS
358 J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359

multiplication of each MSR by each WF. All the WS sum to the overall weighted average. If overall satisfaction index achieved satisfaction scores of 10 out of 10 on all considered variables, the overall weighted average would be 10, and we express OSI as a percentage of that theoretical maximum score. This index is a very important one, from a managerial/ management perspective, because it shows how the Company as a whole is performing.

6. Conclusion and managerial implications The three main aims of this research were (1) to purpose a simplified model to access customers’ satisfaction based on technical and functional quality, and corporate image, by asking customers if they are completely dissatisfied trough completely satisfied with this, and (2) to purpose and use latent segment models methodology as an approach to the modelling of satisfaction with services, and (3) to develop an overall satisfaction index. These models provide the marketing manager with a very flexible analysis tool, easily understood, and so managers interested in building customers’ satisfaction may seek for a better understanding of customers’ behavioural satisfaction, in order to focus on possible marketing actions to improve or maintain customer satisfaction. After reviewing the existing literature regarding service quality and service customers’ satisfaction, this led to the proposal and an
Table 10 Overall satisfaction index (OSI). Items Mean satisfaction Weighting rating factor 7.73 7.94 8.42 8.50 9.28 9.22 9.20 9.48 9.42 9.27 8.12 8.88 8.92 8.18 9.01 8.85 8.82 9.13 162 4.77 4.90 5.20 5.25 5.73 5.69 5.68 5.85 5.82 5.72 5.01 5.48 5.50 5.05 5.56 5.46 5.44 5.64 Weighted score 0.37 0.39 0.44 0.45 0.53 0.52 0.52 0.55 0.55 0.53 0.41 0.49 0.49 0.41 0.50 0.48 0.48 0.51 8.71 87%

Service location Service identification Reception and welcome area Service area Professional knowledge Information clarity Waiting time Liking, attention and interest Professional reliability Interest in customer problems Modernity of procedures and technological supports Credibility and inspired confidence Orientation to serve the customer Spreading of given services Service quality and demand Service quality and expectation The ideal public service Global quality of services Overall weighted average Overall satisfaction index

application with LSM approach. The finding results help on developing a picture of the customers of this Public Service, regarding satisfaction, that is to say, the typology with three classes. Moreover, the LSM, based on observed categorical variables, enables us to analyse organizational quality and customer satisfaction, which is an important theoretical and practical issue for organization managers. Thus, in a managerial sense, those results are actually quite opportune, because organization managers became knowledgeable of how customers’ satisfaction goes on, and therefore results obtained from this study can result in strategic planning strategies enhancing customers’ satisfaction. It was learned that service customers perceive high satisfaction. Most customers surveyed (50.4 percent, the majority) appear to be very satisfied with service quality and corporate image. The analysis allows us to conclude that about 83.8 percent (segments 1 and 2) of customers surveyed revealed highly overall satisfaction. The findings reported here suggest that overall customer satisfaction is real and so the service quality is very good for the majority of customers. It is supported by the latent segments, as discussed, and moreover, by the OSI value of 87 percent (Table 10). This study also focus on identifying socio-demographic factors underlying the different reinforcing behaviours across customer segments, providing managers with even more information. From the socio-demographic profile of customers we can see that gender is not relevant for overall satisfaction and the same happens for customers with average education. Managers must take into consideration segment 2 (33.4 percent) customers, and above all segment 3 (16.2 percent) customers, because they are not very satisfied, or they are unsatisfied, respectively. Finally, because the organization core business is the staff, we correlate technical quality with functional quality (see Table 11); the significant Spearman correlations (p-valueo0.01) highlight the human resources (staff) as the service face, as it should be in such an organization. Thus, as Agus (2004), we also think that managers should emphasize the importance of teamwork in achieving the service goals, that is to say, every worker should be empowered to act on worthwhile suggestions that will ultimately improve customer satisfaction.

7. Limitations and future work We think that this Public Service can be representative of the Portuguese Public Sector, and so the conclusions may be valid for other Portuguese Public Services. Moreover, these models and results can be generalized across other countries and other types of public/private services; to access customers’ satisfaction we can use more or less variables, and these models work well with only categorical variables, with only continuous variables, and with both categorical and continuous variables (mixed case).

Table 11 Technical versus functional Spearman correlations. Professional knowledge Service quality and demand Service quality and expectation The ideal public service Global quality of services 0.68ÃÃ 0.68ÃÃ 0.65ÃÃ 0.68ÃÃ Information clarity 0.70ÃÃ 0.67ÃÃ 0.67ÃÃ 0.68ÃÃ Waiting time 0.57ÃÃ 0.58ÃÃ 0.51ÃÃ 0.51ÃÃ Liking, attention and interest 0.62ÃÃ 0.55ÃÃ 0.52ÃÃ 0.63ÃÃ Professional reliability 0.64ÃÃ 0.58ÃÃ 0.61ÃÃ 0.61ÃÃ Interest on customer problems 0.69ÃÃ 0.64ÃÃ 0.63ÃÃ 0.68ÃÃ

ÃÃ Correlation is significant at the 0.01 level (two-tailed).

ARTICLE IN PRESS
J.R.S. Fonseca / Journal of Retailing and Consumer Services 16 (2009) 352–359 359

It would be important that we could have another survey in the next 2 or 3 years, based on the same questionnaire, in order to study segment stability over time (Fonseca and Cardoso, 2007b).

Acknowledgements The author wish to thank the two anonymous referees, for their many valuable suggestions, which led to a significant improvement of the article. References
Agus, A., 2004. TQM as a focus for improving overall service performance and customer satisfaction: an empirical study on a public service sector in Malaysia. Total Quality Management 15, 615–628. Anderson, E.W., Fornell, C., 2000. Foundations of the American customer satisfaction index. Total Quality Management 11, S869–S882. Bartikowski, B., Llosa, S., 2004. Customer satisfaction measurement: comparing four methods of attribute categorizations. The Service Industries Journal 24, 67–72. Benyon, M., Goode, M.M.H., Moutinho, L.A., Snee, H.R., 2005. Modelling satisfaction automated banking channels: using variable precision rough set theory. Services Marketing Quarterly 26, 77–93. Bhatnagar, A., Ghose, S., 2004. A latent class segmentation analysis of e-shoppers. Journal of Business Research 57, 758–767. Bodet, G., 2006. Investigating customer satisfaction in a health club context by an application of the tetraclass model. European Sport Management Quarterly 6, 149–165. Bonilla, J., Huntington, J., 2005. Leveraging latent class segmentation to optimize marketing campaigns and sales force implementation. 2005 PMSA Conference. Bozdogan, H., 1987. Model selection and Akaikes’s Information Criterion (AIC): the general theory and its analytical extensions. Psychometrika 52, 345–370. Bozdogan, H., 1994. Mixture-model cluster analysis using model selection criteria and a new informational measure of complexity. In: Bozdogan, H. (Ed.), Proceedings of the First US/Japan Conference on the Frontiers of Statistical Modeling: An Approach. Kluwer Academic Publishers, Dordrecht, pp. 69–113. ´ Caro, L.M., Garcıa, J.A.M., 2007. Measuring perceived service quality in urgent transport service. Journal of Retailing and Consumer Services 14, 60–72. Churchill Jr., G.A., 1979. A paradigm for developing better measures of marketing constructs. Journal of Marketing Research 14, 64–73. Cohen, S.H., Ramaswamy, V., 1998. New tools for market segmentation: an introduction to Latent Class Models. Marketing Research 10, 14–21. Collier, J.E., Bienstock, C.C., 2006. Measuring service quality in E-retailing. Journal of Service Research 8, 260–275. Dabholkar, P.A., 2000. A comprehensive framework for service quality: an investigation of critical conceptual and measurement issues through a longitudinal study. Journal of Retailing 76, 139–173. Dempster, A.P., Laird, N.M., Rubin, D.B., 1977. Maximum likelihood from incomplete data via EM algorithm. Journal of the Royal Statistics Society B 39, 1–38. Dillon, W.R., Kumar, A., 1994. Latent structure and other mixture models in marketing: an integrative survey and overview. In: Bagozi, R.P. (Ed.), Advanced Methods of Marketing Research. Blackwell Publishers, Cambridge, pp. 352–388 (Chapter 9).

Figueiredo, M.A.T., Jain, A.K., 2002. Unsupervised learning of finite mixture models. IEEE Transactions on Pattern Analysis and Machine Intelligence 24, 1–16. Fonseca, J.R.S., Cardoso, M.G.M.S., 2005. Retail clients latent segments in ´ ¨ Carlos Bento proceedings. In: Amılcar Cardoso, Gael Dias (Ed.), 12th ˜ Portuguese Conference on Artificial Intelligence, EPIA 2005. Springer, Covilha, Portugal. Fonseca, J.R.S., Cardoso, M.G.M.S., 2007a. Mixture-model cluster analysis using information theoretical criteria. Intelligent Data Analysis 11, 155–173. Fonseca, J.R.S., Cardoso, M.G.M.S., 2007b. Supermarket customers segments stability. Journal of Targeting, Measurement and Analysis for Marketing 15, 210–221. Fonseca, J.R.S., 2007. Latent Class Model for discovering the pattern of the Portuguese electorate feelings and their sight after 2005 elections. In: Lauro, C., Ferreira, C., Saporta, G., Souto de Miranda, M. (Eds.), IASC 07-Statistics for Data Mining, Learning and Knowledge Extraction, August 30–September 1, 2007, Aveiro, Portugal. Garbarino, E., Johnson, M.S., 2001. Effects of consumer goals on attribute weighting, overall satisfaction, and product usage. Psychology & Marketing 18, 929–949. Goode, M.M.H., 2001. Predicting consumer satisfaction from CD players. Journal of Consumer Behaviour 1, 323–335. Gronroos, C., 1984. A service quality model and its marketing implications. European Journal of Marketing 2, 244–249. Harris, L.C., Goode, M.M.H., 2004. The four levels of loyalty and the pivotal role of trust: a study of online service dynamics. Journal of Retailing 80, 139–158. Lazarsfield, P.F., Henry, N.W., 1968. Latent Structure Analysis. Houghton Mifflin, Boston. Magidson, Jay, Vermunt, J.K., 2003. A Nontechnical Introduction to Latent Class Models. Tilburg University, Statistical Innovations, Inc., Nederlands. Marsh, H.W., Hocevar, D., 1985. Applications of confirmatory factor analysis to the study of self concept: first-and higher order factor models and their invariance across groups. Psychological Bulletin 97, 562–582. McLachlan, G.F., Peel, D., 2000. Finite Mixture Models. Wiley, New York. Oliver, R.L., 1997. Satisfaction a Behavioural Perspective on the Consumer. McGrawHill, New York. Oliver, R.L., 1993. A conceptual model of service quality and service satisfaction: compatible goals, different concepts. In: Swartz, T.A., Bowen, D.E., Brown, S.W. (Eds.), Advances in Marketing and Management. JAI Press, Inc., Greenwich, CT, pp. 65–85. Parasuraman, A., Zeithaml, V.A., Berry, L.L., 1985. A conceptual model of service quality and its implications for future research. Journal of Marketing 49, 41–50. Ramaswamy, V., Chaterjee, R., Cohen, S.H., 1996. Joint segmentation on distinct interdependent bases with categorical data. Journal of Marketing Research XXXIII, 337–350. Spreng, R.A., Macoy, R.D., 1996. An empirical examination of a model of perceived service quality and satisfaction. Journal of Retailing 72, 201–214. Truch, E., 2006. Lean consumption and its influence on brand. Journal of Consumer Behaviour 5, 157–165. Vermunt, J.K., Magidson, J., 2002. Latent class cluster analysis. In: Hagenaars, J.A., McCutcheon, A.L. (Eds.), Applied Latent Class Analysis. Cambridge University Press, Cambridge, pp. 89–106. Wedel, M., Kamakura, W.A., 1998. Market Segmentation: Concepts and Methodological Foundations. Kluwer Academic Publishers, Boston. Wu, J., DeSarbo, W.S., 2005. Market segmentation for customer satisfaction studies via a new latent structure multidimensional scaling model. Applied Stochastic Models in Business and Industry 21, 303–309.