Customer Experiences affect Customer Loyalty: An Empirical Investigation of the Starbucks Experience using Structural Equation Modeling

Textbook 2013 62 Pages


Table of Contents

List of Figures

List of Tables

List of Abbreviations

1. Introduction

2. Literature Review
2.1 The Evolution from Products to Services to Experiences
2.2 The Initial Conceptual Model
2.2.1 Customer Experience Quality
2.2.2 Perceived Value
2.2.3 Customer Loyalty
2.2.4 Perceived Wealth

3. Methodology & Research Design
3.1 Assigning Scales to the Individual Constructs
3.2 Pre-Testing the Scales
3.2.1 Data Collection
3.2.2 Scale Purification Process CXQ scale Perceived Wealth scale Perceived Value scale Customer Loyalty scale
3.3 Adjustments and Refinements
3.4 Testing the Measurement Model
3.4.1 Data Collection
3.4.2 Measurement Model
3.4.3 Assessing Model Fit of the Measurement Model
3.4.4 Assessing Validity of the Measurement Model

4. Data Analysis
4.1 Comparison of Competing Models
4.1.1 Interpretation of Structural Model #1 suggesting Full Mediation
4.1.2 Interpretation of Structural Model #2 suggesting Partial Mediation
4.1.3 Interpretation of Structural Model #3 suggesting No Mediation
4.2 Selection of the Best Fitting Model

5. Discussion

6. Conclusion
6.1 Theoretical Implications
6.2 Managerial Implications
6.3 Limitations & Future Research

Reference List


List of Figures

Figure 1.Price of coffee offering from commodity to experience

Figure 2. Progression of Economic Value

Figure 3.Initial Conceptual Model

Figure 4.Refined Conceptual Model after Pre-Test

Figure 5. Measurement Model #1.0 in AMOS

Figure 6.Final Measurement Model #1.2 in AMOS

Figure 7. Structural Model #

Figure 8.Structural Model #

Figure 9.Structural Model #3

List of Tables

Table 1.Existing Conceptualizations of Customer Experience Quality

Table 2.Dimensions of Customer Experience Quality

Table 3.Results of Factor and Reliability Analysis for the CXQ scale

Table 4.Results of Factor and Reliability Analysis for the Product Quality scale

Table 5.Results of Factor and Reliability Analysis for the Perceived Wealth scale

Table 6.Results of Factor and Reliability Analysis for the Perceived Value scale

Table 7.Results of Factor and Reliability Analysis for the Loyalty Intention scale

Table 8.Model Fit of Measurement Models

Table 9.Standardized Total Effects in Measurement Model #

Table 10.Model Fit of Structural Models (Full Dataset)

Table 11.Multi-group effects of Perceived Wealth in Structural Model #1 (e5 fixed)

Table 12.Multi-group effects of Perceived Wealth in Structural Model #2 (e5 fixed)

Table 13.Multi-group effects of Perceived Wealth in Structural Model #3 (e5 fixed)

Mulaik, S.A., James, L.R., Van Alstine, J., Bennett, N., Lind, S. & Stilwell, C.D. (1989). Evaluation of goodness-of-fit indices for structural equation models. Psychological Bulletin, 105, 430–445.

James, L.R., Mulaik, S.A. & Brett, J.M. (1982). Causal analysis: Assumptions, models and data. Beverly Hills: Sage.

List of Abbreviations

Abbildung in dieser Leseprobe nicht enthalten

1. Introduction

The world has changed. Customers’ increasing expectations towards companies make competition continuously harder. What it needs are new business strategies that attract customers and, even more important, business strategies that make customers loyal. This is not a new phenomenon but rather a continuous evolution in business that especially gained momentum in the last years. Over time a development took place from products to services to a post-product, post-service phenomenon which is still evolving (Maklan and Klaus, 2011).

Marketers came to the point where they realized that striving for mere customer satisfaction might not be the panacea to create customer loyalty as it was expected to be. More than 60% of customers who switch to another brand identify themselves as satisfied (Jones, 1996; Reichheld, 1993). Regardless of these academic insights, most companies still rely on it. A report by Euromonitor International based on observations in the American market acknowledges the necessity for marketers to reset strategies. “Provide not only tangible products, but also unforgettable experiences!” (Euromonitor, 2008). In recent years, customer experiences as the ultimate competitive element increasingly gained attention among practicioners and theorists (e.g., Maklan and Klaus, 2011; Verhoef et al., 2008; Gentile et al., 2007; LaSalle and Britton, 2003; Carù and Cova, 2003; Pine and Gilmore, 1998).

It is commonly acknowledged that consumption activities almost always contain experiential aspects (Holbrook and Hirschman, 1982) and it should be the company’s vital mission to make them extraordinary and compelling in order to differentiate from competition and gain a competitive advantage (Pine and Gilmore, 1998). In 2013 we can already find a customer experience focused mindset in some companies’ mission statements (e.g., Dell: “Dell’s mission is to be the most successful computer company in the world at delivering the best customer experience in markets we serve.”) or core values (e.g., McDonald’s: “We place the customer experience at the core of all we do.”).

An often cited example of experience staging and its financial benefits for the company is the American Coffee Company Starbucks. During the transformation from a commodity (=harvested coffee beans) into a good (=roasted, grinded, packaged coffee beans) into a service (=simply served cup of brewed coffee) and finally into an experience (=extraordinary way of ordering, creation, and consumption of a cup of coffee) the price charged increases exponentially (Pine and Gilmore, 2011). Figure 1 shows the different price ranges that are paid on average by customers in the distinct “evolutionary stages” for a cup of coffee in the United States.

Figure 1.Price of coffee offering from commodity to experience (Pine and Gilmore, 2011)

Abbildung in dieser Leseprobe nicht enthalten

We see that it is crucial for managers to extend their mindset beyond product quality and service quality. They need to manage the experience their company offers. But this is easier said than done. The American statistician William Edwards Deming once said “You cannot manage what you cannot measure” and that explains nicely the current shortcoming of customer experience literature. Although “customer experience” has become an omnipresent buzzword in marketing, it still has severe theoretical shortcomings in terms of academic research. In the existing literature, the importance of a successful management of the customer experience is indeed emphasized, but researchers often fail to measure customer experiences holisticly and if they do, like Lax (2012) also recognizes, they usually fail to link customer experience quality to the larger context of business outcomes and customer loyalty. Hence, it is hard to manage customer experiences on the basis of existing models. What is needed as a proper foundation is a conceptual framework with a set of meaningful measurement scales that links the quality of a customer experience to customer behavioral outcomes, especially customer loyalty.

Therefore, the study at hand aims to overcome the aforementioned pitfalls and presents a model that 1) holisticly measures customer experience quality (with the newly developed CXQ scale) and 2) links the different dimensions of customer experience quality to customer loyalty intentions in the form of word-of-mouth and the customer’s willingness to pay more. 3) In order to gain valuable insights for customer segmentation, perceived wealth is introduced as a moderating variable in the model. The study hereby primarily investigates the effect of customer experience quality on the different facets of customer loyalty.

In chapter two, the conceptual model will be presented and acts as a framework for the discussion on existing literature regarding customer experiences, including its implications and gaps. On this basis the hypotheses will be developed.

Chapter three describes the research design and methodology used to empirically test the conceptual model and hypotheses. Structural Equation Modeling (SEM) will be applied in this study to validate the model and test the hypotheses. Based on common sense and the literature discussed in chapter three, the proposed scales to measure the distinct constructs will be presented. Via exploratory factor analysis the scales will be purified and refined to be afterwards transferred into a measurement model which is tested via confirmatory factor analysis. As soon as the fit, reliability and validity of the measurement model is assessed, the relationships as stated by the hypotheses will be integrated into a set of competing structural models and tested in chapter four.

In chapter five, explanations for the observed findings will be discussed critically. During the conclusion in chapter six, the author outlines the theoretical and managerial implications in detail, discusses limitations of the study, and suggests topics for future research.

In advance, the findings of this study are valuable for both theory and practice. It will bring companies one major step forward towards the successful management of customer experiences and allows companies to stage them efficiently and effectively in order to gain a significant competitive advantage. Other researchers can use the CXQ measurement scale for future research and introduce new moderators in order to advise companies in terms of customer segmentation. Additional behavioral outcome or business outcome variables which are integrated in the model might lead to valuable insights as well.

2. Literature Review

Customer experience is a popular, but equivocal marketing buzzword. There are as many different academic conceptualizations as there are scholars and measuring it is a challenge due to its latent nature. Although the literature regarding customer experiences already evolved three decades ago from articles such as Hirschmann and Holbrook (1982), dealing with experiential aspects of consumption, literature still fails to provide a precise terminology and standardized, generalizable approaches (Gentile et al., 2007). To provide the reader with a profound overview of the topic without losing ourselves in the width of diverging definitions, the following literature review will be structured as follows: First, the author will describe the ongoing process in today’s business to increasingly focus on experiences as a progression from products and services. Second, the author presents a conceptual model that combines insights of different studies from the fields of marketing, psychology as well as financial economics. The conceptual model serves as a framework to discuss the different variables, logically relates them to each other based on previous research and visualizes the hypotheses which are consequtively tested in chapter three, four and five.

2.1 The Evolution from Products to Services to Experiences

Over the past three decades, marketing theory underwent several large-scale paradigm shifts. The described paradigm shifts in theoretical literature reflect real-world changes in the competitive landscape that companies face. Maklan and Klaus (2011) conclude that a development from products to services to a post-product, post-service phenomenon took place which is still evolving. It is important to emphasize that the requirements for companies were not substituted over the different phases but rather extended into new dimensions, making competition more and more complex.

In the 1990s the first paradigm shift occurred when marketers took the relational aspects between customer and company into account (Grönroos, 1997; Christopher, 1996). Before, the classic product marketing was largely focused on sales and the creation of fast moving consumer good brands (Merz and Vargo, 2009; Copeland, 1923). Now, instead of bringing products “to market” and considering consumers as targets, firms started to co-create value collaboratively and “marketed with” their customers over an extended time frame. Instead of “delivering value” like in previous eras, the firm’s role was now seen as “proposing value” which was ultimately co-created when the customer uses the firm’s products and services (Vargo and Lusch, 2004). This so-called “value-in-use” is not embedded in a product or service at the moment of exchange, but rather obtained via usage processes (Tynan et al., 2010; Macdonald et al., 2009).

The dichotomous interpretation of goods and services which characterized prior research impeded a combined, simultaneous conceptualization of the two. This was finally resolved when Vargo and Lusch (2004) presented their service-dominant logic. According to Vargo and Lusch (2004), services and goods can be seen as distribution mechanisms for service provision. All economies hereby are considered as service economies (Vargo and Lusch, 2004) and all market offers can be interpreted as customer-centric product-service systems that fulfill customer needs (Ulaga and Reinartz, 2011; Shankar et al., 2009; Baines, 2007). In the following years, value-added services were integrated in the portfolio of many manufacturers to increase their customers’ value-in-use.

Today, as competition quickly adapts, also (value-added) services are increasingly commoditized and no longer sufficient to guarantee a competitive advantage (Meyer and Schwager, 2007; Shaw, 2002; Schmitt, 1999). Several scholars recently identified customer experiences as the ultimate competitive element and their effective management as the most promising strategy for company’s long-term success (e.g., Lax, 2012; Lemke et al. 2010; SAS, 2009; Klaus and Maklan, 2007). Pine and Gilmore (2011) even go so far to claim the rise of the Experience Economy as a logical progression from Service Economy. Addis and Holbrook (2001) interpret the overall economic development within the last decades as an evolution from mass production to mass customization, replacing supply chains with demand chains. The idea behind mass customization is to serve customers uniquely and efficiently. Hereby, mass customizing any good turns a good automatically into a service; mass customizing any service turns a service automatically into an experience (Pine and Gilmore, 2011; Addis and Holbrook, 2001). Figure 2 illustrates this Progression of Economic Value.

Figure 2: Progression of Economic Value, based on Pine and Gilmore (2011)

Abbildung in dieser Leseprobe nicht enthalten

The degree of standardization continuously decreases from commodity to experience. As a company moves up the ladder it is able to fulfill individual needs better and becomes more relevant to its customers. Due to this individual focus and personal relevance, a company which stages experiences can easily differentiate from competitors and is able to charge a premium price for its offerings based on the distinctive value provided, instead of a market price dictated by competition (Pine and Gilmore, 2011).

But how can be evaluated if a company has reached the state of experience staging? How can the quality of a customer experience be measured? One measurement instrument for this purpose which gained more and more attention in the last years is the construct of ‘customer experience quality’.

2.2 The Initial Conceptual Model

The conceptual model shown in Figure 3 presents the proposed relationships among the single constructs of interest: customer experience quality, perceived value, perceived wealth and customer loyalty in the form of the customer’s intention to recommend the company to others and the willingness to pay more.

Figure 3Initial Conceptual Model

Abbildung in dieser Leseprobe nicht enthalten

The single constructs will be subsequently described in more detail. The basic idea in a nutshell is the following: As was found in the studies of Lemke et al. (2010) and Hueiju and Wenchang (2009), perceived value is supposed to play a mediating role between customer experience quality and customer loyalty. Perceived wealth is introduced as a moderating variable. The author hypothesizes that the more wealthy customers perceive themselves, the more dimensions of (customer experience) quality they take into account when evaluating the value they receive and hereby the higher is the effect of customer experience quality on perceived value and, in turn, customer loyalty.

2.2.1 Customer Experience Quality

The existing practically oriented literature on customer experiences claims that customer experiences need to be extraordinary, memorable and compelling in order to generate a competitive advantage (e.g., Pine and Gilmore, 2011; Holbrook, 2007; LaSalle and Britton, 2003; Schmitt, 2003; Addis and Holbrook, 2001). But how is it possible to measure an experience, and moreover, how is it possible to manage customer experiences?

First of all, one needs to understand the nature of customer experiences. Customer experiences blur the traditional dichotomy of goods and services by ultimately focusing on customers’ value-in-use, which is created by the orchestral combination of goods and services during the interaction between customer and company (Poullson and Kale, 2004). Customer experiences are by nature co-created by customers and lead to value perceptions both on a cognitive as well as on an affective level (Prahalad and Ramaswamy, 2004). A customer experience hereby is a holistic personal, customer specific perception of a company’s overall market offering which is generated in a wide array of situations and contains a significant amount of hedonic benefits and emotional value for the customer. According to Bruhn and Hadwich (2012), the academic literature regarding customer experiences can be classified into four streams of research: product experience (e.g., Hoch, 1989), service experience (e.g., Patricio et al., 2011), brand experience (e.g., Brakus et al., 2009), and consumption experience (e.g., Hirschmann and Holbrook, 1982). The study at hand tries to merge the parallel concepts of product experience, service experience, and consumption experience by consolidating them in a holistic quality scale.

Similar to the evolution from products to services to experiences, the scales used for measuring quality have evolved and were continuously extended. Customer assessments of quality are phenomenological in nature and quality is commonly defined (e.g., Caruana et al., 2000; Zeithaml et al. 1996; Parasuramam et al. 1988) as a “perceived judgment about an entity’s overall excellence or superiority” (Zeithaml, 1988, p.3). According to the evolution outlined in section 2.1, product quality can be considered as the very basic dimension of quality. It is commonly agreed on that product quality scales are embedded-value measures and have a rather limited scope of what they measure (Lemke et al., 2010; Parasumaran et al., 1988). Moreover, they are largely limited to cognitive evaluations and hereby insufficient to capture the abundance of feelings a person develops when being a customer of a specific company. Service quality scales, such as the SERVQUAL scale, (Parasumaran et al., 1988), acknowledge more quality dimensions beyond tangible product attributes (e.g. reliability, responsiveness, assurance, and empathy) but still fail to measure affective and especially hedonic components properly (Lemke et al, 2010). Therefore it is necessary to conceptualize a scale which also enables the holistic measurement of customer’s feelings when experiencing to be a customer of a specific company.

A common measurement for customer experiences, although very inconsistent regarding the employed scale items, is the use of ‘customer experience quality´ (e.g, Maklan and Klaus, 2011; Lemke et al., 2010; Hueiju and Wenchang, 2009; Verhoef et al., 2009). Table 1 presents an overview of the most important conceptualizations of this construct, in how far the scholars related customer experience quality to other variables and which study design was used. Research context and sample size are reported as well.

In the author’s opinion, Ting-Yueh and Shun-Ching (2010) so far have created the most convincing scale to measure customer experience quality. Nevertheless, they in return fail to consider product quality as a crucial element of a customer experience. Studies of Maklan and Klaus (2011) and Lemke et al. (2010) have demonstrated that customers evidentially also consider product attributes when judging customer experience quality. These studies did include product quality as a dimension of customer experience quality, but did not capture the affective components of the experience properly. Therefore, the author suggests the creation of a new scale that overcomes the pitfalls of previous research. This new scale for customer experience quality (the CXQ scale) will be developed on the basis of Ting-Yueh and Shun-Ching’s (2010) five major dimensions and additionally contains the dimension product quality (see also section Table 2 presents an overview of each of the six proposed major dimensions of the CXQ scale. Taking the sub-dimensions into account as well, customer experience quality hereby incorporates nine different dimensions. Table 2 summarizes each of the dimensions and explains in how far it affects the overall construct. The single items of the CXQ scale can be found in Table A1 and are discussed and refined in chapter three.

Table 1.Existing Conceptualizations of Customer Experience Quality

Abbildung in dieser Leseprobe nicht enthalten

The CXQ scale includes experiential factors of which companies have direct control as well as factors they cannot control directly. Based on the ratio between controllable and uncontrollable factors, Carù and Cova (2007) introduced a “continuum of consuming experiences” that ranges from experiences which are primarily developed by the company to

Table 2.Dimensions of Customer Experience Quality

Abbildung in dieser Leseprobe nicht enthalten

experiences which are largely constructed by the consumers. For the measurement of customer experience quality this information might be negligible, however, for its management it is crucial. Verhoef et al. (2009) introduced the dimension of time to distinguish among experiences prior, at, and post purchase. Related to this differentiation, the presented CXQ scale can be seen as an aggregation across time and measures the quality of the overall customer experience with a particular focus on the phases of purchase and consumption of the product.

This aggregation is considered to be sensible. As customers rather update their overall impression of a company than dissect their experiences with surgical accuracy, what is important in the end is how the customer subjectively perceives its overall quality.

2.2.2 Perceived Value

The perception of quality regarding experiences is the ultimate foundation of the proposed model but it is not sufficient to explain customer behavioral intentions. Bolton and Drew (1991), Caruana et al. (2000), Oliver (1999), and Sweeney and Soutar (2001) found that quality does not directly lead to behavioral outcomes such as purchase. Rather, it indirectly influences behavioral outcomes via a value perception that mediates the relationship. Perceived value is defined as “the consumer’s objective assessment of the utility of a brand based on perceptions of what is given up for what is received” (Gupta and Zeithaml, 2006). The customer may perceive value in each phase of the customer journey, also including aspects outside the company’s direct control but still related to the company, such as when interacting with other customers (Verhoef et al., 2009). On this basis the author concludes that customer experience quality as conceptualized in this research can be considered an antecedent of perceived value. Moreover, it will be hypothesized that perceived value fully mediates the relationship of customer experience quality and customer loyalty.

2.2.3 Customer Loyalty

Customer loyalty can be investigated either behaviorally or psychologically. Behaviorally, consumers are defined as loyal if they continue to buy the same product over a certain time period (Gupta and Zeithaml, 2006). This is usually measured as repeat purchase frequency or relative volume of purchasing (e.g., Tellis, 1988). However, Jacoby and Chestnut (1978) point out that this kind of behavioral customer loyalty is in some cases merely the result of convenience or high switching costs and hereby might be spurious. Customer loyalty can also be measured by psychological indicators, i.e. the customer’s intention to perform a diverse set of certain behaviors. Intentions can be seen as the psychological antecedents of behavior. In terms of customer loyalty, these intentions comprise repurchase intention (e.g., Reynolds and Arnold, 2000), intention to recommend to others, also known as word-of-mouth (WOM) (e.g., Mattila, 2001), likelihood of switching and likelihood of buying more (e.g., Selnes and Gonhaug, 2000). Zeithaml et al. (1996) merge these four aspects of loyalty into a behavioral-intentions battery with four factors – loyalty (i.e. recommend company to others), propensity to switch, willingness to pay more, and external response to service problems.

Because of the aforementioned risk of a spurious relationship between customer experience quality and behavioral customer loyalty, the author decides to apply the concept of psychological loyalty intentions. With regard to Reichheld (2003) who advocates that complex measures besides customers’ intention to recommend the company to others are unnecessary to capture loyalty, the model will be kept as simple as possible. Therefore not all of Zeithaml et al.’s (1996) identified aspects will be taken into account. Only two factors are employed in the measurement model: 1) intention to recommend the company to others and 2) since people apparently are willing to accept an exponential increase in prices for experiences (Figure 1), also customer’s willingness to pay more will be considered.

Hypothesis 1 and 2 connect the aforementioned constructs in the following way:

H1 Customer experience quality affects customer’s intention to recommend the company to others positively and indirectly through perceived value.

H2 Customer experience quality affects customer’s willingness to pay more positively and indirectly through perceived value.

2.2.4 Perceived Wealth

In financial economics, a consumer’s wealth is defined by several observable variables, i.e. cash balances, government bonds, housing equity, stocks, other assets, and debts. This concept of wealth hereby assesses quantitatively how ‘wealthy’ or ‘rich’ a person actually is. By aggregating this wealth across a nation or a sample, and linking it to aggregated consumption in a longitudinal study, financial economists were able to find evidence for the so called ‘wealth effect’ which is measured by the wealth elasticity of demand (e.g., Peltonen et al., 2012; Campbell and Cocco, 2007). The wealth elasticity of demand describes the proportional change in consumption of a good relative to a change in consumer’s wealth. The wealth effect predicts that an increase in wealth leads to an increase in spending, i.e. people are both willing to purchase a higher quantity of products, which stimulates repurchases, and are willing to accept higher prices.

Peltonen et al. (2012) and Campbell and Cocco (2007) find evidence for the wealth effect in several markets all over the world. However, the effect of wealth on consumption differs significantly across different countries. In order to make generalizable predictions and to be able to roll out the study globally, the author decided to analyze the respondents’ wealth psychologically by measuring perceived wealth instead of measuring wealth by a set of observable financial indicators. This implies that in order to observe a wealth effect, people not actually need to be richer but merely need to perceive themselves to be richer. In the author’s opinion this is a justifiable and reasonable adjustment. To the author’s experience, consumers rarely are up-to-date regarding the volatile exact quantitative value of all their assets and debts but rather have a broad perception of their current financial status in mind. Measuring perceived wealth therefore can be considered a meaningful instrument when aiming to predict consumer behavior. The measurement of perception has the additional advantage that income inequalities across countries are automatically corrected. Usually these inequalities are controlled for by calculating estimates like purchasing power parities or the consumer price index. Unfortunately, as Almas and Shafir (2012) find out, even these corrected estimates are significantly biased. Perceived wealth therefore is a promising alternative for the study at hand. Since consumer perceptions are rather relative than absolute (Ariely, 2009), the respondent automatically relates his or her personal situation to others around him, hereby controlling for international income inequalities.

Jones and Mustiful (1996) investigate the differences in purchasing behavior between lower- and higher-income shoppers regarding breakfast cereals. They find out that compared to higher-income shoppers, lower-income shoppers make more rational purchase decisions as defined by consumer theory, i.e. their purchase behavior is strongly guided by their income and product prices. These findings suggest that lower-income shoppers either evaluate the quality of private label and national brands to be similar or they find the price differential to be of insufficient magnitude to justify the difference in quality. Following up on the first explanation, higher-income customers are more likely to take additional quality attributes into account when making their purchase decisions. Applied to the concept of customer experiences, less wealthy customers are supposed to have a narrower definition of quality and are supposed to care merely about the ratio of price to product quality while more wealthy customers take additional quality dimensions of the CXQ scale into account, increasing the importance of customer experience quality for perceived value.

Maslow’s well known Hierarchy of Needs points towards a similar direction. Although Maslow’s theory may not always hold true, it still provides a general framework for categorizing and prioritizing needs and serves as a common reasoning applied by marketers to segment markets (Schiffmann and Kanuk, 2000; Wahba and Bridwell, 1976). According to Maslow (1987), there are five layers of deficiency needs: physiological needs; safety needs; belongingness & love needs; esteem needs; and self-actualization needs. The hierarchy hereby reaches from the physical requirements for human survival at the lowest level to the desire to accomplish everything that one can in a form of mastery at the highest level. The most basic needs must be satisfied before the individual desires secondary or higher level needs and starts striving for constant betterment. Maslow and Lowery (1998) later differentiated the highest need of self-actualization in more detail and identified four subcategories: 1) cognitive: to know, to understand, and explore; 2) aesthetic: symmetry, order, and beauty; 3) self-actualization: to find self-fulfillment and realize one's potential; and 4) self-transcendence: to connect to something beyond the ego or to help others find self-fulfillment and realize their potential. For the concept of customer experiences, the cognitive and aesthetic aspect is of particular interest. Assuming that the different individual needs are increasingly satisfied with rising (perceived) wealth (Trigg, 2004), consumers who perceive themselves as more wealthy likely desire self-actualization and increasingly take cognitive and aesthetic aspects into account. As argumented by Hueiju and Wenchang (2009), product quality is considered to fulfill basic needs such as physiological and security needs. Service quality represents social needs in the form of belongingness and esteem needs, whereas experience quality includes the even higher needs of self actualization, knowledge/understanding and aesthetics.

The focus of wealthy customers hereby extends from the basic dimension of product quality to higher, additional dimensions of quality, represented in the overall customer experience quality scale when evaluating the perceived value offered by a company and the impact of customer experience quality. The effect of customer experience quality on perceived value hereby increases with higher perceived wealth of the customer.

This moderating effect of perceived wealth in the model is summarized in Hypothesis 3:

H3 Perceived wealth moderates the relationship between customer experience quality and perceived value, such that the wealthier the customer perceives him- or herself, the more the perceived value is affected by customer experience quality.

3. Methodology & Research Design

In order to manage experiences in a way that customer loyalty intentions are maximized, it is crucial to understand which aspects people really remember from the purchase process. Hoch and Deighton (1989) explain that remembered purchase experiences greatly influence future behavior. That means, when deciding to choose a company, individuals first recall their past experiences. This reasoning is in line with Pine and Gilmore (1998) who claim that experiences should be memorable. Therefore, in contrast to existing studies about customer experience quality (see Table 1) in which the measurements were made right after customers/ respondents have left the store, this study will measure customer experience quality with a significant time lag. This approach ensures that only the remembered, memorable experience will be measured and linked to behavioral intentions.

The American Coffee Company Starbucks will serve as an exemplary company to which the questionnaire will be adapted in order to transfer the theory into practice and evaluate the proposed hypotheses. Starbucks was chosen because it has a worldwide network of stores with a consistent corporate design and a global strategy focused on staging experiences. Starbucks is known to many people and has already proven to be a good example in previous research on customer experiences (e.g., Chang and Horng, 2010; Hueiju and Wenchang, 2009).

As outlined in the introduction, SEM will be applied in this study. The concept of customer experience quality and the other constructs entirely based on perceptions are all latent constructs, which makes this study a perfect case to apply this advanced technique of multivariate dependence analysis. SEM enables the researcher to assess the measurement properties and test the proposed theoretical relationships in a unified and integrated manner.

The analysis of a model in SEM hereby consists of two major parts. First, the measurement model in SEM is similar to a factor analysis. On the basis of theory and/or a preliminary exploratory factor analysis which reveals the underlying structure, the researcher specifies in which way observed variables load on latent factors/constructs in the model. Afterwards, in the form of a confirmatory factor analysis, all of the equations are estimated simultaneously and it is tested whether the measurement theory holds true. Second, to test particular research hypotheses a structural model connects the latent constructs via dependence relationships. Again, the entire model is estimated at once. This procedure is similar to conducting factor analysis and a series of multiple regression analyses at once. Endogenuous constructs in the model can be compared with dependent variables in a regression, exogenuous variables are independent variables. However, this analogy has to be treated with caution. Due to the simultaneous estimation of the entire model some of the endogenuous variables might serve as a dependent variable in one relationship/equation, while being an independent variable in another relationship/equation (Malhotra, 2010).

A valuable benefit of SEM is that it explicitly takes measurement error into account, i.e. in how far do the observed variables fail to describe the latent constructs of interest.

However, in order to apply SEM appropriately, two crucial prerequisites should be fulfilled. First, SEM demands a sufficiently large sample size in order to estimate the model properly. This is the case because the -Test, which is used to decide either to accept or reject the model, significantly is influenced by sample size (n). The test is based on , where F is the fit function between the observed sample covariance matrix and the estimated covariance matrix. Therefore, if the sample size is extremely small, the model is always accepted. In return, if n is extremely large the model is always rejected (Blunch, 2008). Recommendations about specific sample sizes fall short because the absolute required sample size depends on several characteristics of the model and cannot be generalized. As a rule of thumb, Malhotra (2010) suggests sample sizes in the range of 200 and 400 if Maximum Likelihood Estimation is used.

Second, the collected data should show multivariate normality. However, this is an assumption seldomly true in practice. Therefore, as compensation if the data deviates from the assumption of multivariate normality the sample size in return should be even larger. Regarding the effects of multivariate nonnormality, Lei and Lomax (2005) found that it does not have a significant effect on the parameter estimates but that nonnormality inflates the statistics. This finding is in line with Henly (1993) who demonstrated that regardless of the sample size, the rejection frequency of a model is substantially higher if it is tested with a sample that contains nonnormal distributions of variables.

Taking these facts into account, the sample size for this study was planned to amount to at least 100 respondents for the exploratory factor analyses and the reliability tests in the pre-test and at least 400 respondents for the final data analysis using structural equation modeling.

In the following, the different steps of the study will be described. The theory presented in chapter two serves as the conceptual foundation as visualized in Figure 3 and as summarized in the hypotheses. The analysis will follow the classic approach of conducting SEM as suggested by Malhotra (2010).

First, in section 3.1 the constructs of interest will be assigned scales. These scales afterwards are pre-tested with an independent sample in section 3.2 where they are refined and purified with the help of exploratory factor analyses and reliability tests in SPSS. Section 3.3 summarizes the findings from the pre-test and presents the necessary adjustments concerning the scales and the overall model. The refined scales then, in section 3.4, are transferred into a measurement model with the help of the software AMOS. An additional large sample collected via an online survey is used to test both the measurement model and the structural model. The measurement model links the observed variables to the unobservable, latent constructs and allows the assessment of construct validty. Only when the measurement model can be considered valid - implying sufficient model fit and reliability - the analysis can proceed and the structural model can be specified and tested.

For more detailed introductions to the principles of SEM the interested reader is referred to Malhotra (2010), Blunch (2008) and Byrne (2001).

3.1 Assigning Scales to the Individual Constructs

In order to measure the different constructs involved in the model, the measurement instruments need to be specified. As described already in the previous chapter, existing scales for customer experience quality fail to acknowledge the full spectrum of the construct. Therefore a new scale, the CXQ scale, will be developed which aims to overcome the pitfalls of its predecessors. In order to generate a battery of items for this new scale, related scales were used as inspiration. The CXQ scale and its proposed items can be found in Table A1.

Perceived wealth requires a new scale as well because, to the knowledge of the author, it has not been measured systematically before. The items of the perceived wealth scales were entirely generated on the basis of common sense and can be found in Table A2.

To measure perceived value and loyalty intentions, already established scales were adapted to the case at hand. The perceived value scale draws all its items from the perceived value indicators scale presented by Dodds et al. (1991). The wording of the items was changed into a first person perspective to make it consistent with the rest of the questionnaire. The scale and its items can be seen in Table A3. The loyalty intentions scale basically is the behavioral-intentions scale invented by Zeithaml et al. (1996), reduced to the two dimensions of loyalty (=intention to recommend the company to others) and pay more (=willingness to pay more). It can be found in Table A4.

All items are measured on 7 point Likert scales. An exploratory factor analysis (EFA) and a reliability test for each scale was conducted in order to 1) refine the new scales and 2) make sure the changes and reductions of the established scales had no negative effects on the usability of the scales.

3.2 Pre-Testing the Scales

A multistage development study was applied on the basis of the scale development paradigm outlined by Churchill (1979). In order to run the analyses, data was collected via an online survey. The data collection and the scale purification process are described in the following.

3.2.1 Data Collection

The data for the pre-test was collected using an online survey. The sample of in total 138 respondents was collected based on convenience by publishing the survey link on the author’s facebook account. Only cases in which respondents indicated that they already have shopped at Starbucks were analyzed. After this pre-selection the final sample consisted of 123 respondents. Inspection of the 5% trimmed means for each variable showed that they did not significantly differ from the regular means (max. ±.26) and the deletion of outliers was accordingly neglected. Ages ranged from 18 to 42 years (M =22.59, SD =3.07) and the majority was German (70.7%), others were Dutch (5.7%) or of any other nationality (23.6%). 66.7% of the respondents were female, 33.3% were male.

3.2.2 Scale purification process

Following the recommendations of Churchill (1979), an iterative scale purification procedure was applied. An initial exploratory factor analysis (Principal Axis Factoring with oblique rotation and factor extraction based on Eigenvalues >1) was conducted to reveal the underlying pattern of factors. In the following, Bartlett’s tests of sphericity were always significant (p<.05) and the Kaiser-Meyer-Olkin Measure of Sampling Adequacy was always >.6, indicating that the data was suitable for factor analyses.

As a second analysis tool, item-to-total correlations were taken into account. This is a common procedure when developing a scale (Kim et al., 2012). Internal consistency reliability in the form of Cronbach’s alpha was considered as a third criterion to evaluate the usability of a scale. CXQ scale

An initial exploratory factor analysis (EFA) with oblique rotation extracted six factors based on Eigenvalues >1. Some problems surfaced when analyzing this pattern matrix. The data shows an abnormal factor loading of CON1 higher than 1. In an EFA with orthogonal rotation this would indicate a severe error; however, in EFA with oblique rotation in which factor loadings are interpreted as regression coefficients instead of correlation coefficients, this is still unusual but it does not indicate a methodological error; rather it merely shows a high degree of multicollinearity and the item can be considered as highly reliable (Jöreskog, 1999). This high degree of multicollinearity in the data can be assumed to be the result of the following characteristic: since the CXQ scale measures the latent construct ‘customer experience quality’ with the help of several latent sub-constructs, a higher degree of multicollinearity than in usual studies with more separated factors can be assumed. This is also the reason why oblique rotation was applied.

Several items load on more than one factor, violating the assumption of unidimensionality. Hence, the items SUR1, SUR2, IMG3, AFF1, AFF2, and AFF3 are eliminated. The items OP2 and OP3 have no significant loadings on either factor and are also eliminated.

An additional factor analysis was run with the remaining items. Now only five factors were extracted. Due to the reduction from six to five factors, IMG2 loads on more than one factor and was eliminated.

A third factor analysis lead to a new pattern matrix with 15 items unidimensionally loading on five factors. Although the pattern matrix looks very promising, this scale cannot serve as the final one. The item-to-total correlations were computed for the remaining 15 items. Items which were poorly correlated (r < .4) to the total score were excluded; these were: COG2, PQ1 and PQ2.

A new factor analysis with the remaining 12 items was run. All items unidimensionally load on four different factors and the item-total statistics demonstrate that all the remaing items significantly correlate (r > .4) with the total score, which is assumed to be customer experience quality (Table 3).

Table 3.Results of Factor and Reliability Analysis for the CXQ scale

Abbildung in dieser Leseprobe nicht enthalten



Type of Edition
ISBN (eBook)
ISBN (Book)
File size
1.7 MB
Catalog Number
Customer Experience Starbucks Structural Equation Modeling Marketing Management


Daniel Gurski, born in 1988, graduated at Maastricht University as Master of Science in International Business with a focus on Strategic Marketing in 2013. He is fascinated by Customer-Centric Service Science and always strives to develop new business strategies that make customers more satisfied and companies more profitable. Daniel hereby tries to bridge the gap between strategy and operations by analyzing business processes from a customer perspective. He worked for global organizations in Germany, the Netherlands and Australia.

Title: Customer Experiences affect Customer Loyalty: An Empirical Investigation of the Starbucks Experience using Structural Equation Modeling