An information threoretic approach to modeling customer satisfaction for low-tech industrial offerings.

ABSTRACT

Researchers have paid a great deal of attention to overall customer satisfaction for high-tech industrial offerings but little attention is paid to customer satisfaction for low-tech offerings. This is significant because for industrial buyers, repurchase intentions have been shown to be correlated with overall customer satisfaction.

Data collected from a cross section of industrial offerings are categorized according to their levels of technology. Optimal models are developed and the results suggest that overall satisfaction for low-tech offerings can best be considered a linear function of satisfaction with the vendor's performance. This optimal model includes no other component of satisfaction. This model differs from other models that have been found in previous studies for other levels of technology. This result is significant to practitioners and researchers because it demonstrates the value of concentrating marketing efforts for low-tech offerings on vendor performance.

1. INTRODUCTION

That researchers are more interested in marketing high-tech offerings than low-tech offerings is demonstrated by the number of journal articles written on each. Online searches on ProQuest for "low tech marketing" and "high tech marketing" turned up 17 and 256 articles, respectively, in peer-reviewed journals. Nevertheless, there remain many low-tech offerings and marketers of low-tech items are disadvantaged by not having a significant body of literature on marketing of those items.

Increased customer repurchases increase the profitability of firms (Fornell, 1992) and maintaining customer satisfaction is key to retaining customers (Mittal, Katrichis and Kumar, 2001). Price has been a traditional form of competition but with increased competitive pressures, TQM, instantaneous worldwide communication, and flexible manufacturing, price competition is no longer sufficient. Product design, quality, performance, and delivery have become more important factors in creating customer relationships (Flint, Woodruff and Gardial, 2002). To be competitive, it is important that today's marketers understand exactly which factors drive customer satisfaction judgments and focus on fulfilling those elements.

High overall customer satisfaction is strongly correlated with increased repurchase intentions, which means that understanding customer satisfaction and the drivers of overall satisfaction is critical to market success (Patterson, Johnson and Spreng, 1997, Shapiro, Slywotzky and Doyle, 1997; Preis, 2000; Kellar and Preis, 2003). Components of overall satisfaction, such as satisfaction with the vendor's performance, satisfaction with the good or service itself, and interpersonal satisfaction (satisfaction with the relationship with the salesperson) are integrated by customers to form an overall evaluation (Patterson, Johnson and Spreng, 1997). While consumer markets have been the subject of considerable research on customer satisfaction, industrial markets have been largely ignored (Patterson, Johnson and Spreng, 1997; Szymanski and Henard, 2001; Homburg and Rudolph, 2001). The purpose of this study is to model the impacts of various components of satisfaction on overall customer satisfaction for industrial offerings. In particular, we examine the effects of components of satisfaction on overall customer satisfaction for low-tech offerings.

2. LITERATURE REVIEW

The most widely accepted model of customer satisfaction is the expectancy-disconfirmation paradigm (Patterson, Johnson and Spreng, 1997). In this model, when a consumer's expectation of performance is met, the expectation is said to be confirmed. When actual performance is below the expected level of performance, the expectation is negatively disconfirmed and when actual performance exceeds the expected level of performance, the expectation is positively disconfirmed. Overall satisfaction is considered a multi-attribute model in which overall satisfaction has multiple components.

Early research into satisfaction investigated satisfaction as a uniform construct across all product categories and in all markets. There was an implicit assumption that satisfaction is evaluated in the same way for all types of offerings. As the investigation of customer satisfaction matures, this assumption has begun to be questioned. Yi (1990), in a review of consumer satisfaction, anticipates that satisfaction varies from one product category to another when he questions, "Are the links between the variables different across consumers, products, or situations?" (p. 107). This suggests that the characteristics of an offering influence which aspects buyers use when making satisfaction judgments. For example, satisfaction with office supplies may be driven by satisfaction with the vendor whereas satisfaction with a fork lift may be more influenced by characteristics of the product itself, such as its reliability (Kauffman, 1994).

Mittal, Ross, and Baldasare (1998) find that "a consumer can be both satisfied and dissatisfied with different aspects of the product" (p. 34) at the same time. Thus, a customer might experience overall satisfaction in spite of components that are dissatisfying. Hypothetically, for example, a customer might experience high overall satisfaction because the salesperson was very accommodating and helpful and the vendor shipped on time, even though the product itself was not optimal. This ability to be simultaneously satisfied and dissatisfied with different attributes of the offering is important to marketers; by understanding the relative importance of attributes, marketers can spend resources on those attributes that increase levels of overall satisfaction while avoiding expenditures on attributes that have little influence on overall satisfaction (Kellar and Preis, 2003).

In industrial markets, components of customer satisfaction have different relative weights than in consumer markets; furthermore, parallel to findings for consumer markets, industrial buyers consider multiple attributes when making satisfaction judgments (Patterson, Johnson, and Spreng, 1997). Kauffman (1994) shows that when industrial buyers evaluate overall satisfaction, product attributes such as physical characteristics, nonphysical characteristics (such as warranty), price, and distribution, vary in importance depending on the product application. The product categories investigated by Kauffman (1994) are capital equipment, administrative purchases, and production process items. Thus, for example, price could outweigh all other characteristics when evaluating items used in production processes but nonphysical characteristics, such as warranty or reliability, might be more important when evaluating capital equipment where a breakdown could idle an entire assembly line. Kauffman (1994) also finds that differentiated products are evaluated differently than undifferentiated (commodity) products.

Low-tech products are not highly differentiated (Shapiro, Slywotzky and Doyle, 1997) and differentiated products are evaluated differently than undifferentiated products. This suggests that purchasers evaluate low-tech products, which are likely to be undifferentiated, differently than they evaluate high-tech products.

Crosby and Stephens (1987) find that a buyer's overall satisfaction with a service provider has three principal components: the service, the service provider, and the salesperson, but they conduct no testing of the weights assigned to each of the components or whether those weights vary depending on the product category. These findings of Crosby and Stephens (1987) from the consumer sector raise the question of whether the same factors are important in the industrial sector and what their relative weights should be. Our research uses these same three factors to model overall satisfaction for low-tech offerings in industrial markets for both products and services.

Product satisfaction is the satisfaction that a buyer feels with the offering and its durability, serviceability, and performance. Vendor performance satisfaction is a buyer's satisfaction with the performance of the vendor organization on such things as on-time delivery, invoicing accuracy, and warranty support. Interpersonal satisfaction is the buyer's satisfaction with the relationship with the salesperson and includes such things as evaluations of the salesperson's trustworthiness, ability to solve problems, communications ability, and personality (Crosby and Stephens, 1987) and is an important component of overall satisfaction (Hempel, 1977; Oliver and Swan, 1989; Crosby and Stephens, 1987; Crosby, Evans and Cowles, 1990).

Homburg and Rudolph (2001) state, "many firms producing and marketing industrial goods are strongly technically-minded implicitly assuming that the product is the most important source of customer satisfaction" (p. 30). They then show that customer satisfaction is a function of a proper subset of the three previously identified components of satisfaction. The composition of the subset for each individual depends on the functional role of that individual within the buying organization. Our study builds on this work by examining the influence of the level of technology on customer satisfaction and, in particular, the importance of the components of overall satisfaction for low-tech offerings within the industrial purchasing setting.

3. DATA COLLECTION AND LEVEL OF TECHNOLOGY ASSIGNMENTS

Using the "URL embedded" design (Bradley, 1999), all members of the National Association of Purchasing Management (NAPM) with e-mail addresses on file with the organization, were contacted by e-mail. The data collection effort was part of a larger more comprehensive study of dyads; each purchasing agent was requested to recruit one salesperson, from whom they had already made purchases, to participate in the study. The salesperson selected was to represent a product or service for which the purchasing agent had decision-making authority. Many items on the salesperson questionnaire differed from items on the purchasing agent questionnaire. Each participant was asked to go online to complete his or her respective questionnaire, independently of the other person.

Single items, each using a seven-point semantic differential scale, were used to assess overall satisfaction, product satisfaction (SAT_PDT), vendor performance satisfaction (SAT_PERF), and interpersonal satisfaction with the salesperson (SAT_REL). The salespeople provided generic descriptions of the items being purchased.

Invitations to participate were sent to 7,298 individuals. Messages were undeliverable to 1,787 addresses due to invalid or improper address information. Responses from 141 recipients claiming ineligibility were received. Thus, 5,370 is the size of the sample frame. In all, questionnaires were received from 193 dyads, yielding a response rate of 3.6%; this low response rate was no doubt due to the length of the questionnaires and the requirement that both parties in each dyad had to respond in order to achieve a useable response.

Demographically, 54 percent of the purchasing agent respondents are male and 46 percent are female. The mean age of purchasing agents reporting their ages is 45 years. As a group they are well educated, with 48 percent holding bachelors degrees and an additional 21 percent holding advanced degrees. The mean number of years of professional purchasing experience of the sample is greater than 14 and the mean number of years of tenure with their present employers is 9. Responses come from at least 41 different states. On all characteristics for which comparisons can be made, the demographics of the sample closely match the demographics of the overall membership of the NAPM, from which the sample is drawn.

Bias is a concern in any sample; the low response rate in this survey would be of greater concern if the demographics of the sample did not match the demographics of the NAPM, as closely as they do. The response rate is in line with other e-mail surveys (Weible and Wallace, 1998). Furthermore, in a study comparing mail and e-mail surveys, Bachmann, Elfrink, and Vazzana (1999/2000) find "evidence that nonresponse bias may be less of a problem using e-mail" than using mail surveys.

The determination of what constitutes a high-tech or low-tech industry is the source of much discussion. Luker and Lyons (1997) quote one analyst as saying, "Everyone knows what [high-tech] is, but no two lists are alike" (p.13). A report from the Congressional Office of Technology Assessment (1982) says that high-tech firms typically devote high proportions of expenditures to R&D activities; by extension, low-tech firms spend low proportions of expenditures on R&D activities. Because R&D expenditures as a proportion of sales will normally approximate this definition and because these figures are readily available on an industry-by-industry basis, we use this latter measure to determine the level of technology of industries. While any macro assessment of technology for an industry will find anomalies at the micro level (Luker and Lyons, 1997), this criterion provides an objective means for classifying industries into high-tech and low-tech categories. Furthermore, the results of such a classification are generally in agreement with most people's understanding of what it means to be "high-tech" or "low-tech." For a detailed description of the methodology used to determine the level of technology of offerings, see Kellar and Preis (2003).

Moderately high-tech industries spend a lower percentage of sales on R&D than high-tech industries, nevertheless, their spending on R&D is significant. Similarly, moderately low-tech and low-tech categories are grouped by the similarity of their spending on R&D. Thus, moderately high-tech industries include electrical equipment, appliances, and components, resins, synthetic rubber, fibers, and filament, and machinery; moderately low-tech industries include paper, printing, and support activities, nonmetallic mineral products, and fabricated metal products; and low-tech industries include food, transportation and warehousing, furniture and related products, and wood products. Because the purpose of the study is to examine low-tech offerings, the observations for high-tech, medium-high-tech, and medium-low-tech offerings are not used in this study. This leaves a sample size of 25.

4. ANALYSIS AND RESULTS

Kellar and Preis (2003) previously demonstrated that high-tech and low-tech offerings should be modeled separately; accordingly, we disaggregate the data corresponding to low-tech offerings from offerings with other levels of technology and restrict the modeling exercise to data from the low-tech group.

Akaike's Information Criterion Corrected (AICc), an information theoretic approach to model selection, is used to select an optimal regression model from among multiple competing models. AICc and related methodologies apply Occam's Razor (Principle of Parsimony) to balance the "goodness of fit" of each model with a penalty factor for the complexity of the model, in order to achieve optimal parsimony (Burnham and Anderson, 2002). Therefore, complex models with large numbers of variables will commonly be dropped in favor of more parsimonious models. Information theoretic methodologies, such as AICc and its predecessor AIC, have been shown to have advantages over more traditional regression model selection methods such as backward elimination and forward stepwise selection; one of these advantages is that all possible models are compared systematically rather than using a heuristic to guess at an optimal model (Bozdogan, 1990). Another advantage of using information theoretic approaches is that they balance a model's parsimony against its complexity to avoid overfitting (the inclusion of too many independent variables) (Burnham and Anderson, 2002).

The information theoretic methodologies (such as AIC or AICc) assign scores to each competing model and the model with the lowest score is preferred. Models with AICc scores that differ from the lowest score by less than a magnitude of 1 are not differentiated statistically, models with scores that differ from the lowest score by magnitudes of more than 1 but less than 2 are considered to be similar, and scores that differ from the lowest score by magnitudes of more than 3 have diminishing evidence of support (Burnham and Anderson, 2002). This means that an analysis will likely result in a portfolio of good models from which to choose an optimal model. The science of the field is then used to decide which model is considered most useful; the selected model is then considered to be optimal. That the portfolio contains more than one model is not a shortcoming of the methodology but rather emphasizes that the data are inadequate to determine a single best model (Burnham and Anderson, 2002).

The use of AICc is advocated over the use of other information theoretic approaches, such as AIC, when the ratio n/K is small, where n = sample size and K = the number of independent variables in the full model + 2. A common rule of thumb is that AICc should be used when n/K<40 (Burnham and Anderson, 2002). In our study, n=25 and K=5; therefore, the ratio of n/K=25/5=5, which is much less than 40 explaining why AICc is used instead of its information theoretic parent, AIC, to select the portfolio of optimal models.

In the analysis performed in this study, three exogenous variables are considered (SAT_PDT, SAT_PERF, and SAT_REL). Seven possible models representing all of the possible subsets of the independent variables exist. Results for all 7 models are shown in Table 1, sorted in ascending order by AICc score.

The first model in Table 1, i.e., the one with the lowest AICc score, is necessarily in the portfolio of models to be considered. For readers who are not familiar with the AICc methodology, we include the second and the third models in the portfolio of models, even though their AICc scores differ by a magnitude of more than 2.50 from the lowest score (ordinarily when a model's AICc score differs from the lowest AICc score by this amount it would not be included in the portfolio of models).

Note that for the three models in the portfolio, SAT_PERF is the dominant variable; in the second and third models, the coefficients of the other included variables are very small, indicating that if one of these models is selected, overall satisfaction will still be determined almost exclusively by SAT_PERF. Since our goal is to identify a useful model and there is no compelling reason to prefer any model other than the one with the lowest AICc score, we select the first model as the most useful. Nevertheless, for the convenience of those readers unfamiliar with information-theoretics, p-values are reported in Table 1.

It should be noted that since the preferred model is a single-variable model, we need not be concerned with multicollinearity.

4.1 Replication of Results

A second survey utilizing a similar methodology to the first survey and distributed to members of the Institute for Supply Management (ISM--formerly NAPM) resulted in a sample of 9 low-tech observations. While the sample size is quite small, the results of the first study were confirmed. AICc found the optimal model for the second data set to be very similar to the model found for the first data set, the differences being small disparities in the values of the coefficient for SAT_PERF and [R.sup.2]. With the second data set, the model for the underlying population is found to be the same for unique samples separated in time, confirming the results of the earlier study and suggesting that the model for low-tech offerings is robust.

5. MANAGERIAL IMPLICATIONS

Studying customers at the attribute level instead of at global or overall levels is of greater value to practitioners because managers work at the attribute level (Wittink and Bayer, 1994). For low-tech offerings, the results of this research clearly show that sales managers and salespeople for industrial offerings could improve their results by maximizing vendor performance satisfaction for low-tech offerings. This is not to say that product characteristics can be ignored, but within limits, industrial buyers of low-tech offerings are most influenced by vendor performance. This suggests that where choices exist, sales managers of low-tech offerings would be well advised to raise satisfaction with vendor performance rather than spend resources on attributes affecting other types of satisfaction.

It is worthy of mention that many previous attempts to understand overall satisfaction have used aggregated data. The implicit assumption in such analyses is that offerings containing differing levels of technology are not distinguished by buyers, but that satisfaction judgments are based on the same factors and are weighted identically for all levels of technology. The importance of this study is that it shows marketing managers of low-tech offerings what factors to improve in order to maximize customer satisfaction. These results could not have been achieved without disaggregating the data into separate levels of technology. The same phenomenon also occurs when examining the relationship between the components of satisfaction and repurchase intention (Kellar and Preis, 2003).

It is reasonable to hope that similar analyses applied to other levels of technology will yield results that will be as beneficial in their realms as these findings should prove to be for low-tech offerings.

Finally, the methodology used in this study may prove to be more valuable than the actual results. The methodology takes advantage of two techniques that are relatively new to the study of satisfaction. In particular these techniques include 1) disaggregation of data and focused models based upon the disaggregated data, and 2) the identification of optimal models using information theoretic statistics and best variable subset selection methodologies. The methodology for such studies is virtually "boiler-plated" in this study.

6. LIMITATIONS

The results of this study should be interpreted with caution. These results should only be accepted to the extent that buyers' satisfactions are accurately reported. Peterson and Wilson (1992) showed that virtually all studies of customer satisfaction are highly skewed to the satisfied end of the scale and our survey results reflect this skewness. Also, these findings are applicable only to low-tech offerings and application of the findings to medium- or high-tech offerings is not justified.

Goods and services are analyzed together; it may be that differences exist in the ways that buyers evaluate goods and services. (Due to sample size limitations, results of modeling efforts that distinguished products from services are not reported in this study. However, it is worthy of mention that the unreported modeling efforts found the models in both product and service categories to be similar. In particular, preliminary results showed that SAT_PERF was the only independent variable in each of the optimal models, and that the coefficients and [R.sup.2]s, while not identical, were not significantly different for the two optimal models. Thus satisfaction judgments for goods and services may be similar for low-tech offerings.) Furthermore, the sample includes buyers of all manner of offerings; it may be that when buyers of single commodity-types are analyzed, satisfaction will be a function of different factors than are shown to be important in this study. The study examines only the three identified components of overall satisfaction and examines their effect on overall satisfaction judgments. In spite of the relatively high amount of variance explained for customer satisfaction, the unexplained variance indicates that significant factors remain to be identified.

The questionnaires used to collect data utilized single items to measure each component of satisfaction; single items may not fully capture the entire range of factors that purchasers include in their judgments of each component of overall customer satisfaction. For example, a purchaser may be extremely satisfied with the technology used in a product but may not be fully satisfied with the reliability of the product; a single-item measure of the satisfaction with the product may not allow respondents the opportunity to fully express their satisfaction judgments. Nevertheless, single-item measures are commonly used in satisfaction research and are reliable (Drolet and Morrison, 2001).

The survey was conducted via e-mail and the Internet. Despite widespread use of the Internet by purchasing professionals not all organizations use it nor are all purchasing professionals comfortable with the Internet. In spite of the apparent similarity between the demographics of the sample and the demographics of the NAPM, the sample could be restricted in its generalizability.

7. IMPLICATIONS FOR FUTURE RESEARCH

Despite the small confirmatory study that was conducted, the findings of this study should be replicated in other research. Other factors may also be important in evaluating overall satisfaction; accordingly, research should be undertaken to determine the roles of additional factors, such as price and brand name.

Similarities of some characteristics of satisfaction are shared between high-tech and low-tech products but not medium--tech products (Shapiro, Slywotzky and Doyle, 1997). This introduces the question as to how technology levels should be clustered optimally. Application of multi-sample cluster analysis prior to model building might reveal characteristics that are common across groups.

This study has categorized technology into 4 distinct levels. The optimal number of categories should be investigated and the differences between categories should be studied to gain a better understanding of how level of technology affects industrial buyer behavior.

This study examines industrial buyers; it would be interesting to model customer satisfaction for consumers using the same data groupings. Our study has combined both goods and services under the rubric low-tech offerings; research should be undertaken to examine whether goods differ from services in the factors that are considered when industrial buyers make satisfaction judgments. Managerially, it would be useful to know whether industrial buyers are aware of the importance they place on satisfaction with vendor performance; if they are, then emphasizing vendor performance in advertising would be beneficial since this speaks to the criteria buyers use to form satisfaction judgments. However, to the extent that buyers errantly believe they place greater emphasis on price, product performance, or relationships with vendors, such expenditures on advertising that stresses vendor performance might be misplaced.

8. CONCLUSION

This study builds on prior research and examines the roles of distinct components of overall satisfaction to determine their effects on overall satisfaction. The findings show that industrial buyers make their satisfaction judgments for low-tech offerings based on satisfaction with vendors' performance. If marketers of low-tech offerings apply these findings, they have the potential to improve overall customer satisfaction which should lead to not only maintaining current customers but also to increasing their repurchase activity. As it costs five times more to acquire a new customer than it does to retain an existing customer (Kotler and Armstrong, 2001), marketers of low-tech offerings should be very interested in taking advantage of these results to increase customer satisfaction while continuing to grow their markets through new customer acquisition. Marketers of other than low-tech offerings also should be interested in these findings and should duplicate the methodology of this study to calculate parallel results for their own industries, companies or offerings.

Table 1. Analysis of All Possible Models

AICc    Constant   Coefficients                     [R.sup.2]

                   SAT_PDT    SAT_PERF   SAT_REL

34.44     3.86                 0.399                  0.414
37.27     3.78      0.030      0.382                  0.414
37.29     3.94                 0.406      -0.018      0.414
40.38     3.91      0.054      0.387      -0.048      0.415
41.20     3.87      0.388                             0.232
43.91     3.86                             0.371      0.144
44.00     3.70      0.353                  0.058      0.233

* statistically significant at p=0.05

** statistically significant at p=0.01

(As noted in the text, statistical significance is presented
here for readers' convenience but has little relevance for
information-theoretic paradigms.)

REFERENCES

Bachmann, D. P., J. Elfrink and F. Vazzana, "E-mail and Snail Mail Face Off in Rematch". Marketing Research, Vol. 11(4), (Winter/Spring), 1999/2000, 10-15.

Bozdogan, H., "On the Information-based Measure of Covariance Complexity and its Application to the Evaluation of Multivariate Linear Models". Communications in Statistics, Vol. 19 (1), 1990, 221-278.

Bradley, N., "Sampling for Internet Surveys. An Examination of Respondent Selection for Internet Research". Journal of the Market Research Society, Vol. 41(4) (October), 1999, 387-395.

Burnham, K.P. and D.R. Anderson, Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach, 2nd Ed., Springer, New York, 2002.

Congressional Office of Technology Assessment, "Technology, Innovation, and Regional Economic Development". Sept. 9, 1982, quoted in D. Hecker, "High-technology Employment: A Broader View". Monthly Labor Review, (June), 1999, 18-28.

Crosby, L. A., K. R. Evans and D. Cowles, "Relationship Quality in Services Selling: An Interpersonal Influence Perspective". Journal of Marketing, Vol. 54 (July), 1990, 68-81.

Crosby, L. A. and N. Stephens, "Effects of Relationship Marketing on Satisfaction, Retention, and Prices in the Life Insurance Industry". Journal of Marketing Research, Vol. XXIV, (November), 1987, 404-11.

Drolet, A.L. and D.G. Morrison, "Do We Really Need Multiple-Item Measures in Service Research?". Journal of Service Research, Vol. 3 (3) (February), 2001, 196-204.

Flint, D. J., R. B. Woodruff and S. F. Gardial, "Exploring the Phenomenon of Customers' Desired Value Change in a Business-to-Business Context". Journal of Marketing, Vol. 66 (4), 2002, 102-117.

Fornell, C., "A National Customer Satisfaction Barometer: The Swedish Experience". Journal of Marketing, Vol. 56 (January), 1992, 6-21.

Hempel, D. J., "Consumer Satisfaction with the Home Buying Process: Conceptualization and Measurement". In H. K. Hunt (Ed.), Conceptualization and Measurement of Consumer Satisfaction and Dissatisfaction .(pp.275-299), Marketing Science Institute, Cambridge, MA, 1977.

Homburg, C. and B. Rudolph, "Customer Satisfaction in Industrial Markets: Dimensional and Multiple Role Issues". Journal of Business Research, Vol. 52, 2000, 15-33.

Kauffman, R. G., "Influences on Industrial Buyers' Choice of Products: Effects of Product Application, Product Type, and Buying Environment". International Journal of Purchasing and Materials Management, (Spring), 1994, 29-38.

Kellar, G. M. and M. W. Preis, "Modeling Increased Repurchase Intentions in High-Tech B2B Markets Using an Information Theoretic Approach". Academy of Marketing Studies Journal, 2003.

Kotler, P. and G. Armstrong, Principles of Marketing (9th ed.), Prentice Hall, Upper Saddle River, NJ, 2001.

Luker, W., Jr. and D. Lyons, "Employment Shifts in High Technology Industries, 1988-96". Monthly Labor Review". Vol. 120:6 (June), 12-25.

Mittal, V., J.M. Katrichis and P. Kumar, "Attribute Performance and Customer Satisfaction Over Time: Evidence From Two Field Studies". The Journal of Services Marketing, Vol. 15(4/5), 2001, 343-356.

Mittal, V., W. T. Ross and P. M. Baldasare, "The Asymmetric Impact of Negative and Positive Attribute-Level Performance on Overall Satisfaction and Repurchase Intentions". Journal of Marketing, Vol. 62(1), 1998, 33-47.

Oliver, R. L. and J. E. Swan, "Consumer Perceptions of Interpersonal Equity and Satisfaction in Transactions: A Field Survey Approach". Journal of Marketing, Vol. 53 (April), 1989, 21-35.

Patterson, P. G., L. W. Johnson and R. A. Spreng, "Modeling the Determinants of Customer Satisfaction for Business-to-Business Professional Services". Journal of the Academy of Marketing Science, Vol. 25(1), 1997, 4-17.

Peterson, Robert A. and William R. Wilson, "Measuring Customer Satisfaction: Fact and Artifact". Journal of the Academy of Marketing Science, Vol. 20 (1) (Winter), 1992, 61-71.

Preis, M. W., Effects of Temperaments and Value Systems on Customer Interpersonal Satisfaction and Repurchase Intention in Dyadic Sales Relationships. Unpublished doctoral dissertation, The George Washington University, 2000.

Shapiro, B. P., A. J. Slywotzky and S. X. Doyle, "Strategic Sales Management: A Boardroom Issue". Strategy & Business (Booz-Allen & Hamilton), (3rd Quarter), 1997, 29-47.

Szymanski, D.M. and D.H. Henard, "Customer Satisfaction: A Meta-Analysis of the Empirical Evidence". Journal of the Academy of Marketing Science, Vol. 29(1), 2001, 16-35.

Weible, R. and J. Wallace, "Cyber Research: The Impact of the Internet on Data Collection". Marketing Research, Vol. 10 (3) (Fall), 1998, 19-24.

Wittink, D. R. and L. R. Bayer, "The Measurement Imperative". Marketing Research, Vol. 6 (4), 1994, 14.

Yi, Y., "A Critical Review of Consumer Satisfaction". In V. A. Zeithaml (Ed.), Review of Marketing 1990, p. 68-123, American Marketing Association, Chicago, 1994

Dr. Michael W. Preis earned his Ph.D. at George Washington University in 2000. Currently he is an assistant professor of marketing at Long Island University/C.W. Post Campus, Brookville, NY.

Dr. Gregory M. Kellar earned his Ph.D. at the University of Tennessee at Knoxville in 1996. Currently he is an assistant professor at Penn State University, Delaware County.