Study Of Philosophical Assumption Analysis

Published Date: 23 Mar 2015

This chapter depicts reflexivity - the way of carrying research and figuring results -, and addresses philosophical reflexivity, methodological reflexivity, and disciplinary reflexivity (Hardy, 2001; Holland, 1999). Reflexivity particularly illustrates the philosophical assumptions, and the executive steps toward the objectives. Thereby, the chapter discusses how the tentative model is refined according qualitative evidences and validated by quantitative supports (Bryman & Bell, 2007, p. 648; Churchill, 1979; Eisenhardt, 1989; Jick, 1979). Figure below briefly illustrates this chapter content.

This section addresses the ontological and epistemological philosophy behind the study. It is necessary to crystallize the philosophical assumptions as the foundation of academic research, which is required to choose applicable methods (Annells, 1996). Given paradigm question, research topic associates with a particular meaning in specific circumstances (Goulding, 2002, p. 36). CEM resembles a complex management approach based on marketing philosophy that utilizes some components such as dealing with brand, interface, employee, and relationship to delight customers.

Schembri (2006) assume realism to construct the meaning of services experience in ontological perspective. In this ontological position, understanding the customer's service experience needs to reflect the customer as the subject that is conjoined with services as the object (Schembri, 2006). Likewise, services experience must be regarded holistically rather than only as performance, process, or outcome. Prahalad and Ramaswamy (2000; 2004) similarly believe all sources of experience are objects that confirm realism in ontology. On contrary, the earlier implicit underlying assumptions indicate a rationalistic philosophy (Vargo & Lusch, 2004). Identification of customer desires in a rationalistic logic keeps us away from a holistic understanding of the customer's experiential meaning (Schembri, 2006). Schembri (2006) argued the rationalistic assumptions are limiting the advance of marketing theory since experience ontologically is a relation between customer and world.

Given epistemology, understanding of human nature in social worlds is acquired by conducting research to capture and interpret the complex ever changing social world. Here, the research methods are applied to interpret social world in terms of meaningful model. Epistemology in connection with the research objectives leads study to get realistic but interpretive understanding about the management role in orchestrating customer experience regarding internal process and proper response. It means subjective factors (experiential modules) should be concerned in epistemological reflexivity (Hirschman & Holbrook, 1982). It thus intensifies the subjectivism in epistemology. Generally, there are three schools of thought based upon different combinations of epistemological and ontological assumptions (Alvesson & Deetz, 2000; pp. 60-74; Johnson & Duberley, 2003):

Neo-empiricism (Thesis) (Epistemic: Objectivism - Ontological: Realism)

Critical theory (Synthesis) (Epistemic: Subjectivism - Ontological: Realism)

Postmodernism (Antithesis) (Epistemic: Subjectivism - Ontological: Subjectivism)

Epistemological subjectivists are researchers who assume it is not possible to observe the behavior of social phenomena neutrally without subjective interpretation of perception (Economic & Social Research Council, 2008). Realism is about the ontological status of the phenomena that assumed to constitute social reality (Economic & Social Research Council, 2008). In the qualitative study, we assume realistic postulation about the ontological status to constitute social reality that they exist independently of our perceptual or cognitive structures (Goulding, 2002, p. 13). Simply, we might not already know all aspects of the phenomenon but we know this reality exists and it can be discovered. Critical theory thus can be the best nominee to be a philosophical underpinning. Critical management study depends on researcher because of subjectivism in epistemology. However, similar to neo-empiricism view in ontology, there is a chance of unbiased and objective data collection and analysis (Alvesson & Deetz, 2000), since social construction lead us to the reality (Myers, 1997, p. 241).

Interpretive epistemology (Subjectivism) is a phenomenological, hermeneutic approach to reveal meaningful structure by stepping in social phenomenon within the context (Boland & Day, 1989). However, it requires observation in natural setting to gain apt understanding (Anderson, 1986; Neuman 1997, p. 68). In addition, interpretive epistemology underlined following issues:

Revealing the complexity by gaining in-depth understanding around the phenomenon

Reflecting the reality based on close participation and high sensitivity to all elements

Capturing the subjective experience of each individual

Addressing the subjectivity by, for example, openness to alternative explanation

Interpretive methods are utilized for exploring and explaining since (Desphande, 1983):

Previous research may not obtain complete understanding

The positivist approach is not proper to cover CEM as a new multi-facets discipline

Lack of understanding necessitates the research of a phenomenological nature

Quantitative approach is not appropriate since cannot figure out inclusive picture

CEM is composed of a series of multiple realities that entirely should be taken into account

In the quantitative phase, testing the model is necessary as far as it is a comparison between reality and the recommended model. Data is rendered as an objective external reality to come up with truth and therefore, it might be perceived as positivism (Charmaz, 2000). Positivism is based on three facts: logicality (mathematical facts; Smith, 1983), objectivity, detached manner (Burrell & Morgan, 1979, p. 21). Epistemology relies on searching regularities and casual relationship between constituent elements of research to test theory and increase prediction power (Neuman, 1997, p. 63). It is also necessary to turn the phenomenon into generalized measurable fact (McGrath. 2008). This approach is acceptable only in case of success. That means, if we fail to verify the model, it is not appropriate and needs to revise (Hindess, 1977, p. 18). Therefore, epistemology can be interpretive in qualitative phase and positivism in quantitative stage (Yin, 1994).

Methodological Consideration

Nancy and Bradley (1999) hinted "theory building begins with a general research question in stage one and ends with a theory and hypotheses that will be tested in more controlled studies in the future". Regarding philosophical assumption, hybrid methodology can be proper for the model building and given objectives. It is appropriate since it simultaneously concentrates on theory building and theory verification. Next section broadly brings some evidences to support this decision.

Why Hybrid Methodology

Academicians portray CEM as an imperative approach for modern marketing, which leads to business success. Despite the importance, the concept has begun to flourish without any discipline (Bitner et al., 2008; Holbrook, 2007; Verhoef et al., 2009) because of insufficient formal academic investigations, lack of empirical direction and theoretical foundations (Caru & Cova 2003; Gentile et al., 2007; Poulsson & Kale, 2004; Pullman & Gross, 2004). Besides, CEM has a complex setting due to impacts of many different factors. In such complex and undeveloped subject, Nachmias and Nachmias (1981) noted multiple data collection techniques are preferable to reliance on only one.

Morgan (1998) declared, "Using either qualitative or quantitative methods in isolation can easily lead to mistaken conclusion". That means the combination of inductive and deductive methods is necessary to enhance the confidence in the results (Eisenhardt, 1989; Jick, 1979). The variety of qualitative methods, in addition to quantities study, reasonably helps to figure out research objectives in more natural setting with comprehensive depiction (Furlong, Lovelace, & Lovelace, 2000, pp. 543-544). Eisenhardt (1989) posed, "grand theory requires multiple studies an accumulation of both theory building and theory testing". Moreover, new categories (not initially anticipated) can arise because of creative potential of mixed method (Eisenhardt, 1989). This methodological approach refers to hybrid methodology that suggests combining qualitative and quantitative methods in logical order or combination. We benefit the potency of one method to improve the fulfillment of the next method (Morgan, 1998). There are some others motivations for using multiple methods (Jick, 1979):

To help to uncover the deviant dimension of a phenomenon

To facilitate refashioning theory and model by looking to different viewpoints

To enrich explanation of the research problem

To ease a synthesis and integration of theories

To serve as the critical test, by virtue of its comprehensiveness, for competing theories

Given convolution of subject matter, it seems hybrid methodology is much more practical than other approaches to build a novel construct and figure out comprehensible model. Likewise, successful studies with similar approach in the context of study (e.g. Pullman & Gross, 2004; Brakus et al., 2009) or in other discipline (e.g. Liao, Murphy & Welsch, 2005; Wilson & Vlosky, 1997) can be references for reasonableness of adopted methodology. Combing the different dataset can be also synergistic (Pandit, 1996). Eisenhardt (1989) asserted, "When a pattern from one data source is corroborated by the evidence from another, the finding is stronger and better grounded". The use of multiple data enhances construct validity (Pandit, 1996), when the results are validated on the same question by multiple methods (Carson, Gilmore, Perry, & Gronhaug, 2001, p. 69; Morgan, 1998).

Developing the Construct

The importance of construct development as the key part of theoretical explanation is underlined in marketing literature (Churchill, 1979; Diamantopoulos & Winklhofer, 2001; Jarvis, Mackenzie, & Podsakoff, 2003; Peter, 1981). There is emphasis on introducing precise and measurable constructs as a basis for strong theory, which is facilitated by a prior pattern in the opening step (Eisenhardt, 1989). Thus, to discover CEM construct, we initially mull over several theoretical and practical evidences earlier than qualitative investigation to crystallize a former tentative construct.

Developing construct can be reflective, formative or both (Jarvis et al., 2003). In reflective model, measures should be correlated, while dropping an indicator from measurement model does not change the whole construct (Jarvis et al., 2003). In contrast, in formative model, the direction of causality is from measure to construct (Jarvis et al., 2003). Developing multidimensional construct for CEM involves two levels; first order is reflective and second order is formative (Type III in Jarvis et al.'s study). CEM is assumed reflective construct consists of reflective components, namely brand experience and interface experience, which are composed of multiple formative indicators.

Research Setting

Research Population

Current study focuses on private service sector. There are three main reasons behind this resolution. In the first place, scholars put extra emphasis on practicality of CEM in service context (see Error: Reference source not found). Secondly, there is tremendous flourish in service industry in Malaysia especially in private sector while it has significant contribution in economy (see Error: Reference source not found). Finally, service sector in Asia have faced lot of difficulties during these years, particularly in dealing with increasing expectation, changing lifestyle, tide competition, and commoditization issues (Lovelock et al., 2002).

On the other hand, service sector furnishes us with apposite data to analyze and draw a conclusion; since, customer experience is more meaningful in service context due to following reasons. First, service experience includes more interactions and accordingly provides more opportunities for customers to experience and become engaged. Second, service experience is more flexible than goods experience because of high customer involvement; hence, the brand experience can be more likely to remember. Third, the numerous touchpoints enrich experience in service context. Fourth, experience is more likely to differentiate the service offering. Finally, in service, co-creating experience, social and relational experience, and personalized experience are more achievable than goods experience. These evidences enrich the result of investigation and subsequently enhance the comprehensiveness.

Unit of Analysis

Unit of analysis usually refers to the level of aggregation of the data during conducting analysis. In the present study, organization level as the lowest independent level is chosen as unit of analysis. Although, experience is investigated in individual level, CEM is practiced by organization, derived by organizational capabilities, and it enhances organization outcomes. This option in line with Kenny (2003) has two characteristic: appropriate level of analysis and independence. The unit of analysis also keeps results away from officious impacts such as compositional effects (Kenny & Judd, 1986).

Hybrid Research Design (mixed methods strategy)

The initial step of research design is the 'division of labor' to integrate the complementary strengths of methods (Creswell, 2003, p. 211), which is accomplished via two primary choices (Morgan, 1998):

Priority: which method is a principal method and which is a complementary one

Sequence: whether the complementary method precedes or follows the principal method

In hybrid methodology, prioritizing and sequencing of qualitative and quantitative methods are crucial to conduct a successful research (Morgan, 1998; Tashakkori & Teddlie, 2008, p. 161). By combining these two, four options are yield: (1) Preliminary qualitative methods in quantities study, (2) Preliminary quantitative methods in qualitative study, (3) Follow-up qualitative methods in quantitative study, and (4) Follow-up quantitative methods in qualitative study.

The Priority Decision Morgan (1998) illuminated "The first research-design decision determines the extent to which either the qualitative or the quantitative method will be the principal tool for gathering the project's data". Conducting research leads to the high range of difficulty by exploiting both quantitative and qualitative methods with same weight or in same time (Morgan, 1998). In traditional triangulation, the results gained from the two methods because of complicated linkage and no effective instruments may be either incommensurate or absolute conflicting (Morgan, 1998). Therefore, the more practical approach is to assign one of the methods as the principal tool and then to allocate the second one as the complementary to effectively assists the principal one (Morgan, 1998; Tashakkori & Teddlie, 2008, p. 161). This division of labor is more realistic and convenient.

The Sequence Decision Morgan (1998) stated, "The more practical strategy is to use the two methods in sequence so that what is learned from one adds to what is learned from the other". That means which method based on its strengths, or probable results should be used first (Creswell, 2003, p. 212). This approach optimizes the effectiveness of the principal method (Morgan, 1998; Tashakkori & Teddlie, 2008, p. 164). Hence, it is valid to use preliminary output to improve the next method. It maximizes the worth of results, when it upgrades the next method, which uses a different data set.

The Priority and Sequence in the current Research

The preferable hybrid approach can be qualitative investigation that follows by quantitative study. This pattern will be implement by preliminary qualitative study provides complementary assistance in developing a larger quantitative study. Scholars recommended the exploratory research as the prior approach for the majority of descriptive studies to build a theoretical model (Churchill & Iacobucci, 2002, p. 92; Creswell, 2003, p. 215; Yin, 1994, p. 138). On the word of Kotler (2006, p. 122), "The objective of exploratory research is to gather preliminary information that will help define problems and suggest hypotheses". Along with the objectives, exploratory study in early stage expand understanding, provide insight, and develop hypotheses until it lets us to continue the study by testing the hypotheses and validating the construct in a deductive study (Churchill & Iacobucci, 2002, p. 93).

This approach is against traditional triangulation as comparison between different methods, researcher, theory, or results. Priority and sequence decision emphasizes on complementarily characteristic of each method as well as flexibility, manageability, productivity, and lack of difficulty (Morgan, 1998). The data collection begins with a qualitative method to crystallize the definition, and domain, improve the model, and develop the content of the questionnaire. The qualitative results must be treated as tentative until they are confirmed by quantitative research (Morse, 1996). The strengths of qualitative methods are utilized during exploratory study to conduct more fruitfulness quantitative research later. Further, we make use of the quantitative data to verify and expand on what is learned through the qualitative study. Then the quantitative inquiry covers a much larger sample than in-depth qualitative research to explore the generalizability, and transferability of the results (Morgan, 1998).

Why Qualitative Approach is Chosen in the First Phase?

The most common design in hybrid methodology is preliminary qualitative studies plus complementary quantitative research (Morgan, 1998). Qualitative research has been arisen in response to limitations in conventional quantitative management research. Scholars recommend the qualitative study in the initial stage of research to overcome insufficient understanding around unfamiliar phenomenon (Churchill, 1991, p. 132). For example, qualitative study is recommended in service and social experience context (Otto & Ritchie, 1996; Verhoef et al., 2009). Morse (1991, p. 120) also highlighted some characteristics of subject matter in qualitative research, which have close agreement with our research objectives. Regarding that, inductive method can be applied when:

The concept is immature due to lack of theory or theoretical researches

The available theory may be inaccurate, incorrect, inappropriate, or even biased

There is need to explore and describe the phenomena or develop the theory

The nature of the phenomena may not to be suited to quantitative investigation

It seems qualitative approach can be the best nominee to overcome immaturity in CEM concept and generate formal theoretical understanding. Moreover, because of inconsistency in the subject matter, it will be practical to use inductive methodology to explore the phenomenon and integrate the various options into more comprehensive definition. Flexibility is another reason behind choosing preliminary qualitative approach (Churchill, 1991, p. 132), since we can take advantage of data to come up with novel themes and restructured the model (Eisenhardt, 1989).

Researcher Function and Time Horizon

Researcher has active role in data gathering, data coding and interpreting. In general, it is tried to be skilful in collecting and summarizing data in appropriate manner and self-critical accounts. Moreover, to have accurate data, we try to conduct research in bias-free situation and have precise estimation. In qualitative stage, the researcher's role as an instrument for data collection is yielding maximum knowledge to accomplish holistic understanding in natural setting (Creswell, 2003, p. 200). Therefore, it requires direct involvement not only in analysis but also in all data collection steps. Given time horizon, the study will be conducted as cross-sectional study and data are gathered just once, over a period of three months in first phase and another three months in second phase.

Qualitative Study

Success of hybrid methodology highly depends on qualitative phase (Jick, 1979). Qualitative data contributes as the critical counterpoint to quantitative methods, which drawn from firsthand records to characterize the phenomenon, crystallize holistic picture of research topic and report relevant details. The upshots ameliorate the initial propositions to hypotheses (Carson et al., 2001, p. 42). According Creswell (2003, pp. 198-199), we consider following reflection in conducting qualitative research. Firstly, we more emphasize on the process, rather than outcomes. Secondly, researcher works as an instrument for data collection and analysis to reach holistic understanding. Thirdly, qualitative research needs fieldwork and direct involvement. Fourthly, we have to be descriptive with a sense of gaining deep understanding. Finally, the inductive process has to lead the research to build concept, tentative hypotheses, and model from details in realistic way.

The qualitative study, based on exploratory research design, intends to investigate the phenomenon, and provide the rigorous insights (Churchill, 1991, p. 130; Yin, 1994). Churchill (1991, p. 130) recommend exploratory research to discover vague problem and discover the possibilities of conjectural statement that here is called tentative model. To build a proper foundation for quantitative study, based on Selltiz, Wrightsman, and Cook (1981, p. 21) and Yin (1994), we justify the applicability of exploratory case study (analysis of selected cases) in the initial phase of research:

It is a proper way to answer research question, study contemporary event, and investigate a phenomenon prior to theory development;

The concept novelty necessitates submitting some propositions before developing hypotheses;

In respect of the various possible variables for CEM construct, antecedents and consequences, we require setting priorities to purpose the model;

Given limitation in literature, we crucially should gain enough insight into the problem;

Exploratory study helps to capture the domain and generate the right items (measurement)

Our approach in qualitative research is in line with exploratory study, which is based on 'Experience Survey' and 'Analysis of Selected Cases' (Churchill & Iacobucci, 2002, p. 95). Churchill (1991, p. 135) defined experience survey as the key informant survey, which figures out valuable insight through who has association whit a particular marketing effort. The target of this approach is coming up with relationship and overall variables picture or tentative explanation (Selltiz et al., 1981, p. 94). Analysis of selected case is an approach to explore for explanation or features that can be common with other cases (Churchill, 1991, p. 143). Thereby, we start qualitative study by choosing convenient cases, which are potentially capable to run CEM (see, 5.4.2). Then, we collect data about the construct and domain, examine the tentative model, and explore the alternative characterization via experienced interviewees. Qualitative data analysis package NVivo 8 will be used to enable effective data management. The qualitative study procedure is summarized in Figure ‎5 -2.

Figure ‎5‑2: Qualitative Study Procedure

Qualitative Research Design and Data Collection

Interview can be a practical instrument; since, it results sensible inference through more controllable procedure (Yin, 1994, p. 80). Interview is also the foremost source of data collecting toward theory building (Lillis, 1999). In our case, in-depth interview with non-probability sampling is an applicable technique (Churchill, 1991, p. 135). Interview is carried out face-to-face (one-to-one conversations) in the interviewees' self-setting (Denzin & Lincoln, 2005, p. 661; Morgan et al., 1990). To make best use of in-depth interview, we design semi-structured questionnaire with following characteristics.

According to Furlong et al. (2000, p. 536), semi-structured interview is a proper technique while we have prior idea about questions. Semi-structured questionnaire thus can extract maximum relevant information in specific time because of existing tentative pattern (Yin, 1994, p. 85). It also aids to examine the authenticity of tentative model and prior understanding about propositions. Additionally, to know what is beyond the tentative mode, some open-ended questions empower the interview to have necessary flexibility for further exploration (Furlong et al., 2000, p. 536). The significant advantageous of this method are the modification and completeness (Silverman, 2000). It also systematically saves time and offers specific elaboration in text production and analysis.

The standard theme of interview is adopted from Kohli and Jaworski (1990). After a brief description, interviews start open-ended and finalize with more relevant probe questions (Dick, 2005). Thus, there are two kinds of questions. First, the opening questions investigate whether the concept elements are significant and to what extent they are meaningful and practical (Sekaran, 2006, p. 236). Then, interview follows by semi-structured questions to examine, and validate the propositions in details. The respondents are also asked to tell their own stories and experiences (Yin, 1994, p. 84). This kind of information associates with informant's experience (Dick, 2005) and leads to find out memories, meanings, and interpretations might be hard to discover in other ways (Carson et al., 2001, p. 73). It is also predictable that after or before some questions a brief explanation is necessary to clarify the intention of questions (Kohli & Jaworski, 1990). Interviews are conducted separately and they last 60-120 minutes. The Interview Protocol - Appendix 4 - is set up to have a systematic procedure.

Due to get access to useful information, it is essential to render the right image and convince interviewees the study is non-threatening. This issue can be handled by give the overall summary of the research procedures and objectives in order to minimize the risk of 'defensive or self-conscious behavior' (Waddington, 1994). Additionally, offering some motivations can be practical to encourage respondents for more commitment and collaboration. We consider providing a short report to the participating firms as prospective motivation. It supposes to let them to benchmark themselves within the industry. Keeping with ethical issues, we utilize direct approach and disclose the purpose of study.

Choosing Appropriate Cases in Qualitative Stage

In private service, consumer banking is nominated as the sample frame for qualitative study. In the retailing of financial services, customer behavior is characterized by full range of possible experience, which leads to achieve results that are more comprehensive. The focus on personal banking is also reasonable to analyze because of various possible interactions and accordingly various meaningful experiences. Moreover, specific characteristics of consumer banking and Malaysian banking perspective are likely to generate more holistic result (see, Error: Reference source not found).

The study is set along with a similar research in financial service (e.g. O'Loughlin & Szmiginthe, 2005) with multiple cases (Yin, 1994, p. 45). It is based on conducting a series of interviews with key informants in banking (Gummesson, 2000, p. 179). Non-probability sampling is preferable, because it is convenient to meet sampling goals. The samples are chosen through purposive and snowball sampling (selective sampling; Patton, 1990; Sandelowski, Holditch-Davis, & Harris, 1992, p. 279) as one kind of theoretical sampling (Eisenhardt, 1989). Purposive sampling is nominated for choosing banks (Sekaran, 2006, p. 277; Denzin & Lincoln, 2005, p. 378) and snowball sampling for finding branch managers as representatives of the banking (O'loughin & Szmigin, 2005). This approach intensifies the diversity of sampling to find different properties (Charles, 1994, p. 99; Dick, 2005).

At least three or four sets of data are essential to come across the concept (Martin & Turner, 1986). Accordingly, we plan to conduct 12 interviews in six data sets. Five of these sets are part of initial data collection attempt and the sixth one for cross-validation. This is in line with Perry (1998) and Eisenhardt (1989) who respectively suggested four to fifteen and four to ten cases for interview. We project theoretical saturation by five data sets, albeit the openness to increase cases. If there is any contrast between two respondents in a data set, we add another interview session to enhance validity.

We nominate five local banks from Bank Negara list as our sampling frame (see, Appendix 6) as well as one foreigner bank for cross-validation. According a study by Goh in 2005, Maybank is the largest commercial bank in Malaysia in terms of asset, net profit, shareholder equity, and number of employees while CIMB holds the second place. Public Bank is the second rank in terms of profit and shareholder equity. RHB bank and Hong Leong Bank also have notable asset as well as considerable development in customer management. Additionally, HSBC is the largest foreigner bank regarding asset, profit, shareholder equity, and number of employee (Goh, 2005). In light of these information Maybank, CIMB, Public Bank, RHB Bank, and Hong Leong Bank are chosen as the first set of data source and HSBC is selected for cross validation. HSBC is leading banks in CEM; thus, verifying the qualitative result with HSBC would be reasonable decision (Lee, 2008; Wing, 2008).

The interviews will be conducted with the managers of major branches, since they have close interaction with customers, they have extensive familiarity with consumer banking, and they are well updated with bank marketing strategy regarding consumer banking [i] . In this case, snowball technique is utilized to facilitate sampling. The technique suggests the initial respondents recommend other potential interviewees who can contribute to the study (Bryman & Bell, 2007, p. 200). Recommending conversant is applicable in financial services (O'loughin & Szmigin, 2005). However, to refine the process, a purposive approach prioritizes the recommendations (Schmidt & Little, 2007). Subsequently, by saturation we move to sorting and coding in the clearest way.

Data Capturing, and Coding in Qualitative Stage

During qualitative research, we anticipate mass of detailed material. Ensuring about the capturing and recording relevant data would allow us for easy retrieval in analysis. Miles and Huberman (1984) expressed initial coding is compulsory to linking data and weave a story. In order to have systematic coding, some tools from Grounded Theory are utilized. In spite of GT, we do not begin with field data and allow the theory to emerge from the data; instead, we begin with tentative model from previous studies. That means core category does not emerge from data set. The role of literature review is also unlike GT, especially in developing categories. Instead, we follow preferable method by Miles and Huberman (1994) to set provisional codes, namely start list, which is based on the tentative model.

To reduce the risk of observer bias, initial data capturing is fulfilled by field remarks through audiotaping plus simple note taking (Dick, 2005). The main aim of coding data is to determine possible categories toward a reasonable theme and variables (Dick, 2005). We start with Open Coding to transfer raw data to appropriate structure. It brackets raw qualitative data into incidents data file (Van de Ven & Poole, 1990) and helps to narrow down the transcripts by labeling and categorizing (Pandit, 1996). As a pre-analytic process, it is the underpinning to identify the concepts, their properties, and dimensions (Strauss & Corbin, 1990, p. 102). Data are firstly broken down and then compared. To categorize, comparable evidences are clustered together under the same conceptual label and lead to the more abstract level to build categories (Pandit, 1996). We employ single coding in contract with multiple coding since the interviews are primarily one-on-one. Then, data is refined in the context of the categories and linked (Dey, 1993, p. 29). It leads to more precisely definition in terms of categories with minor adjustments. Then, selective coding formulates the connections and relationship between a variable and its properties as well as category and its sub-categories.

Data Analysis in Qualitative Stage and Model Developing

The analysis is based on theoretical propositions that systematically guide the process (Yin, 1994, pp. 103-104). Analysis of qualitative data involves data reduction, data monitoring, conclusion drawing, and verification. After initial stage of analyzing during data coding, we turns to searching for patterns, discovering what is important and what is to be learned, and deciding what to tell others (Schurink, 2004, p. 20). Similarly, Patton (2002, p. 432) recommended constructing the model after reducing the volume of raw information, sifting trivia from significance, and identifying significant themes.

Qualitative data analysis is inductive reasoning and theorizing to address initial propositions (Yin, 1994). Data analyzing requires apt judgment and creativity to transform data into theory by going beyond description. Taylor and Bogdan (1998) specified each researcher could follow personal way in data analysis. Therefore, the results of qualitative analysis usually are not exclusive because it ultimately depends on the analytical intellect and style of the analyst (Patton, 2002, p. 433). However, we try to make it reasonably systematic to follow standard routine in analyzing qualitative data such as data reduction and coding, identification of themes and developing concept (Denzin & Lincoln, 2005, p. 453; Taylor & Bogdan, 1998). We start analyzing during data collecting, and coding (e.g., matrix of categories, Yin, 1994, p. 103). That means, analysis is not simply one of the later stages of research but it is a pervasive activity throughout the research (Coffey & Atkinson, 1996, p. 11).

Microanalysis of the data and progressive refocusing begin subsequent to preliminary coding to establish some categories. Linking codes creates subcategories and tying them to the properties and dimensions. We draw distinction or parallels between coded result and tentative subcategories. Developing categories into general analytical model with relevance outside the setting can be the next step. Next attempt includes the process of integrating and refining categories to form a larger theoretical scheme. Analysis looks for integration of variables and categories to formulate the primary model (Pandit, 1996). The major categories are synthesized into a theoretical scheme for more analysis. In this case, comparison is suggested (Dick, 2005). It means draw an analogy between the emerged patterns quickly leads to theory. Dick (2005) recommended constant comparison, involves: comparing data set toward data set and comparing data set to the tentative pattern. Relationships between variables begin to emerge by the result of comparison (Eisenhardt, 1989).

Analytic strategy follows 'pattern matching' approach and looks for theoretical replication for stronger conclusion (Yin, 1994, p. 106). The results comparison of the emergent model with the pattern and literature enhances internal validity and generalizability with deeper insight into the model (Eisenhardt, 1989; Yin, 1994, pp. 106-108). Besides, we utilize 'explanation building' strategy to address some research question, reflect theoretical proposition, and build hypotheses (Yin, 1994, p. 110). We utilize NVivo to facilitate data filtering, restructuring, and analysis (Basit, 2003). It helps to simplify and speed up the analysis without losing flexibility (Lindsay, 2004, p. 486).

Qualitative Errors

There are two chief possible errors: sampling and non-sampling error. It is tried to reduce sampling error with accuracy and flexibility in selecting cases (see 5.4.2). Non-sampling error is usually referred to the mistakes made in the acquisition of data - interview. We try to reduce the three types of non-sampling errors: data acquisition errors, non-response errors, and selection bias. It is possible to avoid data acquisition error by choosing appropriate respondents and recording correct responses. Data acquisition error is minimized by (1) fitting equipment, (2) having right procedures for interpretation, (3) recording data accurately, (4) avoiding interviewees from misinterpretation of terms, (5) helping respondents to answer the questions, and (6) concerning sensitive issues. Non-response errors happen when responses are not achieved from some cases. In this case, the methodology is flexible enough to choose extra samples to reach reasonable saturation. Finally, selection bias refers to limitation in choosing some members of the target population. Making sure about key informants aids to fulfill error free results in terms of selection bias.

Reliability and Validity in Qualitative Stage

Although there are some criticism for reliability and validity of qualitative method, we can still utilize various trustworthy approaches to verify research result. Qualitative result can be evaluated against four research quality criteria: construct validity, internal validity, external validity, and reliability (Pandit, 1996). Consistence and systematic data coding enhance reliability (Furlong et al., 2000, p. 542). Start list, interview protocol, systematic coding, and precise analyzing also increase the reliability, especially Diachronic Reliability (Carson et al., 2001, p. 58; Yin, 1994, p. 33).

Construct validity is guaranteed by synergetic views of phenomena and the multiple sources of data (Pandit, 1996; Yin, 1994, p. 33). To maximize this validity, we watch over establishing and implementing clear as well as precision research design (Pandit, 1996). Validity is intensified during coding by establishing causal relationships between categories and their properties and core category (Pandit, 1996). The theoretical reasons for why the relationship exists are basis for the validity (Eisenhardt, 1989). Credibility is enhanced by distinguishing between real relationship and false one. Literature is also helpful while finding rest on limited cases (Eisenhardt, 1989). The closeness of interviewees to the tentative model, also leads to more internal validity. However, Yin (1994, p. 33) hinted internal validity is not issue for exploratory study.

External validity (transferability) Cross-validation study through quantitative method reveals any possible error to rich more generalizable result. We are also aware of ignoring routine finding in support of novel result (Fielding & Fielding, 1986), which causes less stable and generalizable finding and less external validity. In order to overcome this common mistake, scholars suggest systematic qualitative data coding. According to Pandit (1996) and Eisenhardt (1989), well definition of a priori constructs based on literature can enhance the external validity as well. Literature helps us to establish the valid domain to generalize our finding and sharpen external validity (Pandit, 1996).

Quantitative Study

Instrument Design and Data Collection

Subsequent to qualitative study, quantitative method helps to verify the emerging model (Morgan, 1998). The hypotheses come up after refining the construct and developing the propositions (Eisenhardt, 1989) and we employ survey with written questionnaires to examine them. The qualitative investigation helps to validate the tentative properties, while the interviewees assist to contract or expand the tentative dimensions. As a result, we can reach the more reasonable as well as clear dimension and sub-dimensions to operationalize the concept and build up the self-constructed questionnaire. To assess the quality of questionnaire, we run a polite study among 30 MBA students (Zarantonello et al., 2007). Given scaling, all constructs will be deliberated by indicators with a seven-point Likert type, which ranged from strongly disagree to strongly agree (Shamdasani, Mukherjee, & Malhotra, 2008). Questionnaires are supposed to distribute in two ways: (1) personally administered questionnaires for respondents in Penang, and (2) mail questionnaire for respondents who are out of Penang territory. All respondents are furnished with brief presentation and electronic version of questionnaire (PDF Fillable Form) that is distributable (returnable) via internet and email after completion. This format saves time by automatically merging in single data set (i.e. XML format).

Sampling in Quantitative Stage: Methods and Size

There are two main considerations in respect of sampling: (1) choosing a proper sample frame in line with research objective, and (2) the appropriate sample for statistical analysis [ii] . Malaysian hotel industry is appointed whit reference to earlier argument in Error: Reference source not found. To remind, besides, the importance of hotel service in Malaysian tourism industry (Nasution & Mavondo, 2008), it can reflect the concrete and complete picture of CEM in service. On the other hand, it covers some aspects of service experience, which are not accessible in consumer banking (e.g. entertainment or escapist experience).

Thereby, hotel industry (four, and five-star hotels) is the next sample frame for quantitative study. The number of star is considered due to its widely usage in official classification by the countries' governments [iii] , the Tourism Board according to a research World Tourism Organization (WTO) and International Hotel & Restaurant Association (IH&RA) (Lau et al., 2005). Limiting the study to four, and five-star hotels is based on common approach in previous studies (e.g. Punjaisri & Wilson, 2007). In addition, these hotels are more likely to enhance branded customer experience. Likewise, the international characteristic of these hotels let us to test Western-philosophy in Malaysian context (Punjaisri & Wilson, 2007). We concern this frame in all previous stages, especially in mulling over global and regional literature, reviewing the best practices, and emerging tentative model.

According to Ministry of Tourism in 2008, there are 92 Five Star and 98 Four Star listed hotels in Malaysia. Among 190 hotels, 127 samples is reasonable size (Sekaran, 2006, p. 294). On the other hand, SEM requires 100-150 samples (Hair et al., 2006, p. 741). Regarding small response rate in Malaysia sending questionnaire to all relate population is necessary. There is a hope by direct data gathering rather than mail survey (in Penang and nearby territories) response rate will be increased.

Methods of Analysis

After data gathering, handling blank response, categorizing data, and creating data file, the data analysis is started to address the 'getting a feel for data', 'testing goodness of data', and 'hypotheses m development' (Sekaran, 2006, p. 306). Analysis is started by testing reliability (Cronbach's α reliability coefficient) to check internal consistency for entire variables. We employ factor analysis to explain variability among observed variable. It is necessary to build a testable model and assure the concepts are measured correctly. To examine respondent bias we run t-test for early and late respondents to validate research result as cross-sectional study.

Clarifying the descriptive statistic is the next step (e.g. Frequency, Central Tendencies, and Dispersion). Then, we turn into inferential statistic and start - for example - with Pearson Correlation Matrix. The hypotheses testing initiate with Multiple Regression Analysis by estimating model authenticity. Further data analysis is conducted through structural equation modeling (SEM) (Joreskog & Sorbom, 1993; Shamdasani et al., 2008). A two-stage approach (Anderson & Gerbing, 1988) was adopted - first, estimating the measurement model and obtaining the standardized regression coefficients model, and second, estimating the structural model. A total disaggregation approach was followed for all constructs (Bagozzi & Heatherton, 1994).

Structural Equation Modeling (SEM)

SEM is an apt choice for realistic theory testing rather than theory development. It is a mostly confirmatory to determine if the model is valid. There are several additional reasons to justify the SEM application: (1) the power of SEM in modeling constructs as latent variables, (2) the strength in creating more realistic models, (3) simultaneous model analysis, (4) time saving analysis, and (5) flexibility (Chin, 1998). SEM is also common approach within the context (e.g. Brakus et al., 2009; Kao et al., 2007; Pullman & Gross, 2004). SEM involves two components: (1) a measurement model to estimate the correlations/covariance matrix between constructs, and (2) a structural model to estimate the structural coefficients between constructs. In the modeling, first, CEM is considered as endogenous variable and respectively antecedents address exogenous variables; then, CEM become exogenous and consequence becomes endogenous. AMOS by SPSS is utilized for analysis.

Reliability and Validity in Quantitative Stage

Given reliability, Split-half reliability is also utilized to examine internal contingency. Inter-item consistence reliability is examined as well. In respect of internal validity, qualitative study validates the research variable before the quantitative phase. The qualitative study, best practice review, and broad literature review leads us to the optimum content validity in terms of well-delineated dimensions (Harris, 2002). Construct validity is a necessary condition for model building and it achieve through developing latent variable in SEM (Jarvis et al., 2003). It verifies how well the underlying construct is being measured (Churchill, 1979). Convergent validity will be examined through the correlation among the components of CEM (Lages, Lages, & Lages, 2005) and discriminate validity is examined among antecedents.

Ethical Issues

Worldwide-accepted values hearten us to obey the certain code of conduct in all stages of research. The code of conduct consists of value, right, and responsibility that leads to proper actions and infers as Ethic (Furlong et al., 2000, p. 38). Accordingly, five ethical principles, adopted from American Sociological Association (ASA), are observed in this study:

Professional Competence (Maintain the levels of competence, utilize the right capabilities);

Integrity (Honest, fair, respectful… and not jeopardizing, false, misleading, deceptive);

Professional and Scientific Responsibility;

Respect for People's Rights, Dignity, and Diversity;

Social Responsibility

The right of privacy is a vital consideration in current research; regarding the privacy of all respondents in publishing the information is considerable (Zikmund, 2003, p. 79). Back to Sekaran (2003, p. 51), we have to balance our data needs with classified information and ensure firms that the information will remained confidential. We also concern courtesy and professional behavior in interview planning. We avoid putting any pressure on respondents, applying deceiving subject and using the result to harm respondents, or abusing the research outcomes. These aspects of privacy plus legitimacy of research increases the sense of collaboration in respondents (Zikmund, 2003, p. 80).

Moreover, we try our best to be honest, trustworthy, and careful in all data gathering stages. For instance, on the subject of deception, we clarify the real purpose of study by considering the right of inform (Zikmund, 2003, p. 81). Besides, demonstrating the goal of research can be helpful to gain cooperation from interviewees and respondents. In respects of professional competence and scientific responsibility, we have to ensure accuracy by objectivity and scientific research methods (Zikmund, 2003, p. 81). To concern code of ethics in terms of professional responsibility, we follow specific precepts that mostly borrow from The American Association for Public Opinion Research (2005):

Scientific Approach: We are responsible to develop research designs and instruments to satisfy the reliability and validity in data gathering and analyzing via only scientifically accepted tools; however, we are aware of limitation in our techniques and capabilities. We are mindful of avoiding methods mislead survey respondents or misrepresent research.

Representation: We manifest our methods and findings in appropriate detail. We do not reveal/use the names of respondents or organization except with respondents' permission.

Concluding: Our conclusion and interpretations of outcomes should agree with the data.

Dissertation: We hold confidential all findings for the Universiti Sains Malaysia, except when the distribution of the result is authorized by USM. In addition, we also consider our responsibility to the academicians as well as society to disseminate possible findings.