There has been extensive research conducted in the area of service quality and the factors that encourage customers to repurchase a product. In some service settings the servicescape can have a significant effect on what a customer experiences and the subsequent perceptions and behaviour of that customer. The material presented in this document has been extracted from a larger doctoral thesis research study of five hundred coffee shop patrons, which was undertaken to measure the effect of servicescape on consumer's perceptions of service quality, and further investigated the potential significance of the inclusion of the 'servicescape' in a model relating customer service quality with customer behavioural outcomes in a service setting.

The results contribute to marketing theory by empirically showing that the addition of the servicescape construct with other service quality constructs improves the prediction of customer post-purchase perceptions and behaviours with a parsimonious model identified.

However, and of greater significance in contributing to marketing theory, a more complete general service quality model relating customer service quality with post-purchase customer behaviours, the PAWs model of Future Customer Behaviour, was developed combining a specific set of service quality constructs, which may be of value to researchers and practitioners alike in studying various service sector organisations.

The aim of this paper is to review how the service environment or 'servicescape' affects customer post purchase behaviour. Specifically, the research addressed the impact of servicescapes on customer perceptions of service quality. The research developed and tested a new theoretical framework, PAWs model of Future Customer Behaviour. Therefore, PAWs model should provide service providers with a means of evaluating the relationship between the retail establishment and their consumers.

The service sector has grown to become one of the dominant drivers of economic growth (Bigné, Moliner & Sanchz 2003 and Brady, Cronin & Brand 2002). In countries such as Australia, India and the United States of America, the service industry is the fastest growing industry and the largest sectors. The Australian economy has 8.8 million (85 percent) of employed Australians in service industries and accounting for about three-quarters of Gross Domestic Product (GDP), (Australian Bureau of Statistics 2007 and Veldre 2007).

In India the services sector contributes significantly to the countries Gross Domestic Product (GDP), accounting for 56 percent during 2008-09 and according to Ranjeet S. Mudholkar, principal advisor of the Financial Planning Standards Board India (2009), 'with innumerable surveys revealing the immense growth prospects in India and the country being projected as one of the high-growth economies in the world, there is an increasing realisation for the need of more professionals and experts in financial planning' (Mudholkar 2009).

In 1993 the services sector in the United States of America accounted for 66 percent of the Gross National Product (GNP), while 75 percent of all jobs resided in service industries, and 85 percent of all new jobs were created from service providers (Henkoff 1994; Rust & Oliver 1994; Shugan 1994 and Daniels 1993). Since that time the services sector continues to be the fastest growing sector in the USA accounting for 83 percent of the Gross National Product in 2006 (Howells & Barefoot 2007).

The importance of the services sector has resulted in intense competition between service providers (Brady, Cronin & Brand 2002). From an economic perspective, a service providers understanding of consumers evaluative processes contributes to their profitability and market performance from future customer behaviour (Zeithaml 2000; Rust & Zahorik 1993 and Koska 1990). From a social viewpoint, however, service providers have the opportunity to contribute to a consumer's well being as their service is likely to impact on a customer's quality of life (Sirgy 1996 and Sirgy & Lee 1996).

Service Quality (SQ) has become an increasingly indispensable aspect for service providers in managing a successful business operation in today's competitive service market (Blose, Tankersley & Flynn 2005 and Schneider, Holcombe & White 1997) with consumer's perceptions of the actual quality of the customer service delivered and received in relation to the product often influencing a service provider's service quality (Perrone 2009). Consumer reactions to SQ can result in a number of positive benefits, such as:

• higher consumer retention rates;
• attraction of new customers through word-of-mouth referrals;
• successful cross-sell and up-sell opportunities by service providers;
• service provider market share increases;
• improvements in staff morale and productivity;
• a decrease in staff turnover;
• reduction of operating costs; and,
• improved financial performance

(Duncan & Elliot 2004; Ranaweera & Neely 2003; Jamal & Naser 2003; Alexander, Dimitriadis & Markata 2002; Yavas, Bilgin & Shemwell 1997; Buttle 1996; Kim & Kleiner 1996 and Zahorik & Rust 1992).

Since the 1970's service quality in the retail environment has provided marketers with a vast field of research. Kotler (1973) emphasised the importance of retail marketing strategies by focusing on the relevance of atmospherics in influencing consumer behaviour. Marketers have been subsequently exploring ways and means of producing a coherent framework for analysing retail environments (Gilboa & Rafaeli 2003).

'Servicescape', which refers to the physical surroundings in a service environment, was a term first introduced by Mary Jo Bitner (1992). The physical surroundings are important in service settings because customers are affected by these surroundings (Bitner 1992). This study addressed how the servicescape can influence the way consumers react to a product or service provider, and contributes to our understanding of consumer behaviour (d'Astous 2000). This study describes the theoretical foundation and the methodology proposed on the service environment and how that environment relates to subsequent customer perceptions and behaviours.

In order to measure how the environment relates to subsequent customer perceptions and behaviours, the proposed model has three types of variables: service performance, post purchase perceptions and future customer behaviour. The concepts of each variable are provided as follows:

• Customer perceptions of service performance provide service providers with a comprehensive assessment of their business operations, align performance measures, facilitate communication with staff and their customers and provide feedback for training and develop learning (Klenke & Trickey 2002). The three independent constructs associated with service performance are product delivery (the product), functional service quality (often referred to as customer service) and servicescape service quality (the retail environment);
• Post purchase perceptions indicate how customers reflect service performance in terms of customer satisfaction and their perceived relationship with the service provider. There are two mediating constructs associated with post purchase perceptions, customer satisfaction and relationship strength;
• Future customer behaviour is measured by the intentions of customers' to re-purchase and future word-of-mouth behaviour. Service providers who can predict future purchase behaviour, based on their relationships with their customer's by reducing or eliminating potential negative evaluations of their company, will reduce those customers' intentions to switch resulting in financial gains for their firms (Verhoef 2003).

Hence, this study investigated the retail service environment, which was empirically tested in coffee shops and provides for a wider understanding of consumers responses to environmental cues in the retail servicescape. There are numerous research papers on servicescapes however this study supports a model of service quality in which the servicescape is a fundamental part. The outcome of a review of the 'servicescape' literature and its possible impact on service quality, as perceived by customers, has identified a gap in existing service quality models.

The study addressed this gap by, developing and testing a new theoretical framework and, adding the construct servicescape service quality, to other service quality constructs, product delivery and functional service quality and their prediction of customer satisfaction, relationship strength, retention intentions and word-of-mouth. The origin of this theoretical framework can be traced to the seminal conceptualisation of service quality advanced by Grönroos (1990, 1982). Specifically, Grönroos' (1990, 1982) conceptualisation put forward that service quality is the result of customer perceptions of the interaction that takes place during customer service delivery (functional service quality) and product delivery (technical service quality), which is the outcome result of the production process.

When technical and functional service qualities are combined, as the literature suggests, the addition of other constructs such as the physical environment (servicescapes) plays a key factor in the development of service quality perceptions (Dann & Dann 2004; Schiffman & Kankuk 2000; Schiffman, Bednall, Watson, & Kanuk 1997; Zeithaml, Parasuraman, & Berry 1990; Grönroos 1988; Parasuraman, Zeithaml & Berry 1985; Zeithaml, Parasuraman, & Berry 1985; Lovelock 1983; Normann 1983 and Grönroos 1982).

This study therefore developed a framework detailing the relationship between service quality constructs and the servicescape in determining the predictive powers of future customer behaviour by;

• contributing to the marketing literature specifically relating to service quality and customers post purchase behaviours;
• providing a practical element for the services industry with a model relating to improved predictive relationships; and,
• as the review of the literature indicated, there is little empirical evidence to indicate that the servicescape is tested in service quality models that seek to predict future customer behaviour.

The review of the literature revealed that there has been considerable debate associated with service quality due to the conflicting results reported in the literature with many adaptations to the gap based SERVQUAL (Parasuraman, Zeithaml & Berry 1988, 1985) model (Cui, Lewis & Park 2003; Aldlaigan & Buttle 2002; Newman 2001; Bahia & Nantel 2000; Harmen & Vriens 2000; Lee, Lee & Yoo 2000; Van Dyke, Prybutok & Kappelman 1997; Buttle 1996; Lam 1995; Gagliano & Hathcote 1994; Cronin & Taylor 1994; McAlexander, Kaldenberg, & Koenig 1994; Brown, Churchill, & Peter 1993; Cronin & Taylor 1992; Saleh & Ryan 1992; Bitner 1990; Carman 1990 and Babakus & Mangold 1989). The apparent result of the complexity surrounding service quality is that some researchers and academics appear compelled to develop alternative conceptualisations of the construct and to continue to study service quality (Al-Hawari & Ward 2006; Aldlaigan & Buttle 2002; Broderick & Vachirapornpuk 2002; Soteriou & Stavrinides 2000; Frost & Kumar 2000; Dabholkar, Shepherd & Thorpe 2000; Oh 1999; Sweeney, Soutar & Johnson 1997; Philip & Hazlett 1997; Spreng & Mackoy 1996; Dabholkar 1996; Dabholkar, Thorpe & Rentz 1996; Berkley & Gupta 1994; Bowers, Swan & Koehler 1994; Cronin & Taylor 1994; McAlexander, Kaldenberg & Koenig 1994; O'Connor, Shewchuk & Carney 1994; Headley & Miller 1993; Babakus & Mangold 1992; Lytle & Mokwa 1992; Mattsson 1992; Brogowicz, Delene & Lyth 1990; Carman 1990; and Haywood-Farmer 1988).

The purpose of this study was to develop and define such an alternative model. The proposed conceptualisation therefore draws on existing literature to develop a theory, which represents a fusion of the current literatures. One reoccurring theme that continued to emerge from the literature was the conceptualisation of service quality from a consumer's perception of product delivery and customer service outcome. This interaction is a critical part of the overall service product delivery and essential to customers' perception of service quality (Perrone & Ward 2006).

This study drew on a seminal and empirically tested model, the Grönroos (1984, 1982) Model of Perceived Service Quality that divided service quality into two dimensions, technical service quality and functional service quality. The combining of those two constructs and the addition of servicescape service quality provided an informed measure of service quality in developing a new proposed service quality model, PAWs Model of Future Customer Behaviour, to include the servicescape construct and customers post purchase perceptions and behaviours.

In one of the earlier research studies, the importance of managing the atmospherics of the retail environment was recommended to retailers as a means of targeting a customer's experience in this way (Kotler 1973). Further physical environmental research associated with retail providers established that a significant impact on the physical environment could be separated into specific categories; ambient factors, design factors and social factors (Wagner 2000; Bitner 1992 and Baker 1987). As a result of these servicescape models, perceived service quality encompassed all the physical elements influencing the service outcome, delivery and environment and the affects they have on the consumer.

Research has identified and supported numerous studies relating to the impact of consumer behavioural responses to environmental cues resulting in varying positions to marketing variables concerning the interior retail servicescape (Turley & Milliman 2000). Research further indicated that no previous empirical studies have examined the servicescape service quality aspect of the interior retail service setting using the constructs of the proposed theoretical framework with-in this study. The creation of a 'new model' therefore focuses on service based around customers' requirements of the internal retail environment thereby providing retailers with a clearer representation of consumer behaviour.

Based on this foundation, this study presents a framework beginning with the models of three independent constructs, product delivery, functional service quality and servicescape service quality. The two mediating constructs, customer satisfaction and relationship strength. These two dependent post purchase perception constructs indicated how customers reflect a service performance in terms of customer satisfaction and their perceived relationship with a retail service provider. Retention intentions and word-of-mouth communications made up the two dependent future customer behavioural constructs reviewed.

As is evident from the preceding literature review, there is considerable research examining the conceptualisation and operationalisation of perceived service quality. The following sub sections provide a review of the theoretical framework constructs identified.

In developing the theoretical frameworks there were seven key constructs identified in the literature (product delivery, functional service quality, servicescape service quality, customer satisfaction, relationship strength, retention intentions and word-of-mouth). The theoretical framework distinguishes these constructs into three categories, Service Performance (product delivery, functional service quality, servicescape service quality), Post Purchase Perceptions (customer satisfaction, relationship strength), and Future Customer Behaviour (retention intentions and word-of-mouth). The aim of this study was to develop a new conceptual model, which empirically tested service quality relating to future customer behaviour.

The three service performance independent constructs measure consumer's expectations of service quality.

• Product delivery or the technical aspect of service quality directly evaluates the quality of a product that is delivered by an organisation (Grönroos 1984, 1982). In a coffee shop setting for example, it is the finished product or outcome dimension such as glasses, plates, drinks, and food. Product delivery is also identified as a perceived level of service performance that is produced by an organisation in dealing with customer expectations of their interactions with staff (Grönroos 1984, 1982).
• Functional service quality or customer service relates to the level of service a customer receives from staff or how those customers will view the service of an organisation (Grönroos 1984, 1982). Therefore, functional service quality, in the literature, is sometimes called the process related dimension (Grönroos 1984, 1982), and can be exampled by the following statements, “I received a welcome upon arrival, the staff listened to my needs, staff was polite and staff had good product knowledge”. Functional service quality subsequently is the seller-buyer interaction and how consumers are provided for by the organisation (Grönroos 1984, 1982).
• Servicescape service quality measures the general facilities of the retail outlet. Servicescape refers to the overall physical environment of a retail setting for example, the décor, lighting, temperature, music and furniture (Wagner 2000; Bitner 1992 and Baker 1987). Therefore, servicescapes, from a performance perspective, are those physical evidences that organisations effectively utilise to portray aspects of their product delivery where consumers experience the entire physical environment. The servicescape variable in the theoretical framework has been represented as a single construct comprising all of the physical elements. The literature reviewed did reveal that some researchers, (Wagner 2000; Bitner 1992 and Baker 1987), made reference that the retail servicescape consisted of more than one component in its formation, appearance, atmosphere and facilities, which has been reflected in the study.

The two post purchase perception mediating variables measure consumer's perceptions of service quality as manifested in two perception variables:

Customer satisfaction relates to the overall satisfaction a consumer has with an organisation. The literature reviewed was analysed to provide an indicative representation of research studies resulting in customer satisfaction outcomes as they relate to retail industries because of its potential influence on consumer behavioural intentions and customer retention (Anderson & Fornell 1994; Bitner & Hubbert, 1994; Bolton & Drew 1994; Rust & Oliver 1994; Anderson & Sullivan 1993; Cronin & Taylor 1992; Fornell 1992; Oliver & Swan 1989; Oliver 1981, 1980, 1977, and Olson & Dover 1979). Customer satisfaction was perceived as being evaluative and an emotion based response to a retail encounter. Research then suggested customer satisfaction reflected the degree to which a consumer believes that the possession and/or use of a product evoked positive feelings (Jamal & Naser 2003). Some typical statements consumers may ask of an organisation would be, “has staff met my expectations, were services I received adequate, and did I enjoy the product(s)?”

Relationship strength the second post purchase perception construct measures the level of relationship between the consumer, an organisation and its staff (Bove & Johnson 2001; Naude & Buttle 2000; Ward & Smith 1998; Storbacka, Strandvik & Grönroos 1994; Crosby, Evans & Cowles 1990 and Gummesson 1987), by asking the questions as consumers, “Do I have a good rapport with staff? Do I trust the staff? Are staff approachable? Do they make me feel good about my purchases?”. In building relationships with customers, the literature reviewed suggested retail providers have focused on customer satisfaction as one measure in that process with other noteworthy predicting constructs of trust and commitment (Hennig-Thurau, Gwinner & Gremler 2002; Bove & Johnson 2001; Hennig-Thurau & Hansen 2000; Dorsch, Swanson & Kelly 1998; Ward & Smith 1998; Hennig-Thurau & Klee 1997; Hunt & Morgan 1994; Crosby, Evans & Cowles 1990 and Swan & Nolan 1985).

Researchers indicated that those retail service providers, through customer interactions, might be able to add value to their products by promoting a strengthening of those consumer relationships, which in turn may increase retention rates (Beerli, Martin & Quintana 2004). Relationship strength is believed to act as a mediating variable, as it is represented in this study, having an effect on behavioural intentions (Forrester & Maute 2001).

The two future customer behaviour dependent constructs of the theoretical framework measure subsequent consumer behaviour.

Retention intentions concerns a customers future purchase intention being a cognitive component of attitude. “Do I plan on returning to this retailer? When I purchase again will I make that purchase from the same organisation?”. Therefore, retention intentions is a behavioural intention with a subjective probability that consumers will perform some form of behaviour. Consequently, a customer's behavioural intention to continue to patronise an organisation may therefore be based upon their evaluation of the retail provider with a subjective probability the customer will repurchase the product in the future (Fishbein & Ajzen 1975).

Retention intentions may be influenced by a customer's intention to remain with a firm, knowing their product quality, which is based upon the customer's perceptions of the firm's technical (Grönroos 1990), functional (Grönroos 1990) and servicescape, (Bitner 1992) service quality as mediated by customer satisfaction and relationship strength.

Word-of-mouth, the second dependent construct measures a consumer's future referral intention. “Will I refer others? If asked to recommend an organisation would I tell them about the one I went to?” Word-of-mouth theory in the literature suggested that it is a result of consumer responses to the use of products or services associated with cognitive processes, such as perceptions of value and equity evaluations, which consumers experience and is shared as meaningful information with others (Westbrook 1987). Therefore, given this reasoning word-of-mouth is an activity that is likely to influence positive influences of a particular shopping experience (Babin, Darden & Griffin 1994). In this study, word-of-mouth communication is a future customer behavioural variable in the theoretical framework and may take into account the actual quality of the delivery of products provided by the retailer.

The proposed theoretical model contains seven constructs, three independent variables, two mediating variables and two dependent variables, as shown in Figure 1, illustrated model and Figure 2 structural model. The model postulates that the relationships between the constructs provides measures of service performance as evidenced through the service performance of the three independent constructs, Product Delivery, Functional Service Quality and Servicescape Service Quality. These three variables when combined with post purchase perceptions two mediating variables, Customer Satisfaction and Relationship Strength is proposed to predict customer post purchase behaviour.

Figure 1. Original theoretical illustrated model (Model 1)

Figure 2. Original theoretical structural model (Model 1)

Future customer behaviour is assessed, as noted, using two dependent constructs, Retention Intentions and Word-of-Mouth. In the coffee shop scenario used to test the model, the relationship between a customer and the retailer should be statistically related to: Service Performance as assessed by the three independent variables Product Delivery, Functional Service Quality and Servicescape Service Quality leading to Post Purchase Perceptions, as assessed by the two mediating variables Customer Satisfaction and Relationship Strength and then the outcome, Future Customer Behaviour as assessed by the two dependent variables Retention Intentions and Word–of- Mouth.

As there have been numerous research papers on servicescapes as well as the other service quality variables, this study provides for a wider understanding of consumers post purchase behaviour through the development of a new service quality model that combines servicescapes with those specific service quality variables. Therefore, this study addressed this gap to empirically support a service quality model on future customer behaviour and by testing a new theoretical framework.

This section provides a brief description to the method used in the collection and analysis of the data for this study. This study employed exploratory and descriptive research design methods in a two stage process with Stage 1 (exploratory), utilising focus groups and personal interviews, and Stage 2 (descriptive), using quantitative methods for the main data gathering by administering a structured self administered survey. The purpose of Stage 1 was to inform the construction of the test instrument for Stage 2 (Kvale, 1996; Thompson, Locander & Pollio 1989 and Hudson & Ozanne, 1988).

In Stage 1 qualitative methods were used to initially explore consumer's behavioural responses to coffee shop patronage. During the process of conducting interviews and focus groups, insights and perspectives were obtained to provide initial contextual information that enhanced the researcher's knowledge about the research being conducted. The Stage 1 process also assisted the researcher in clarifying the phenomenon being investigated and to identify with the research issues (Hair, Bush & Ortinau 2003; Zikmund 2003; Churchill & Iacobucci 2002 and Cooper & Emory 1995). For this exploratory phase there were a total of 43 participants representing service providers and customers. Two traditional focus group formats were conducted with 17 participants.

During the Stage 1 process an initial survey was constructed arising from a progression of interview and focus group data, consultation with other academics, researchers, consumers, retailers and colleagues thereby developing an understanding of customer attitudes and behaviours when using a product and their criteria in selecting a retail service provider (Edvardsson 1998 and Hayes 1997). The interview process had five primary goals:

• obtain insights into how servicescapes affect consumer behaviour in a retail coffee shop setting
• gain a better understanding of how consumers view their relationship with their preferred retail coffee shop
• gain a better understanding of customers future purchase intentions
• allow the researcher to be fully immersed into the topic
• allow the researcher to have a wider scope into the soundness of the proposed test instrument (Churchill 1995).

The expression of free flowing discussion of the focus groups, coupled with the flexible format, allowed the participants to discuss the research topic with conviction, limited anxiety and inner personal beliefs and feelings as a spontaneous and mutual expression (Sarantakos 2005; Zikmund 2003; Greenbaum 2000 and Blankenship, Breen & Dutka 1998). The focus group format was also advantageous to the pilot study as it provided critical comments and perspectives required in the development and administration of the data to be captured by the test instrument (Aaker, Kumar & Day 2001; Kinnear & Taylor 1996 and Churchill 1995).

In Stage 2, the quantitative data gathering phase, the test instrument was developed to measure the seven main variables of the study with the information obtained from the findings of Stage 1 and the literature reviewed for the study. The data gathering process entailed surveying members of the public who were a random sample from patrons of coffee shops in the greater Melbourne, Victoria Australia area. The data sample consisted of 500 usable test instruments that were stratified by obtaining 50 responses from participants in each of ten different coffee shop locations.

During data analysis descriptive statistics, Exploratory Factor Analysis (EFA) and Confirmatory Factor Analysis (CFA), were conducted, which was a first step prior to analysing the research data using structural equation modelling. Descriptive analysis was performed using SPSS 15.0, which prepared the data providing details describing a set of factors for this study (Sekaran 2003). This process was followed by EFA and then assessment of the measurement model using CFA. Finally, hypotheses testing of the relationships using Structural Equation Modelling (SEM) were conducted with AMOS 7.0 software. As noted Structural Equation Modelling was used to analyse the data and as this research was grounded in a positivism theoretical design, SEM was considered appropriate as the relevance of positivism is complementary to SEM analysis (Guba & Lincoln 1994 and Hunt 1991).

The data gathered during Stage 2 was used to test the competing models developed for this study with the data analysis process described in the next section. During the analysis four models were evaluated to determine best fit indices and by testing the various models to see which set of variables were most significant. This included eliminating all non significant variables and pathways resulting in the most parsimonious model for this study. The process of testing the different models included assessing the original theoretical model with subsequent assessing of the structured models to find the most parsimonious fit model (Kline 1998 and Hulland, Chow & Lam 1996). Figure 3 provides a summary of the design process undertaken in Stage 2 to develop optimum measurement of the variables.

Figure 3. The overall method

The pre-analytical process of the data preparation process ensured that the data were complete and accurate by coding, transcribing or the entering of the data into a computer database and accounting for missing responses to further ensure the results could be meaningfully interpreted from the data (Trochim 2003 and Malhotra 1999). This process included conducting tests for outliners, descriptive statistics of demographic variables, undertaking data transformation and ensuring scale validity and reliability were additionally explored to ascertain a better representation of the data (Sekaran 2003).

The demographic profile of the respondents from the coffee shop samples are shown in Table 1. The table provides the age/gender breakdown for all 500 surveys that were administered. Of the 700 consumers asked to participate in answering the survey, 500 usable test instruments were completed resulting in a response rate of 71.4 percent, which was quite high for business research (Neuman 2003 and Malhotra 1999).

Table 1. Age /Gender Crosstabulation

Exploratory Factor Analysis (EFA) was utilised to examine the factor structure and its fit with the data thus ensuring the items of the survey actually measured the variables they were designed to measure. EFA measured the factorisability of each group of items providing an overview of the factor structure. To control for possible confounding of the data sample, the initial sample of 500 participants was divided into 200 (EFA) and 300 (CFA) data sets respectively. This process controlled the effect of the confounding of participant data in relating the findings of the EFA to the CFA.

Components analysis was then utilised for initial extraction of factors followed by an oblique rotation. The EFA test was conducted to identify items that did not significantly load on factors (Keen 2007; De Vaus 2002; Fabrigar, Wegener, MacCallum & Strahan 1999 and Hair, Anderson, Tatham & Black 1995).

EFA was conducted to verify that the indicator variables of this study aligned with the constructs, which were theoretically proposed including the scale reliability of each of those factors, EFA confirmed that most of the indicator variables did align with the theoretically proposed constructs. Of the seven main constructs, Product Delivery split into two factors, Products Delivery and factors, SSS Facilities, SSS Appearance and SSS Atmosphere. For the variable Word-of-Mouth two positive indicator items aligned with the construct, Retention intentions forming a re-labelled construct Positive Word-of-Mouth (PWOM), while the three negative indicator items remained together thereby being labelled Negative Word-of-Mouth (NWOM).

As the three negative indicator items resulted in a lack of negative correlation with PWOM factor, it was determined that they would be subsequently removed from the study. As theoretically proposed the dependent constructs of Retention Intentions and Word-of-Mouth, through the literature reviewed, are two distinct behaviours therefore in keeping with the theory, the constructs were subsequently split into two distinct variables thereby creating a ten factor model with two dependent constructs, RI and PWOM, as was theoretically proposed. Those items that did not load on a factor were subsequently deleted thereby producing a more than adequate pool of items for each of the factors for further analysis to be performed through CFA.

Following EFA analysis the original theoretical model was developed into four models for comparison using confirmatory factor analysis and structural equation modelling. Following is a brief description of each model:

• Model 1, the original theoretical model (Figures 1 & 2), was assessed as a first step approach in the evaluation process with subsequent assessing of the structural models, (Models 2, 3 and 4) using CFA and SEM (Kline 2005 and Hulland, Chow & Lam 1996). The theoretical model was analysed to establish data fit with relative changes taking place in factor loadings of the theoretical model, which are observed in the subsequent competing models during the testing process (Kline 2005).
• Model 2, the revised theoretical model, (Figures 4 & 5) is a modification to the original theoretical framework with a resulting ten factor model by splitting product delivery into two distinct constructs, product and product delivery, splitting servicescape service quality into three constructs, servicescape service quality facilities, servicescape service quality atmosphere and servicescape service quality appearance. The variable word-of-mouth partially combined with the variable retention intentions with two positive indicator items moving to that variable while the three negative word-of-mouth items remained as a separate construct, which was subsequently removed from the study.

Figure 4. Revised theoretical illustrated model (Model 2)

Figure 5. Revised theoretical structural model with RI/PWOM & NWOM

• Model 3, the revised ten factor model, PAWs model of Future Customer Behaviour (Figures 6 & 7), shows the removal of the construct, NWOM, and the splitting of the RI /PWOM variables. The revised ten factor model was re-run by separating the three retention intention items and the two remaining positive word-of-mouth items into two distinct factors as reviewed in the literature and shown in the original conceptual model (Figure 1).

Figure 6. PAWs illustrated model of Future Customer Behaviour (Model 3)

Figure 7. PAWs structural model of Future Customer Behaviour (Model 3)

• Model 4 tests the model without the servicescape variable Figures 8 & 9 to determine if the exclusion of the servicescape variable significantly affects the relationship between service quality as measured by the other variables of the model and the impact on customer perceptions and future behaviours.

Figure 8. Illustrated model without the servicescape variable (Model 4)

Figure 9. Structural model minus SSSQ with RI and PWOM split (Model 4)

Confirmatory Factor Analysis (CFA) was used to demonstrate empirically that the hypothesised model fits the data and CFA confirmed the factor structure and the identification of the underlying factors as produced by the Exploratory Factor Analysis (EFA) (Bollen 1989 and Anderson & Gerbing 1988). The distinction between EFA and CFA is not always clear cut therefore CFA results were used to incrementally re-define the measurement models (Gerbing & Hamilton 1996; Bollen 1989 and Anderson & Gerbing 1988). To confirm that the measurement items were adequate using CFA, the proposed model for this research was estimated by conducting evaluations of its overall statistical fit, goodness-of-fit criteria and the subsequent assessment of fit of the model.

The conceptual model developed puts forward service quality as comprising seven primary constructs, product delivery, functional service quality, servicescape service quality, customer satisfaction, relationship strength, retention intentions and word–of-mouth. A CFA test was then conducted on the four models with the results tabulated at Table 2 for comparison.

Table 2. CFA summary of model fit statistics for the four proposed models

In summary, during the CFA process all factor loadings were estimated (freed), so that items were allowed to load only on one construct with no cross loadings. The research also met some basic criteria for CFA testing; 1) sample size, 300 test instruments, sixty percent, of the original sample of 500 surveys were used, which was more than adequate (Hair, Anderson, Tatham & Black 1995 and Anderson & Gerbing 1988), and, 2) as there was normal distribution of the variables, the maximum likelihood discrepancy function was appropriate for the complexion of the data collection process (Bentler & Dudgeon 1996).

Other assessment factors evaluated were Cronbach alpha, the significance of Barlett's Test of Spherity and the KMO, variance extracted and goodness of fit indices for the models satisfying the acceptable requirements based on these tests for good model fit. The fit indices (Table 2) for the four models all fitted well with the survey data with Model 3, PAWs model of Future Customer Behaviour, proving to fit the data best.

The next step in the evaluation process involved comparing models (Kline 2005and Anderson & Gerbing 1988) using Structural Equation Modelling (SEM) as an analytical tool, which has been a successful approach for marketing research (Baumgartner & Homburg 1996). In order to decide which SEM model was the most parsimonious, the process of eliminating non-significant pathways and subsequently variables did yield the most significant and parsimonious model as presented in Figures 10 & 11. Those variables and items that were removed were in relation to Product and Product Delivery as the data indicated those two constructs and relating items had little effect on the overall outcome.

The most parsimonious model is the simplest that gives a high R² but not necessarily the highest R² thereby providing a trade off between complexity and power as the R² shows the best predictor model. Table 3 presents and compares the five models R² with all five indicating they had significant results. Model 5, the most parsimonious model, is highlighted.

Table 3. Summary of R² for the five proposed models presented

Figure 10. Most Parsimonious illustrated model (Model 5)

Figure 11. Most Parsimonious structural model (Model 5)

The testing indicated that Model 3, the PAWs model of Future Customer Behaviour, presented fundamentally sound with the theoretical framework including having acceptable fit indices. The results indicated that the PAWs model of Future Customer Behaviour was an acceptable model and would therefore be a reliable model to use in the testing of other service settings. The PAWs model also revealed that the servicescape construct did have a marginally better result than Model 4, which was tested without the servicescape construct for this particular study.

• Model 5 however, did present with a slightly greater R2 for the variable PWOM. Therefore, for this research Model 5, with the removal of those non-significant paths was the most parsimonious model, for this study on coffee shops.

Having determined and established the final structural equation model, Model 3, examination of the path coefficients and the significance levels between the constructs of the model was conducted. A critical ratio, an index used to determine the significance or non-significance of a structured path, were deemed significant if they were greater than a conventional threshold of  ±1.96 (Hair, Anderson, Tatham & Black 1995). The critical ratios of all paths were significant. The standard regression weights and path significance for Model 3 PAWs model of Future Customer Behaviour and Model 5, most parsimonious model are presented in Tables 4 & 5.

Table 4. Standardised regression weights and path significance levels

Table 5. Standardised regression weights and path significance levels