Customer Satisfaction Measurement on the Internet

Table of Contents

Abstract

Table of Contents

List of Abbreviations

List of Figures

List of Tables

1. Introduction
1.1 Definition of Subject Matter and Problem Statement
1.2 Subquestions
1.3 Intended Academic and Practical Contribution
1.4 Chapter Conclusion

2. Customer Satisfaction Models
2.1 The Expectancy Disconfirmation Model
2.2 The Equity Theory
2.3 Ongoing Research
2.4 Chapter Conclusion

3. Customer Satisfaction Measurement
3.1 Measurement Approaches
3.1.1 Event-specific Methods
3.1.2 Attribute-specific Methods
3.1.3 Methods to assess Important Product Attributes
3.2 The Research Process
3.3 Chapter Conclusion

4. The Internet
4.1 Internet Services
4.2 Internet Research Methods
4.2.1 E-mail Surveys
4.2.1.1 Advantage and Disadvantages
4.2.2 WWW-Surveys
4.2.2.1 Advantages and Disadvantages
4.2.3 Online Focus Groups
4.2.3.1 Advantages and Disadvantages
4.3 Internet-specific Target Groups
4.3.1 Newsgroups and Mailinglists
4.3.1.1 Advantages and Disadvantages
4.3.2 Virtual Communities
4.3.2.1 (Potential) Advantages and Disadvantages
4.4 Chapter Conclusion

5. Customer Satisfaction Measurement on the Internet
5.1 Online Measurement of Derived Satisfaction
5.1.1 E-mail Surveys
5.1.2 WWW-Surveys
5.1.3 Comparability/Combination of Methods
5.2 The Online Measurement of Exceptional Experiences
5.3 Online Methods to Assess Important Product Attributes
5.3.1 In-Depth Interviews
5.4 ‘CS’ Research using Internet-specific Target Groups
5.4.1 Virtual Communities
5.5 The Online Research Process
5.5.1 The Decision-Making Framework
5.6 Chapter Conclusion

6. Online Customer Satisfaction Surveys In Practice
6.1 Methodological Considerations
6.1.1. Definition of the Population
6.1.2 Data Collection Method and Questionnaire Design
6.1.3 Sampling Frame
6.1.4 Sampling Procedure and Sampling Elements
6.1.5 Data Collection
6.2 Survey Results
6.2.1 Demographics
6.2.2 E-mail Surveys
6.2.2.1 Differences between Countries
6.2.2.2 Differences between Firm Sizes
6.2.3 WWW-Surveys
6.2.3.1 Differences between Countries
6.2.3.2 Differences between Firm Sizes
6.2.4 Online Methods to Assess Important Product Attributes
6.2.4.1 Differences between Countries
6.2.4.2 Differences between Firm Sizes
6.3 Chapter Conclusion

7. Conclusion

References

Appendix

The Questionnaire

The Reminder

Abstract

Based on the Expectancy Disconfirmation Model as the underlying construct, methods to measure customer satisfaction with products and the steps to be undertaken in the research process are investigated. The measurement of Derived Satisfaction using (dis)confirmation was found to be the appropriate approach to ‘CS’ measurement. The Critical Incidents Technique can be used to assess the influence of exceptional experiences and the customers’ evaluation of value-added services. The method of Focus Group interviewing is most appropriate for the exploration of important product attributes with customers while In-Depth Interviews structure salespeople- and executive interviews. It was also shown that during the research process, several points specific to ‘CS’ measurements need to be accounted for.

The Internet services currently used by marketing and social researchers include E-mail, mailinglists, newsgroups, Internet Chat, the World Wide Web (WWW) and Virtual Worlds. Virtual Worlds, being most advantageous for observational studies, are not useful for customer satisfaction research. Virtual Communities, in turn, have some promising characteristics for future use. Internet research methods based on these services include E-mail surveys, WWW-surveys and online Focus Groups. Common advantages of E-mail- and WWW-surveys include administrative and response speed, cost savings and global reach of respondents. Their greatest common disadvantage is the non-representativeness of the respondents for the larger population as well as their self-selection. Unless access is restricted to a known population, probability sampling is impossible when using the World Wide Web. Because of its serious disadvantages, the use of online Focus Groups is not (yet) advisable.

Based on these insights, the Internet was found to be an advantageous medium for customer satisfaction studies only if specific conditions are met. Companies need to investigate on a case-by-case basis if the online measurement of customer satisfaction is possible in their specific situation. The recommendations that were made are summarized in the online research process and the decision-making framework.

The results of a survey among market research agencies show that practitioners are to a large extent aware of the limitations within which the Internet can be used for customer satisfaction surveys. However, especially WWW-surveys sometimes are conducted in a way that does not lead to representative results.

List of Abbreviations

illustration not visible in this excerpt

List of Figures

Abbildung 1 The Expectancy Disconfirmation Model

Abbildung 2 Methods of ‘CS’ Measurement

Abbildung 3 The ‘CS’ Research Process

Abbildung 4 Example of an E-mail Survey

Abbildung 5 Example of a WWW-Survey

Abbildung 6 Example of an Online Chatroom

Abbildung 7 The Online ‘CS’ Research Process

Abbildung 8 The Decision-Making Framework

Abbildung 9 Responses by Type of Survey

Abbildung 10 Respondents per Country

Abbildung 11 Number of Employees per Company

Abbildung 12 Postings in Newsgroups/on Mailinglists

Abbildung 13WWW-Surveys: Usage of a Sampling Frame

Abbildung 14 WWW-Surveys: Quota Sampling

Abbildung 15 The Interviewing Method

List of Tables

Tabelle 1 Advantages and Disadvantages of Internet Research Methods

Tabelle 2 Summary of the Recommendations

Tabelle 3 E-mail Surveys in Practice

Tabelle 4 E-mail Surveys by Country

Tabelle 5 E-mail Surveys by Firm Size

Tabelle 6 WWW-Surveys in Practice

Tabelle 7 WWW- Surveys by Country

Tabelle 8 WWW- Surveys by Firm Size

Tabelle 9 Usage of Online Focus Groups

Tabelle 10 Online Focus Groups: Country Differences

Tabelle 11 Online Focus Groups: Firm Size Differences

1. Introduction

This introduction gives an overview over the subject and contents of this thesis. Firstly, a definition of the subject matter is made, followed by an introduction of the problem statement. Subsequently, the subquestions necessary for the answering of the problem statement and the chapters in which they will be dealt with are introduced. Next, the intended academic and practical contribution are explained. The aim and organization of the thesis are shortly summarized in a chapter conclusion.

1.1 Definition of Subject Matter and Problem Statement

Within this thesis, two popular subjects are combined: Customer satisfaction measurements and marketing research on the Internet.

The importance of customer satisfaction has gained considerable intention in the marketing literature. “As the cornerstone of the marketing concept, customer satisfaction has been embraced by practitioners and academics alike as the highest-order goal of the company” (Peterson & Wilson, 1992, p. 61). With increasing global competition, accelerating customer choice opportunities and demands, customer satisfaction has become a vital goal for the survival of the company. Individual countries as well as the European Union have recognized this importance and started to develop their own customer satisfaction indices in order to provide their companies with a standard benchmark within their industry and country (Bruhn, 1998).

Since its opening to private and commercial use in 1995, the Internet has been growing tremendously (Agrawal 1998). Because of this growth, the new medium has also gained the interest of (marketing) researchers. Coomber (1997) summarizes that “the existence of the Internet and the World Wide Web (WWW) clearly provides new horizons for the researcher. A potentially vast population of all kinds of individuals and groups may be more easily reached than ever before, across geographical borders and even continents”.

While Internet marketing research is gaining popularity and studies concerning this subject are finally emerging on a larger scale, no specific investigations of customer satisfaction measurements on the Internet could be found. The problem statement to be answered within this thesis therefore is the following:

How can customer satisfaction measurements be realized on the Internet?

In order to find an answer to this question, a restriction has to be made because the satisfaction formation appears to be different for products and services. Due to the more advanced findings concerning the satisfaction formation with products as opposed to services, the satisfaction with products is investigated within this thesis.

1.2 Subquestions

Several aspects need to be investigated in order to find an answer to the problem statement in section 1.1. Subquestions 1 to 3 deal with the subject of customer satisfaction and its measurement while subquestions 4 and 5 concern the new medium ‘Internet’ and its current use in marketing research. Lastly, the two subjects are combined by finding answers to subquestions 6 and 7. Below, the questions are presented in the order in which they will be answered.

1. Which theoretical construct models customer satisfaction ?
2. Which approaches exist to measure customer satisfaction?
3. Which research steps have to be taken during the measurement process?
4. What is the Internet?
5. Which marketing research methods are already applied on the Internet?
6. Which of these methods are suitable for a measurement of customer satisfaction on the Internet?
7. Is there currently a discrepancy between theory and practice?

In order to frame the problem and provide a reference point to the discussion, subquestion 1 will be answered in chapter 2. Based on this definition, the relevant measurement approaches (subquestion 2) will be introduced and evaluated in chapter 3. Subsequently, the research steps that need to be followed will be investigated (subquestion 3).

After the first part of this thesis has then been clarified, a definition of the Internet (subquestion 4) will be given in chapter 4 to frame the second part of the discussion. After the relevant Internet services have been introduced, subquestion 5, investigating the marketing research methods already being applied on the Internet, is answered. Based on the accumulated answers, the question of how customers satisfaction measurements can be conducted on the Internet (subquestion 6) is dealt with in chapter 5. Lastly, a survey among marketing research companies in Australia, Canada, the UK and the US has been conducted in order to find out weather a discrepancy between the propositions made in chapter 5 and the practice of companies exists. It could furthermore be investigated if more developments take place in practice which demand the ‘backup’ of academic research findings. The results are presented in chapter 6.

1.3 Intended Academic and Practical Contribution

While some literature on marketing research on the Internet can be found, academic research does not yet quite catch up with an investigation of its applications. Research concerning WWW-Surveys and E-mail surveys has been scattered and oftentimes originated in the field of social sciences. Academic studies concerning online Focus Groups could not be found at all, while the more general studies regarding computer-administered communication also frequently originated in the field of social sciences.

The intended academic contribution of this study is therefore to accumulate these scattered findings in a first step in order to provide an overview over the current state of knowledge. Secondly, these findings have been combined with the current knowledge in the field of customer satisfaction. Based on these insights, recommendations for online customer satisfaction measurement have been developed. Schillewaert et al. (1998, p. 320) state that “future studies should be aimed at developing a comprehensive framework for describing when to use and when not to use the various sampling methods for WWW surveys”. This study provides a framework not only for WWW-studies but for Internet studies in the context of customer satisfaction. The recommendations given can furthermore be used as reference points for future studies.

Bandilla and Hauptmanns (1998) furthermore state that it is of vital importance for research in marketing and the social field to learn the usage of the new medium ‘Internet’. According to the authors, its chances and restrictions need to be acknowledged early, so that the knowledge about how to conduct methodologically correct studies on the Internet has accumulated once penetration rates are comparable to those of the telephone. Buchanan and Smith (1997, p. 141) add to that point by stating the following: “What makes investigations of its [Internet-mediated research] validity crucial .. is the fact(s) that such research is currently being done”.

The intended practical contribution of this thesis is to provide marketing researchers with a theoretical framework to guide their research efforts in order to make their studies as methodological sound as possible. The survey conducted has shown that customer satisfaction surveys are already conducted on the Internet. This thesis can be used as a guideline for all companies considering the measurement of their customers’ satisfaction on the Internet.

1.4 Chapter Conclusion

The importance of customer satisfaction has long been recognized by companies. Because of its tremendous growth within the past 4 years, the Internet has been increasingly interesting for marketing researchers. While a small body of research has been amounted concerning (marketing) research on the Internet in general, no investigations of online customer satisfaction could be found. This thesis therefore investigates how customer satisfaction measurements can be realized online.

After the concept of customer satisfaction and its measurement have been clarified, the Internet with its services and the current marketing research methods based on them are introduced. Based on these insights, the question of how customer satisfaction measurement can be realized online is answered. A survey among marketing research companies aims to bridge the gap between theory and practice and attempts to identify developments in practice which need the ‘backup’ of academic research.

2. Customer Satisfaction Models

The construct of customer satisfaction (‘CS’) has been researched extensively during the past decades. However, as of today, no generally accepted model has emerged (Johnson et al., 1995, Berger & van Mens, 1997, Kaapke & Hudetz, 1998).

The following chapter gives an overview over the two process theories currently discussed in theoretical satisfaction research: The Expectancy Disconfirmation Model and the Equity Theory (Müller, 1998). Despite considerable agreement on these two process models, research over some (additional) model constructs and their relation is still going on. Subsequently, an overview over continuing research is given. The chapter closes with a short summary of the theories discussed.

2.1 The Expectancy Disconfirmation Model

Richard Oliver led customer satisfaction research with the Expectancy Disconfirmation Model. This model has consistently been validated in empirical research (Engel et al., 1995 Peter & Olson, 1996) as well as build upon by various scholars (e.g. Tse & Wilton, 1988, Oliver & DeSarbo, 1988, Halstead et al., 1994, Spreng et al., 1996, Gupta & Stewart 1996, Wilton, 1997, Patterson et al., 1997).

According to this model, a customer’s satisfaction has three antecedents: Pre-purchase expectations, perceived product performance and confirmation or disconfirmation of these expectations. While the role of affect has not yet been resolved clearly, there is consensus over the existence of an emotional reaction to the intensity of satisfaction experienced (Müller, 1998). As of today, this model has been dominant in theoretical ‘CS’ research (Müller, 1998, Berger & van Mens, 1997, Gupta & Stewart, 1996, Engel et al., 1995, Boulding et al., 1993).

Expectations have been defined differently by various researchers. Tse et al. (1988), treat expectations as the likelihood of an event as well as an evaluation of its goodness or badness. Müller (1998) summarizes the following expectation-concepts that can be found in the ‘CS’-literature: Expectations as ideal product performance expectations, minimal expectations, and productspecific norms. According to the current literature (Müller, 1998, Klingebiel, 1998, Berger & van Mens, 1997), expected product performance defined as a product’s most likely performance (‘predictive expectations’) is the most common presumption used in customer satisfaction research. Engel et al. (1995, p. 275) support this statement with the motivation that “this is the logical outcome of the pre-purchase alternative evaluation process”. In correspondence with these authors, expectations will be treated here as ‘predictive expectations’. As will be explained in section 2.3, expectations are furthermore assumed to be growing over time.

Perceived performance expresses the performance of the various product attributes as recognized by the customer. There is general consensus that expectations as well as perceived performance are not formed on an aggregate product level but for each product attribute separately (Oliver, 1993). Halstead at al. (1994) state that this separate recognition allows for the assessment of the contribution each attribute makes to the overall satisfaction judgment.

According to Engel et al. (1996), (dis)confirmation is the result of a comparison of expectations against perceived performance. Consumers make these comparisons using better-than, worse-than heuristics (Oliver, 1997). Positive disconfirmation occurs whenever a consumer’s perceived performance exceeds his^[1] expectations, resulting in some degree of satisfaction. Negative Disconfirmation occurs when expectations exceed product performance, resulting in dissatisfaction. The intensity of the (dis)satisfaction experienced by the consumer is related to the intensity of the experienced (dis)confirmation (Patterson et al., 1997). Finally, under confirmation performance equals expectations, also resulting in satisfaction (Peter & Olson, 1996). However, this can be regarded as a more neutral stance, not being very extreme (Engel et al., 1995).

The degree of satisfaction/dissatisfaction experienced by the customer in turn triggers an emotional reaction on his part as a result of the unexpected discrepancy between expectation and perceptions (Müller, 1998). This affective reaction then influences the customers’ repurchase intentions, complaint behavior and word-of-mouth communications (Patterson et al., 1997, Gupta & Stewart, 1996, Peter & Olson, 1996).

According to De Ruyter et al. (1997), a growing number of studies have also shown a direct influence of product performance on customer satisfaction. However, Halstead et al. (1994) provide an overview over studies showing a wide disparity of findings. In agreement with more current findings (De Ruyter et al., 1997, Berger & van Mens, 1997, Oliver, 1993), perceived performance will be treated here as exerting both, a direct and an indirect influence on satisfaction via (dis)confirmation. Figure 2.1 gives an overview over the relationships described.

illustration not visible in this excerpt

Abbildung 1 The Expectancy Disconfirmation Model

Source: Adapted from Müller (1998),

Berger & van Mens (1997).

Although the Expectancy Disconfirmation Model in ‘CS’ research is oftentimes taken as the underlying construct for products and services, for reasons that will be given in section 2.3 it is regarded here as being valid only for products^[2].

2.2 The Equity Theory

Compared with the Expectancy Disconfirmation Model, this theory has gained less attention within customer satisfaction research (Müller, 1998). One of the first researchers to propose this model within the customer satisfaction context were Huppertz, Arenson and Evans (1978). Continuing research related to this model has mainly been undertaken in ‘CS’ research related to services (e.g. Patterson et al., 1997). This can be explained by the fact that, due to the simultaneousness of production and consumption, it is thought to be easier for service customers to evaluate the input/output ratio of the interaction as it is for consumers of products (Müller, 1998).

The Equity Theory is based on the presumption that in every social interaction, each partner has to make contributions in the form of inputs and in turn faces outcomes of the transaction. The input/outcome combination is then rated according to its fairness as it is perceived by the customer (Huppertz et al., 1978). A customer will feel as having been treated fairly whenever his input/output ratio is proportional to that of the exchange partner. Specifically, high satisfaction is predicted to occur whenever high inputs result in high outcomes compared to the input/performance ratio of the exchange partner. Low satisfaction will result when outcomes decrease proportionally to those of the interaction partner (Oliver & De Sarbo, 1988). Experienced dissatisfaction further triggers negative emotions which rise in tandem with the increase in perceived inequity (Müller, 1998). In contrast to the Expectancy Disconfirmation Model, perceptions are only found after the purchase has already taken place (Berger & van Mens, 1997).

Within in the Equity Theory, different exchange partners have been proposed as possible references for a customer making the comparison. According to Fisk and Young (1985, in: Müller, 1998), customers compare their input/output ratio to that of another, specified customer. Oliver and Swan (1989, in: Müller, 1998) do regard the partner as the one actually participating in the interactive process, but supplement the model with the concept of disconfirmed realistic expectations which directly influence expectations.

Two main problems hamper the use of the equity theory as an underlying model for customer satisfaction research. Firstly, the quantification of the inputs as well as the outputs in a given transaction is rather difficult (Müller, 1998). This is due to the fact that the customer is more often than not not knowledgeable of the inputs and outputs of the provider. Related to this problem is the question whether customers can construct fairness judgments at all based on this incomplete information. Müller (1998) concludes that for complex exchange evaluations (as the evaluation of a product with multiple attributes), the Equity Theory does not provide enough explanation.

2.3 Ongoing Research

As mentioned in the opening paragraph, the debate over the ‘CS’ construct is far from being resolved. This section gives an overview over the most important issues currently being discussed.

Customer Satisfaction with Services

Research on the different determinants on customer satisfaction has found mixed results depending on the type of product that had been tested. As early as 1988, Oliver and De Sarbo noted that “different product categories could have unique responding techniques” (p. 505). Halstead et al. (1994) agree in that they state that the manner in which consumers form satisfaction judgments varies between product categories. They note that “the services literature is rich with evidence on how evaluation processes differ between products and services” (p. 118). The evaluation process for services is perceived as being more difficult (compared to products), based on different types and sources of expectations as well as based on the evaluation of the process as well as an outcome (Halstead et al., 1994). Specifically, prevailing in the service satisfaction literature is the dominant role of performance in service evaluation (De Ruyter et al., 1997, Boulding et al., 1993). While service quality and customer satisfaction in the past have been used interchangeably in the service literature, there is growing evidence of a conceptual difference between the two concepts (De Ruyter et al., 1997, Oliver, 1993). However, the position of ‘service quality’ within the model is still unresolved.

The Influence of Experience

Related to the discussion above, Halstead et al. (1994, p. 126) state that “the inclusion of both performance and disconfirmation variables may be unnecessary depending on the product category and/or the specific attributes customers use to evaluate products and services”. They go further in explaining that in the case of customers’ ability to form concrete pre-purchase expectations (e.g. with prior experience), mainly disconfirmation effects may be found. However, when the formation of expectations cannot take place (e.g. due to a lack of product familiarity), only performance evaluation may occur. Johnson & Fornell (1991) agree that the influence of perceived product performance and expectations varies according to the consumers’ experience. That state that eventually, with extreme customer experience, “current performance and expectations may coincide” (Johnson & Fornell 1991, p. 278). Unfortunately, the influence of experiences is not yet clarified. Nevertheless, it can already be acknowledged that customer satisfaction and its antecedents are not static but are likely to change over time.

Related to the above discussion, one word of caution seems to be imperative: It is often believed that customer satisfaction directly leads to loyalty. However, one should be aware that customer satisfaction is an important but not the only factor influencing loyalty. According to Berger & van Mens (1997) other factors influencing loyalty include reputation, switching costs, choice alternatives, price, the relationship between the customer and the company, and the customer’s freedom of choice.

2.4 Chapter Conclusion

This chapter gave an overview over the customer staisfaction models currently discussed. Because research has not yet resulted in a generally accepted definition of the ‘CS’ construct (Bruhn, 1997), an overview over ongoing research was also given. It was specifically recognized that the ‘CS’ construct seems to be different for services and products. Moreover, the role of expectations has yet to be clarified.

Based on these insights, I find it most meaningful to take the Expectancy Disconfirmation Model as the underlying construct for customer satisfaction measurement used in this thesis. Especially, the shortcomings of the Equity Theory, the prevalence of the Expectancy Disconfirmation Model in ‘CS’ research and the lack of empirical evidence in ongoing research have led me to this conclusion. Furthermore, drawing from ongoing research, a certain (yet to be specified) dynamic nature of customer satisfaction formation is also acknowledged. As noted in section 2.3, ‘CS’ formation with services seems to be different from the process described in the Expectancy Disconfirmation Model. Therefore, the following chapters with their resulting conclusions only hold for the satisfaction measurement relating to products.

While customer satisfaction as a latent construct is neither directly measurable nor observable, the next chapter will deal with the issue of it indirect measurement via multiple proxies or indicators. Lingenfelder & Schneider (1991) note that the usefulness of a hypothetical construct for applied research depends decisively on the ease/difficulty of its operationalization.

3. Customer Satisfaction Measurement

After having agreed on an underlying theoretical model to be used within this thesis, the process of customer satisfaction measurement will now be described. One problem with customer satisfaction measurement is the fact that in practice, a large amount of different approaches exist (Klingebiel, 1998, Ramos, 1996) of which a considerable number is not based on any theoretical foundation at all (Peterson & Wilson, 1992). This can partly be attributed to the complexity of the (potential) concepts related to the construct of customer satisfaction. However, the success of any research for a large part depends on its theoretical foundations because “theory .. summarizes what is known about an object of study and states the uniformities that lie beyond the immediate observation ..” (Cooper & Emory, 1995, p. 43). Although the Expectancy Disconfirmation Model might not capture all the antecedents of customer satisfaction, it is based on extensive research and empirical validation and therefore superior to a purely intuitive approach.

This chapter firstly introduces different measurement approaches of customer satisfaction and evaluates them from a theoretical viewpoint. Reference point throughout the whole chapter will be the Expectancy Disconfirmation Model that had been accepted as being valid within this thesis in Chapter 2. Subsequently, the sequence of steps that have to be taken during the research process will be explained. The chapter closes with a summary of the conclusions drawn.

3.1 Measurement Approaches

Because its aim is to investigate a precisely specified problem and the statement of the degree to which customer satisfaction is present, a ‘CS’ investigation is a descriptive research (Churchill, 1995). Repeated measures (longitudinal analysis) on a regular basis are regarded as absolutely necessary because of the dynamic nature of customer demands and market characteristics as well as the company’s need to track the progress of its improvement actions (Homburg & Werner, 1996, Töpfer, 1996). The time frame chosen for repeat measurements will be determined by the frequency of repeat purchases and the substitution possibilities of the customer, as well as market dynamics, and the speed at which improvements of weak points can be conducted by the firm (Aaker et al., 1998, Töpfer, 1996).

In the literature, two methods of measuring customer satisfaction can be found: Objective methods and subjective methods.

Objective methods measure observable quantities that are independent of the investigator’s interpretation. Approaches include the analysis of sales turnover, market share, the degree of customer migration and the repurchase rate. However, these methods have two serious drawbacks: Firstly, their relation to customer satisfaction is not clarified theoretically; they are not included in the Expectancy Disconfirmation Model. Secondly, and related to the first point, it is also accepted throughout the literature that these methods are lacking validity (Töpfer, 1996, Lingenfelder & Schneider, 1991). Based on these insights, objective methods are not regarded as appropriate measurements within this thesis.

Subjective methods make use of a pre-defined construct of customer satisfaction and attempt to measure it via indicators (Lingenfelder & Schneider, 1991). These methods can further be classified into attribute-specific methods and event-specific methods.

Figure 3.1 gives an overview over the most common approaches. Please note that within the subjective methods, only approaches relating to product satisfaction are taken into consideration.

illustration not visible in this excerpt

Abbildung 2 Methods of ‘CS’ Measurement

Source: Adapted from Klingebiel (1997).

3.1.1 Event-specific Methods

Event-specific methods rest on the assumption that a customer’s satisfaction is largely dependent on the incidents he experienced with the company. These methods are based on so-called ‘story telling’ whereby customers are asked to report their experiences with the company/product in question in an unstructured way (Bruhn, 1997). The timing of the investigation is critical because customers must have been able to form an evaluation over the product as well as being able to remember that incident in detail (Eversheim, 1996, Töpfer, 1997). It should be noted that event-specific methods are very unlikely to generate a complete picture of the customers’ satisfaction (Homburg & Werner, 1996, Töpfer, 1996). While most of the methods reported in the literature relate to services only, the Critical Incidents Technique can be used for ‘CS’ measurement with products also.

The Critical Incidents Technique only focuses on occasions where customers made exceptional, non-routine experiences (Stauss, 1995, in: Bruhn, 1997). Because the experiences under investigation are non-routine, they are believed to be stored in memory for a long time (Eversheim, 1997). Hayes (1992) adds to that point by stating that a critical incident is always specific to one single behavior or product characteristic. Customers are questioned via open-ended, standardized questions which facilitate the recall of these special incidents (Bruhn, 1997). Although the influence of critical incidents is not modeled within the Expectancy Disconfirmation Model, it was acknowledged in section 2.3 that experience is likely to have some influence on customer satisfaction while the specific nature of this influence had not been fully clarified. Therefore, the Critical Incidents Technique can be used to complement measurements based on the Expectancy Disconfirmation Model. Moreover, related to the definition of products and services given in Chapter 2, Bruhn (1997) acknowledges that manufacturing or trading companies also engage in value-added functional services in order to support their marketing and sales. The Critical Incidents Technique can also be used to see if exceptional influences have been made with the value-added services of the company and if this aspect needs more attention.

3.1.2 Attribute-specific Methods

Attribute-specific methods are based on the assumption that the customer forms his product evaluation via individual product attributes (Eversheim, 1997). This view is in conformance with the Expectancy Disconfirmation Model. Attribute-specific methods are especially suitable for standardized, timely and cost-effective measurements of features which are usually expected by the customer (routine attributes) (Eversheim, 1997). These methods can further be sub-classified into direct or indirect measurements.

Direct measurements approach satisfaction or its components in a straightforward way. Methods include the measurement of product performance, product performance and its importance, (dis)confirmation, directly reported satisfaction and the decompositional method (Bruhn, 1997, Kotler, 1994, Lingenfelder & Schneider, 1991).

As its name implies, the analysis of product performance only takes the influence of performance but not that of (dis)confirmation into account and is therefore incomplete from a theoretical viewpoint.

Accordingly, the measurement of performance and its importance suffers from the same drawback. Moreover, there are some problems with the direct questioning of attribute importance that will be discussed in section 3.2.

The measurement of (dis)confirmation in turn does not account for a separate influence of performance as modeled in the Expectancy Disconfirmation Model and is therefore inferior in situations where performance has a greater influence than expectations.

In the case of directly reported satisfaction, overall or multiattribute satisfaction is measured via satisfaction scales. This approach shortens the measurement process because expectations and performance do not need to be measured separately. On the other hand, the disentanglement of the separate influences of (dis)confirmation and performance is inhibited which reduces its information content (Bruhn, 1997). Consequently, the separate influence of performance again cannot be accounted for.

For reasons given in the next section, the measurement of overall satisfaction is a special case and should always be included in the measurement instrument.

Making use of the decompositional method, customers have to rate their satisfaction with different sets of pre-specified combinations of product attributes. Attribute combinations are constructed in a way that they differ at the individual attribute level between groupings. Via a decompositional statistical analysis (e.g. Conjoint Analysis) the relative importance of the different attributes is assessed after the questioning has been conducted.

It can be criticized that this method does not allow for the separate assessment of (dis)confirmation and performance. Furthermore, by setting the individual product attributes at different levels, customers are expected to express their satisfaction with a product performance that they did not experience. According to the Expectancy Disconfirmation Model, this is not possible.

Indirect measurements do not measure satisfaction directly but only its antecedents (derived satisfaction) or infer from its consequences (complaint-analysis).

Derived Satisfaction measures the degree to which a certain attribute was expected as well as experienced (Bruhn, 1997, Kotler, 1994). This method acknowledges both influences on customer satisfaction and therefore also allows for the measurement of the separate influence of performance. However, according to the Expectancy Disconfirmation Model, expectations as an indirect influence on customer satisfaction are completely mediated through (dis)confirmation. Another option therefore is to measure the degree of (dis)confirmation as well as performance (Klingebiel, 1998, Eversheim, 1997, Lingenfelder & Schneider, 1991). It follows that this approach best presents the Expectancy Disconfirmation Model because it accounts for the direct influence of (dis)confirmation and the possible separate influence of performance on satisfaction. Furthermore, as will be explained in section 3.2, if supplemented with an evaluation of overall satisfaction, this method allows for the assessment of the importance of individual attributes after the data have been collected.

With regard to complaint analysis, it should be noted that in general, only about 5% of unsatisfied customers ever complain (Aaker et al., 1998, Eversheim, 1997, Kotler, 1994), which severely limits the method in is predictive ability.

3.1.3 Methods to assess Important Product Attributes

As will be explained in the next section, a vital stage of the ‘CS’ research process is the determination of those product attributes that the customers perceive as most important. Methods to assess attribute importance include the Critical Incidents Technique, In-Depth Interviews and Focus Groups.

Bruhn (1997) states that the Critical Incidents Technique can also be used for the detection of important product attributes. However, its main disadvantage in this context lies in its inability to detect important product attributes with which the customer has not made exceptional experiences.

In-Depth Interviews are another way to explore the importance of product attributes. Interviews are conducted in an informal, unstructured way that allows the interviewer to probe into the respondent’s answers. As the expression ‘in-depth’ implies, the main advantage lies in the fact that each respondent has the opportunity to provide comprehensive and deep information. Accordingly, Dutka (1994,) classifies the method as being appropriate for capturing the ideas and viewpoints of key executives and salespersons within the organization. This method in turn is less appropriate for the questioning of customers because it is oftentimes stated that they do not mention attributes that are self-evident to them. Only the direct interaction in a focus group allows for the direct clarification of attribute importance for the larger group of customers.

Focus Groups are defined as “a research technique that collects data through group interaction on a topic determined by the researcher” (Morgan, 1997, p. 6). A group usually consists of 8-12 relatively homogeneous respondents to ensure optimal group interaction (Churchill, 1995). The moderator is responsible for leading the discussion and stimulating participant interaction. Used in the exploratory phase, the group discussion is relatively structured with a high level of moderator involvement because of its strong preexisting agenda (Morgan, 1997). A minimum of two group interviews are advisable to ensure that the outcome of one group is not subject to its composition (Morgan, 1997). The biggest advantage of the method is this group-dynamism coupled with cost and time-effectiveness. Morgan (1997, p. 15) states that “in an era where issues of consensus and diversity are of intense interest .. , the discussion in Focus Groups can provide direct data on these exact issues”. It follows from these arguments that focus group interviews are the most useful as well as practical method for customer interviewing in the exploratory phase of ‘CS’ research.

Concluding it can be said that a number of measurement approaches are available to conduct customer satisfaction measurements in practice. However, when investigated with reference to the Expectancy Disconfirmation Model, only two methods are regarded as appropriate: The Critical Incidents Technique and the measurement of Derived Satisfaction using (dis)confirmation. All of the other methods only cover either (dis)confirmation or performance or even measure staisfaction in a direct way.

Furthermore, for exploratory investigations, In-Depth Interviews are appropriate for salespeople and executive interviews while Focus Groups are advantageous for customer interviewing.

In the next section, the customer satisfaction research process will be described with reference to these four methods.

3.2 The Research Process

While the measurement of customer satisfaction follows the steps described in general marketing research, each of them requires actions specific to ‘CS’ research.

This section starts with an overview over the steps in ‘CS’ measurement. Subsequently, each of them will be discussed shortly. Whenever a difference in proceedings between the measurement of Derived Satisfaction and the Critical Incidents Technique exists, the methods will be discussed separately.

The customer satisfaction research process consists of seven steps. Figure 3.2 on the next page gives an overview over their sequence.

Step 1: Research Goal and Target Group

The first decision to be made is the definition of the research goal. This will usually include the determination of the current satisfaction level of customers, its monitoring over time, the detection of problem areas within the company and the control of program effectiveness (Hayes, 1995, Homburg & Rudolph, 1995). Subsequently, the target group needs to be defined. The totality of a firm’s current customers should only be targeted if the company’s profit is spread relatively even among them. However, it is oftentimes stated in the marketing literature that approximately 20% of a company’s customer account for 80% of profitable sales (Wayland & Cole, 1994, in: Klingebiel, 1998, Hanan & Karp, 1989). In that case, they should naturally be the customers of most interest. Furthermore, whenever possible, lost customers should be included in the investigation because they can be very helpful in detecting weak points (Homburg & Rudolph, 1995, Töpfer, 1996).

illustration not visible in this excerpt

Abbildung 3 The ‘CS’ Research Process

Source: Adapted from Homburg & Rudolph (1995), Churchill (1995).

Step 2: Explorative Investigation

If the ‘CS’ measurement is conducted for the first time, the product attributes that are most important to the customers need to be assessed in an explorative investigation. This stage is vital to the success of the whole investigation because the appropriateness of the attributes included strongly influences the validity of the results (Eversheim, 1997). Etter (1996, p. 4) states that “attribute importances are relatively stable over time .. [and] .. are unlikely to change appreciably over a span of several years”. In the same vein, Homburg and Werner (1996) propose that this stage should be repeated all 3 to 4 measurement cycles to capture new trends. Important product attributes can be detected via In- Depth Interviews with salespeople and executives and complemented with the investigation of industry-specific literature. Essential is the direct investigation of customers because managers and employees can never have a complete picture of what is important to the customer (Eversheim, 1997, Bruhn, 1997, Goodman et al., 1992). As explained in section 3.1.3, the most appropriate way to do this is via Focus Groups. The interviews are usually tape-recorded and completed with the researcher’s notes.

Step 3: Sampling Design and Data Collection Method

The appropriate sampling frame for a customer satisfaction measurement is the customer database. A census seems appropriate whenever the target group is relatively small, otherwise, the sample size needs to be decided upon^[3]. When measuring Derived Satisfaction, results are usually required to be representative for the larger population of customers. In that case, a probability sample needs to be drawn. If only general insights are wanted (Critical Incidents Technique), a non-probability sample is sufficient^[4]. However, if no sampling frame is available (e.g. for firms selling mass products), a probability sample cannot be drawn. Usually, a quota sample is then taken. Having information about the (demographic) composition of its customer base, the company exactly replicates this composition in the sample it draws. There are proponents and opponents to the idea whether this constitutes a representative sample or not^[5] (Ronig, 1998).

According to Dutka (1994, p. 61), “telephone interviews and mail questionnaires are the chief methods of collecting data for customer satisfaction research”. Unless the sample size is very small, personal interviews are very cost- and time-intensive. Another drawback of this method is the interviewer bias, which is less intense during telephone surveys. Telephone surveys permit superior quality control, elicit large response rates and fast turnaround times (time between data collection and return). Mail surveys, in turn, are superior when customers are difficult to reach; they allow the customer to choose his own responding time and are less expensive. Their greatest drawback is the low response rate that questions the representativeness of the returned questionnaires (Werner, 1997, Fowler, 1997, Dutka, 1994). Peterson & Wilson (1992, pp. 64-65) during their meta-analysis of customer satisfaction studies have found that “satisfaction data collected using different [data collection] modes are not comparable, on average, personal or telephone interviews appear to increase satisfaction ratings by approximately 10-12 percent relative to self-administration”. Based on this argument, the use of personal administration methods can be questioned.

During the last ten years, computer-assisted telephone-interviewing (CATI), computer-assisted personal interviewing (CAPI) and self-administered questionnaires have also been used in a market-research in context in Europe^[6] (Ronig, 1998). The feasibility of computer-assisted self-administred questionnaires for ‘CS’ research will receive considerable attention in the remainder of this thesis.

Step 3a: Questionnaire Design and Pretest

When measuring Derived Satisfaction, the development of a questionnaire is the next step to be conducted. The questionnaire should be standardized so that results are comparable between measurement cycles. Other issues in questionnaire development include questionnaire length, degree of disguise and the choice of scale with its reliability and validity^[7]. In addition to questions about product attributes and overall satisfaction, most questionnaires ask about demographic data for segmentation purposes, an open-ended question to capture information missed during the exploratory phase and word-of-mouth behavior (Goodman et al., 1997). Whenever possible, questions should be asked in a random order in longer questionnaires because of respondent fatigue (Dutka, 1994). Prior to data collection (step 4), a pretest needs to be undertaken to test the feasibility of the method chosen and the validity and reliability of the questionnaire.

Step 5: Data Analysis

Data analysis is very different for the Critical Incidents Technique, which follows a qualitative approach, and the measurement of Derived Satisfaction, which follows a quantitative approach.

Data gathered with the help of the Critical Incidents Technique can be analyzed manually or with the help of a computer program for quantitative analysis (Dutka, 1994). In any case, data are ‘coded’ by focusing on a common verb or adjective (Hayes, 1992). The resulting clusters are then named according to their underlying satisfaction item. If possible, the obtained satisfaction items are classified in a second step so that they form specific customer requirements. By having the process conducted by at least 2 people, the interjudge agreement^[8] can be calculated (Hayes, 1992).

In turn, the analysis of quantitative data (Derived Satisfaction Measurement) is more complicated. While response distributions to ‘CS’ measurements tend to be negatively skewed (mainly positive answers) (Froböse & Schmickler, 1998, Engelmann & Müller, 1997, Dutka, 1994, Peterson & Wilson, 1992), the normality of the data should always be checked before conducting any parametric tests.

Furthermore, satisfaction ratings might not be comparable between attributes (Froböse & Schmickler, 1998, Engelmann & Müller, 1997). It is therefore more sensible to compare the attributes to some benchmark as opposed to each other. The benchmarking measure mentioned most frequently in the literature is the competing firm (Rothenberger, 1997, Engelmann & Müller, 1997). In that case, this information has to be asked for separately on the questionnaire, which in turn increases its length. Results can also be compared to previous ‘CS’ measurements.

Lastly, the importance of individual product attributes is vital for management decisions. While it can be measured directly in the questionnaire (direct reporting), this approach suffers from low validity in that customers tend to rate everything as important (Zacharias, 1998, Werner, 1997, Grisaffe, 1993). Furthermore, it also contributes to questionnaire length. A more useful approach is the indirect assessment of attribute importance via regression- or Chaid-analysis (based on a chi-square test)^[9] after the data collection has taken place (Zacharias, 1998). Regression analysis can also be used to assess whether expectations about or performance of a product attribute contributes more to customer satisfaction, which can be important information for management decisions. Subsequently, the research report can be written (step 6). There it is important not to overwhelm the reader with methodological detail but present the results in an appealing way.

3.3 Chapter Conclusion

This chapter gave an overview over the methods used to measure customer satisfaction and the steps to be undertaken in the research process.

To conclude, the measurement of Derived Satisfaction using (dis)confirmation is the appropriate approach to ‘CS’ measurement within the context of the Expectancy Disconfirmation Model. The Critical Incidents Technique in turn can be used to cover the (yet) unclear influence of experiences and the customers’ evaluation of value-added services. All other methods will not be considered further within this thesis because they did not account for either the effect of (dis)confirmation or performance or measured satisfaction in a direct way.

Furthermore, the method of Focus Group interviewing was found to be most appropriate for the exploration of important product attributes with customers, while the In-Depth Interview is advantageous for salespeople- and executive-interviews. Lastly, it was shown that during the research process, several points specific to ‘CS’ measurements need to be accounted for.

[...]

---

^[1] For the ease of writing, only the male form will be used while all propositions are equally valid for males and females.

^[2] Note that there are rarely any ‘pure’ products. The classification as a product or service is made according to the degree of tangibility present where products tend to be more tangible than services. In this thesis, the following definition will be accepted: Services “include all economic activities whose output is not a physical product or construction, is generally consumed at the time it is produced, and provides added value in forms .. that are essentially intangible concerns for its first purchaser“ (Quinn et al., in: Zeithaml & Bitner, 1996, p. 5).

^[3] For a discussion on the appropriate sample size, see Churchill (1995).

^[4] Probability samples include simple random, (dis)proportionate stratified, and system or area cluster samples; nonprobability samples include convenience, judgment and quota samples.

^[5] See Ronig (1998) for a discussion on the issue.

^[6] For a complete discussion, see Ronig (1998).

^[7] Validity: Degree to which the scale measures what it is supposed to measure; Reliability: Similarity of results using independent comparable measures. For a complete discussion, see Hayes (1992).

^[8] The Interjudge Agreement is an Index ranging from 0 to 1 (perfect agreement). Usually, an index of approximately 0.8 is used a s the cutoff. For a detailed discussion, see Hayes (1992).

^[9] See Zacharias (1998) for a discussion.