Decreasing Returns in Customer Loyalty: Does It Really Matter to Delight the Customers?

ABSTRACT - Because of the contradicting results between customer satisfaction and loyalty, practitioners and researchers alike are now arguing that firms should take quality beyond customer satisfaction to customer delight or 100% satisfaction. This study questions the effectiveness of the 'customer delight’ or '100% satisfaction’ in increasing customer loyalty. A two-threshold, utility-oriented model of the effects of satisfaction on loyalty is developed. The empirical results show that trying to delight customers or provide 100% satisfaction may not be worth the effort because there are points where the effect of customer satisfaction on loyalty levels off.

Citation:

Paul-Valentin Ngobo (1999) ,"Decreasing Returns in Customer Loyalty: Does It Really Matter to Delight the Customers?", in NA - Advances in Consumer Research Volume 26, eds. Eric J. Arnould and Linda M. Scott, Provo, UT : Association for Consumer Research, Pages: 469-476.

Advances in Consumer Research Volume 26, 1999      Pages 469-476

DECREASING RETURNS IN CUSTOMER LOYALTY: DOES IT REALLY MATTER TO DELIGHT THE CUSTOMERS?

Paul-Valentin Ngobo, The University of Montpellier

[The author gratefully acknowledges the defying comments of the ACR Reviewers and the support of the marketing students at the Department of Marketing, The University of Montpellier.]

ABSTRACT -

Because of the contradicting results between customer satisfaction and loyalty, practitioners and researchers alike are now arguing that firms should take quality beyond customer satisfaction to customer delight or 100% satisfaction. This study questions the effectiveness of the 'customer delight’ or '100% satisfaction’ in increasing customer loyalty. A two-threshold, utility-oriented model of the effects of satisfaction on loyalty is developed. The empirical results show that trying to delight customers or provide 100% satisfaction may not be worth the effort because there are points where the effect of customer satisfaction on loyalty levels off.

INTRODUCTION

"We must take quality beyond customer satisfaction to customer delight" (Chairman and CEO of Eastman Kodak quoted by Chandler, 1989)

"Although the elimination of defects is critical to continuing customer satisfaction, increased productivity, and decreased costs, it is customer delight that is the key to survival in today’s markets" (AMP Canada’s Quality Program Manager, quoted by Wittaker, 1991).

Customer retention is critical to business success in today’s competitive environment. This importance has led marketing scholars and practitioners to recommend that firms improve their customers’ satisfaction because satisfaction is a key to customer loyalty and retention (Fornell et al. 1996a). Indeed, many studies report a positive association between customer satisfaction and customer loyalty (see Fornell et al. 1996a; Hallowell, 1996; Bolton, 1998). However, researchers and practitioners have also begun to express concern on the process that underlies the satisfactionBloyalty relationship as the empirical evidence shows that improvements in customer satisfaction do not always result in proportional improvements in loyalty (see Mittal et al. 1998; Auh and Johnson, 1997). As a result, researchers (Oliver et al. 1997) as well as practitioners (Mendelsohn, 1998) are now arguing for customer delight or 100% satisfaction in retention programs. As shown by the practitioners’ opinions at the beginning of the paper, the motto of many business executives now seems to be " No delight, No retention". If this viewpoint is "true" then only very satisfied customers should be more loyal and retained by the firm. The moderate and the low in satisfaction should not be or should be in a lesser extent. The purpose of this paper is to investigate this "received view". Specifically, the question we address is the following: does it really matter that firms delight or provide 100% satisfaction to their customers to achieve greater customer loyalty? The rest of the paper is organized as follows. The first section presents a theoretical review of the satisfaction-loyalty studies. The second section theorizes on the relation of satisfaction and loyalty. Throughout this paper we defend the following thesis: customer delight may not be worth the effort because there are threshold points where the effect of satisfaction on loyalty increases at decreasing rates and the firm stops reaping the benefits of customer satisfaction in terms of customer loyalty. The third section is thus an attempt to demonstrate this position with empirical data. The paper concludes with a discussion of the implications of the research.

CUSTOMER LOYALTY: FROM CUSTOMER SATISFACTION TO CUSTOMER DELIGHT

In this paper, customer satisfaction is conceptualized at the customer level as an emotional response to the consumption experience (Hunt, 1977). In this sense, measures of satisfaction should describe emotional feelings such as delight, gladness, enjoyment (Hausknecht, 1988). Customer loyalty, here, refers to a customer’s psychological predisposition to repurchase from the same firm again (Auh and Johnson, 1997; Dick and Basu, 1994). Loyalty is different from customer retention because the latter represents the actual product/service repurchase over time. Thus, our measures of loyalty should describe only repurchase intentions in the next purchase occasions.

The 'linear’ view of the customer satisfactionBloyalty relationship

Most of the approaches relating customer satisfaction with loyalty are based on an explicit or implicit linear assumption (Bearden and Teel, 1983; Fornell, 1992). (See Figure 1).

The rationale is that satisfaction strengthens brand attitudes and this causes the size of the consideration set to shrink (Sambandam and Lord, 1995). Indeed, a satisfying brand should remain accessible until the subsequent choice occasion. The linear relationship beteen satisfaction and loyalty (see Figure 1a) has been supported empirically by many authors (e.g. Fornell et al. 1996a; Bearden and Teel, 1983; Hallowell, 1996). However, is the relation that simple?

The 'customer delight’ view in customer loyalty

A lot of research has begun to question the linear view (Coyne, 1989; Mittal et al. 1998; Fornell et al. 1996b; Anderson and Mittal, 1997; Jones and Sasser, 1995; Chandler, 1989; Whittaker, 1991). Yet, there is ongoing debate as to the nature of the relationship. Coyne (1989) posits that the loyalty curve is relatively flat after a first threshold (of satisfaction) is passed and climbs rapidly after customer satisfaction has passed a second threshold (see Figure 1b). This has been supported by Oliva et al. (1992). Jones and Sasser’s (1995) typology of customers is also consistent with that view. According to these authors, loyalty would increase marginally over moderate to high levels on the satisfaction continuum and then increases considerably at higher levels of satisfaction. Anderson and Mittal (1997) also suggest that in the extreme level of satisfaction, delighted customers tend to ignore competing brands in favor of one which has delighted them in the past. Oliver and Colleagues (1997) found that customer satisfaction and delight were acting in parallel in their model, indicating that customers make a difference between their being satisfied and their being delighted. However, in the two samples (Park visitors and Symphony concert clients) customer satisfaction is the major antecedent to repurchase intentions. Delight impacts repurchase intentions of symphony concert clients only and less than satisfaction.

Thus, to our knowledge, there is no empirical study which demonstrates formally of the superiority of customer delight over customer satisfaction.

FIGURE 1

DECREASING RETURNS IN CUSTOMER RETENTION

Studies on customer delight seem to neglect one important characteristic of human behavior: the saturation effect. Micro-economists have long recognized that there are decreasing returns associated to such activities as consumption and production. A few studies on customer loyalty have begun to integrate that dimension. For instance, Fornell et al. (1996b) find that at the customer level the relation between customer satisfaction and economic performance is characterized by diminishing returns. Rust et al. (1995) found that the dissatisfied customers of a hotel’s chain had a probability of 45% of returning, those who were satisfied had a probability of 95% and those who were delighted (i.e. 100% satisfied) had a probability of 97%. However, these authors recognize the decreasing return idea since they admit that the biggest benefits by the company were derived from converting customers from dissatisfied to satisfied and not towards delighted. The problem with the nonlinear approach to the satisfaction-loyalty issue however is the lack of a theoretical foundation that would assist researchers in their specifying of the appropriate functional form. Without such a theory, misspecifications are likely to occur in empirical studies. Auh and Johnson (1997) have developed such a theory based on Howard’s (1977) three-stage theory of consumer choice. They propose that at low levels of satisfaction customers have relatively large consideration sets because they are in an early stage of problem solving. As satisfaction keeps increasing customers reduce the size of their consideration sets and begin to utilize rules stored in memory to simplify their decisions. However, albeit small, the size of consideration sets is likely to be greater than one because some consumption contexts are related to more than one alternative and also because of the switching possibilities in highly competitive industries. These authors therefore propose that the positive effects of satisfaction on loyalty levels off at a certain point. In other words, custmer satisfaction has a negative cubic relationship with loyalty. Their results however do not demonstrate the superiority of the cubic relationship over the quadratic positive relationship (i.e. the customer delight view) at least in terms of explained variance.

A THEORY OF THE CUSTOMER SATISFACTIONBLOYALTY RELATIONSHIP

This study is more related with the 'decreasing returns’ perspective than with the 'increasing returns’ view. The objective is to draw on previous conceptualizations (e.g. Auh and Johnson, 1997) and to place them within the expected utility theory. Yet, our conceptualization still differs. For example, Auh and Johnson (1997) adopt an information seeking perspective where to be loyal an individual is supposed to have crossed the various choice stages established by Howard (1977). Loyalty is actually established only when customers have come to the routine behavior stage, where the emphasis shifts from external information search to internal search. More importantly, Auh and Johnson (1997) propose that firms should transfer customers who are moderately satisfied to the very satisfied segment where satisfaction impact on loyalty is greater. This however contradicts our propositions (see Conclusion).

In this paper, we adopt a utility perspective to the customer satisfaction-loyalty issue. There are two major arguments in favor of such an approach (Anderson, 1998): the need for a parsimonious model that captures the fundamental relationship between customer satisfaction and loyalty and the fact that a utility-oriented approach provides a means of integrating the host of the reasons why an individual might choose to repurchase from the same firm or the same brand again. Utility is regarded here as the ability of a brand or a firm to meet the customer’s requirements. In general, it may be posited that customers engage in loyalty either because they need it or they desire it. By need, we mean a cognitive-based loyalty where loyalty is related to the extent to which the customer perceives the need to repurchase the same brand or from the same firm given the anticipated benefits associated with this brand (e.g. the value for money). A relationship with a firm or a brand may also be kept because the customer likes the brand and/or identifies himself/herself with the firm (Fournier, 1998). In this case, loyalty is primarily determined by a generalized sense of positive regard for and attachment to the brand (Geykens et al. 1996). In sum, different reasons make customers loyal with a brand. Customers might develop loyalty behaviors when they feel that the brand has superior features and attributes, what has worked in the past is likely to work in the future, they want to save the time devoted to making purchase decisions, new alternatives are not worth the effort, they feel that the brand can meet the self-esteem need, one’s need for a sense of discipline and order, and the need for self-respect, etc. All these reasons or utilities then explain why the customer would be loyal. The basic model proposed here explains how customer satisfaction is associated with loyalty through these various reasons.

FIGURE 2

THE PROPOSED MODEL OF THE CUSTOMER SATISFACTION-LOYALTY RELATIONSHIP

To remain quite present in the debate on customer delight, we primarily will treat customers’ loyalty which results from their satisfaction with previous consumption or service experiences. In this sense, we exclude the habitual buying behavior as a means to reduce perceived risk and effort. Indeed customer loyalty occurs when the customers exhibit repeat purchase behavior which is the result of a strong positive attitude toward the brand, which in turn makes them more resistant to brand switching (Dick and Basu, 1994). Habitual consumers however simply buy repeatedly, even if they do not necessarily have a strong positive attitude toward the brand as part of their risk and effort reduction strategy. Their behavior is therefore somewhat 'independent’ from their satisfaction. This being, we propose that the relation o customer satisfaction and loyalty should exhibit three features (see Figure 2).

First, loyalty should increase more after a certain level of satisfaction has been reached (a minimum threshold). Indeed, any consumption experience is not supposed to lead to customer loyalty. It is the expected utility of such a behavior that will be determinant. This expected utility will be a function of customer satisfaction. This is consistent with Bolton’s (1998) proposition that customers engage in a relationship on the basis of its future value (i.e. a cost- benefit analysis). This value is primarily determined by prior customer satisfaction and eventually the new information obtained by the customer. However, we limit our conceptualization to customer satisfaction only. Note that this minimum level may differ across individuals on the basis of their accumulated experiences. Yet, it will 'trigger’ customer loyalty if the customer can derive enough utility by repeat purchasing the brand. As a result, brand attitude should be enhanced and the customer should start to reduce the number of brands considered on a purchase occasion, with the satisfactory brand being recalled first.

Second, within points f₁ (the minimum) and f₂ (the maximum), loyalty should increase considerably as the utility of engaging in a relationship with the brand keeps increasing. Indeed, there should be a phase where satisfaction and loyalty should be increasing more due to the increase in the derived utility (e.g. better attributes) from the relationship. Within this phase, the customer should also be trusting, preferring the firm’ products more and the perceived benefits of the relationship with the brand should outweigh its costs. The impact of satisfaction on loyalty therefore should be greater within this phase and this should result in an enhanced brand attitude, and in a decrease in the size of the consideration set.

Third, there should be a 'ceiling effect’ zone whereby loyalty simply levels out after a certain level of satisfaction has been reached. This ceiling effect zone corresponds to any point where high levels of customer satisfaction don’t increase loyalty levels beyond lower levels anymore. In this case, each additional unit of customer satisfaction will bring less in incremental loyalty than the previous unit did.

Formally, the relation of customer satisfaction and loyalty can be expressed as follows:

EQUATION (1a)

where L=loyalty activity; SAT=Level of customer satisfaction; f₁=the minimum threshold of satisfaction that is necessary to induce loyalty and f₂=the saturation level where satisfaction has diminishing returns on loyalty. Equation (1a) posits that the loyalty function has two thresholds (f₁ and f₂) showing customer loyalty initially increasing more after the first threshold (f₁) is reached at an increasing rate (i.e. within the two thresholds: f₁ and f₂) and then less increasing at a decreasing rate after the second threshold (f₂) is reached.

EQUATION (1b)

Equation 1b indicates that loyalty should be greater within than out of the two thresholds (f₁ and f₂) and it should also be greater in the 'ceiling effect’ area than in the 'defection’ zone. Indeed, there are limits on time and effort associated to any activity. An increase in loyalty the customer chooses to engage in should depend on the marginal utility s/he derives from such an activity relative to any other activity (e.g. brand switching). When loyalty brings back more benefits than any other activity, the customer shoud increase his/her commitment to this brand. For instance, a customer who receives poor quality would be more predisposed to switch. As satisfaction increases, the customer perceives the utility to repurchase the same brand again. But, at very high levels of satisfaction, loyalty might levels out. Different reasons might explain such a phenomenon. The first is the saturation effect. Indeed, when the customer is exposed to increasing levels of a same attribute or brand, the marginal utility of that attribute or brand might decrease. What can occur when the customer buys the same brand constantly. The second reason is related to competitive intensity within the industry. Indeed, when the firm improves customers’ satisfaction, competitors usually do the same or use offensive strategies to "break" some established habits (Fornell, 1992). For example, increasing competition might move the standards of product evaluation up. This in turn might decrease customer relative utility with the current brand. In either case, the marginal utility that will be derived from brand switching can become greater than the marginal utility of loyalty.

DEMONSTRATION OF THE POSITION AT HAND

In order to discuss the decreasing effects of satisfaction on loyalty, we use four samples of customers. Sample 1 is composed of banks’ clients, sample 2 of car insurance policyholders, sample 3 is composed of clients of a newly established (4 months) retailer and sample 4 is composed of buyers of cameras. Except for sample 4 (where the data was collected in a larger city) all the respondents were interviewed by marketing students from the local university as part of a marketing research course. All samples were conveniently composed. Convenience samples are appropriate for this research given that we are interested in whether or not our samples have enough variance that can be explained by the theoretical model. To test models, we require measures of satisfaction and loyalty (see Table 1). Satisfaction. Four 7-point Likert items were selected from those suggested in previous works (see Bitner and Hubbert, 1994; Oliver, 1997).

Loyalty was measured with five 7-point Likert items in terms of repurchase intentions and Word-of-mouth intentions (Nayarandas, 1996). The idea being that loyalty results in a stronger brand attitude (Dick and Basu, 1994). This strong and positive attitude is translated into strong repurchase and word-of-mouth intentions. The cut-off value of 0.40 was retained for the item-total correlation. All the measures have acceptable reliability. Table 2 summarizes the main characteristics of the data.

MODEL SPECIFICATION ISSUES

Empirical Models. In order to test whether customer satisfaction has a linear, a quadratic or a two-threshold relationship with customer loyalty, we specify the following equations:

(2) L_i = a_o + b_SATLSAT_i + e_i   (Linear Model)

(3) L_i = a_o + b_SATLSAT_i +   b_SATQ SAT_i2 + e_i  (Quadratic Model)

(4) L_i = a_o + a_oS1S₁(f₁) + a_oS2S₂(f₂) +   b_SATM0SAT_i + b_SATL0[SAT_i X S₁(f₁)]   + b_SATHi[SAT_i X S₂(f₂)] + e_i (Proposed Model)

Where L_i represents loyalty of customer i; SAT_i is customer satisfaction of individual i. Two dummy variables are used in Equation 4. S₁ equals '1’ if satisfaction is less than the minimum threshold f₁ and '0’ otherwise; S₂ equals '1’ if satisfaction is greater than the maximum threshold f₂ and '0’ otherwise. The term a_o is the constant term; b_{variable name} is the coefficient with different subscripts indicating the 'linear’ (L), 'quadratic’ (Q), 'moderate satisfaction’ (Mo), 'low satisfaction’ (Lo) and 'high satisfaction’ (Hi) and e_i is the error term. Equation 2 is the appropriate specification to capture the linear relationship. It implies that loyalty keeps increasing as satisfaction increases. The quadratic components were constructed with mean-centered scores. Equation 3 represents the quadratic relationship. It proposes that loyalty increases differently depending on satisfaction values. The '100% satisfaction’ view will be accepted if: (1) Equation 3 fits the data better than the other models and (2) the quadratic component is significant and positive. Equation 4 is the appropriate specification for testing the hypothesized functional form. This dummy-variable equation allows us to define three different groups of customers according to their levels of satisfaction:

Group 1: Low satisfaction: if f₁ =1 and f₂ =0, then,

            L_i = (a_o + a_oS1) + (b_SATLo + b_SATMo)*SAT_i + e_i

Group 2: Moderate satisfaction: if f₁ =1 and f₂ =0, then,

            L_i = a_o + b_SATMo + *SAT_i + e_i

Group 3: High Satisfaction: if f₁ =1 and f₂ =0, then,

            L_i = (a_o + a_oS1) + (b_SATHi + b_SATMo)*SAT_i + e_i

All the differences in the parameters are assessed in reference to Group 2. The constants a_oS1 and a_oS2 are supposed to capture the differences in the origins of the regression lines across the three groups of customers. The coefficients b_SATLo and b_SATHi account for the variation in the effects of customer satisfaction on loyalty across the different levels of satisfaction.

Estimation Procedure. Traditional regressions were used for Equations 2 & 3 while Equation 4 was estimated using segmented regression. One important methodological issue in estimating equation 4 is determining the minimum and the maximum thresholds of satisfaction. We estimate equation 4 for each possible couple of values within the satisfaction scale values [1, 7] and select that which maximizes the variance explained by the model. The proposed model will be accepted if equation 4 has superior fit to equation 2 & 3 and exhibits the proper pattern for the coefficients: (1) the coefficient for the dummy variables are significant, (2) that for the individuals high in satisfaction (group 3) is less than the coefficient for the individuals in group2 but greater than that for the individuals low in satisfaction, and (3) the coefficient for the individuals moderate in satisfaction is greater than that for the individuals low in satisfaction.

TABLE 1

THE MEASUREMENT VARIABLES

TABLE 2

SAMPLE CHARACTERISTICS

THE FINDINGS

The linear Model. Table 3 presents the estimation results of equation 2.

The estimates indicate that the linear model cannot be rejected. Almost all the parameters (except for the constant in the insurance policyholders) are significant. This model therfore is going to be used as a comparison standard for equations 3 & 4.

The quadratic model. Table 4 presents the estimate results of equation 3.

Table 4 shows that the quadratic model provides significant (except for the retailer data) and meaningful results. The quadratic component is negative in the bank and insurance samples and positive in the camera data. However, it is necessary for us to carry out a formal test before drawing any conclusion on the existence of any effect. Two criteria will be used: the adjusted R² and the traditional F-test (see Kleinbaum et al. 1998, p. 391). The F criterion tests whether the difference in R² differs significantly from 0 while the adjusted R² indicates the variance explained by the model once the number of parameters is taken into account. The results of these tests show that the quadratic model explains more variance than the linear model in (1) the bank data: adj. R² is 0.66 versus 0.65 in the linear model (about 1.5% increase in R²) and this is confirmed by the F test: F(1; 70)=4.24, p<0.05, (2) the insurance data: adj. R² is 0.70 versus 0.69 in the linear model, the F test is slightly significant: (F(1; 50)=3.57, p<0.10, and (3) in the camera data: adj. R² is 0.65 versus 0.63 in the linear model and the F test is also significant: (F(1; 221)=6.31, p<0.01. The linear model however still explains more variance than the quadratic model in the retailer data: adjusted R² is 0.36 in the linear as well as the quadratic model and this is confirmed by the F test: F(1; 47)=0.76, p>0.25. Thus, for the rest of the analysis, the quadratic model becomes the reference point for model selection in the bank, insurance and camera samples while the linear model remains the reference point for the retailer data.

TABLE 3

THE ESTIMATION RESULTS OF EQUATION 2

TABLE 4

THE ESTIMATION RESULTS OF EQUATION 3

Estimation results of the proposed model. Table 5 presents the estimation results of equation 4.

The F test shows that the proposed model performs better than the quadratic model in (1) the bank data at least in terms of the adj. R² (0.68 versus 0.66) and less in terms of the F test which is slightly not significant (F (3; 67)=1.47, p>0.10 and (2) in the camera data: adj. R² equals 0.68 versus 0.65 in the quadratic model and this is confirmed by the F test (F(3; 218)=9.58, p<0.001. Note that there is an increase of about 6.15% of the R² in the camera data in comparison with the quadratic model. The proposed model however cannot be accepted for the insurance data even though it has a greater adjusted R² value (0.72 versus 0.70 for the quadratic model). Indeed, the F value is not only too low F(3; 47)=0.58, p>0.25) but also not all the key parameters are significant. The quadratic model therefore cannot be rejected for the insurance data. The linear model also can’t be rejected for the retailer data as it is shown by the adj. R².

DISCUSSION OF THE RESULTS

A Two-threshold Model of the Customer SatisfactionBLoyalty Relationship. The empirical results indicate that loyalty of the bank clients exhibit a minimum (= 4) and a maximum (=6.5) thresholds. That is customer satisfaction has a greater impact on loyalty after satisfaction has reached a score of 4 and it does not increase loyalty further after satisfaction has passed a score of 6.5. The proposed model fits the data quite well in the camera sample. The minimum value of satisfaction is at 4 while its maximum is at 6. Satisfaction has less impact on loyalty for the low in satisfaction (L=1.84 + 0.35 SAT). In the second group, however, satisfaction has a greater impact on loyalty (L= -.97 + 1.70 SAT). In the third group, loyalty increases at a decreasing rate (L=1.84 + 0.65SAT). This is consistent with the model’s predictions.

A Quadratic Negative Model and a Linear Model of the SatisfactionBLoyalty Relationship. The qudratic negative model, which is also characteristic of the decreasing effects, fits the insurance data better than any other model. The maximum of loyalty (repurchase intentions) is reached at a satisfaction score of 5.37. This implies that, in this sample, when customer satisfaction has reached a score of 5.37, increasing customer satisfaction beyond that point would not increase loyalty further. Our results indicate, however, that satisfaction has a linear relationship with customer loyalty for the retailer data.

In sum, our data supports three different models: the two-threshold model, the quadratic negative model (i.e. with decreasing returns) and the linear model. What might explain this difference?

TABLE 5

THE ESTIMATION RESULTS OF EQUATION 4

An Attempt at Explanation

A theoretical explanation? The most striking fact here is the inability of the proposed model to fit the retailer data. We should however bring a little more precision for a better comprehension of this relation. This linear relationship is obtained for the individuals living in a medium-sized city where the retailer investigated is the main outlet, established 4 months ago, covering a relatively large area. Indeed, the establishment of supermarkets is the subject of a strong regulation in this country. As such, and consistent with Anderson’s (1994) results, we hypothesize that repurchase intentions here are probably more a problem of switching costs than of real customer satisfaction. This is somewhat confirmed by the variance explained by satisfaction (38%), indicating that more is probably due to the lack of competitive alternatives in the area.

A statistical explanation? The other models reveal the existence of decreasing returns or ceiling effects of satisfaction on repurchase intentions. However, these ceiling effects are represented by two different models: the two-threshold model and the quadratic negative model. We believe that this is due to the sample size. One can note that as the sample size increases, the significance of the parameters in equation 4 also increases. For example, in the insurance data (n=53) not all the parameters are significant while they are slightly in the bank data (n=73) and quite significant in the camera data (n=224). This however needs to be investigated further.

CONCLUSION

The question we were addressing in this paper was whether or not firms should delight or provide 100% satisfaction to their customers to increase customer loyalty. An examination of the literature and the empirical results indicate that care should be taken. The proposed model argues that there exists a minimum threshold to be reached for satisfaction to have greater positive effects on loyalty. This was supported in the bank and the camera data. This study also shows that there exists a linear relationship between customer satisfaction and loyalty. However, this linear relationship is important only within a specific segment of satisfaction or when customers do not have a large choice set. For example, our results indicate that the individuals in the retailer data are likely to go back to this supermarket because of the distance between the various supermarkets in this area.

Managerial Implications. Most significant in this study is the fact of observing that satisfaction can exhibit decreasing returns on loyalty. In this sense, our study is consistent with all the authors who defend this position (Fornell et al. 1996b; Auh and Johnson, 1997). This indicates that at very high levels of customer satisfaction (i.e. customer delight or 100% satisfaction), loyalty is likely to level off. This contradicts the executives who argue that quality must be taken beyond customer satisfaction to customer delight. The findings therefore indicate that managers who conside that the relation between customer satisfaction and loyalty is linear might be surprised if their investment on satisfaction does not result in a proportional increase in customer retention. A linear approach to the satisfaction-loyalty issue may result in an underestimation of the impact of satisfaction on loyalty or an overestimation of the impact of customer delight on loyalty (Anderson and Mittal, 1997). The manager who views the effect of satisfaction on loyalty linearly may think that it is not important to invest in customer satisfaction or that it is better to delight customers. However, if managers consider the existence of minimum and maximum thresholds of satisfaction, they could allocate their resources better and reap the benefits of satisfying their customers. For this reason, it is important to identify segments of customers in terms of their level of satisfaction. Managers should turn the low in satisfaction into the moderate in satisfaction and keep them there as it is at this level that satisfaction has a greater impact on loyalty. Note that this is consistent with Bolton’s (1998) proposition that organizations should focus on customers in the early stages of the relationship. If customers’ experiences are not satisfactory, the relationship is likely to be short because the customers would not have seen any reason to stay with the firm. Our results therefore contradict those studies which recommend systematically that firms improve satisfaction of all the customers (Jones and Sasser, 1995) or that firms delight their customers to improve customer retention (Oliver et al. 1997). This is not to say that customers should never be delighted. Rather, we mean that managers should always assess the costs and the benefits of such investments and therefore they have to take the "customer delight fashion" carefully.

Limitations and Suggestions for Future Research. Our analyses concern only four industries and bear on convenience samples. Although the findings are encouraging for further research and most of them join previous studies (Fornell et al. 1996b; Auh and Johnson, 1997), they need to be questioned before their generalizability is established. The next step should be an investigation of these relationships across different industries and customers. One can expect that the minimum and maximum thresholds of satisfaction or the slopes of the relationships within the same thresholds will differ depending on the competitive intensity of each industry or individual difference variables such as involvement. Moreover, researchers could investigate the differential drivers of satisfaction in each segment of customers. For example, do customers utilize the same comparison standards in every segment (Ngobo, 1997)? how does customers’ trust varies across these segments? Finally, in this research we estimated the models by the ordinary least squares method. However, this method assumes that the data has a normal distribution. What seldom is the case for customer satisfaction data. An estimation by nonparametric methods such as the Kernel regression or the stochastic spline regression seems to be an interesting approach insofar as it does not make assumptions on the distribution of the data.

REFERENCES

Anderson, Eugene Walter (1994). "Cross-category variation in customer satisfaction and retention," Marketing Letters, vol. 5 (1), 19-30;

Anderson, Eugene Walter (1998), "Customer Satisfaction and word-of-mouth," Journal of Service Research, vol. 1(1), 1-14.

Anderson, Eugene Walter and Vikas Mittal (1997). "The SatisfactionBProfit Chain: How solid are the links?" Working paper The National Quality research Center, The University of Michigan Business School

Auh, Seigyoung and Mihael D. Johnson (1997). "The complex relationship between customer satisfaction and loyalty for automobiles" In Johnson Michael D., Andreas Hermann, Frank Huber, and Anders Gustafsson (eds.), Quality, satisfaction and Retention: Implications for the Automotive Industry, Wiesbaden, Germany: Gabler.

Bearden, William O. and Jesse E. Teel (1983), "Selected determinants of consumer satisfaction and complaint reports " Journal of Marketing Research, vol.20 (February), 21-28;

Bitner, Mary Jo and Amy R. Hubbert (1994). " Encounter satisfaction versus overall satisfaction versus quality: the customer’s voice " In Service quality: New directions in theory and practice (Eds.) Roland T. Rust and Richard L. Oliver, Sage Publications, London,72-94;

Bolton Ruth (1998), "A Dynamic Model of the Duration of the Customer’s Relationship with a Continuous Service Provider: The role of satisfaction" Marketing Science, vol. 17 (1), 45-65;

Coyne, K. P. (1989), "Beyond service fadsBMeaningful strategies for the real world" Sloan Management Review, vol. 30, 69-76;

Dick Alan S. and Kunal Basu (1994). "Customer loyalty: Toward an integrated conceptual framework" Journal of Academy of Marketing Science, vol. 22 (2), 99-113;

Fornell, Claes (1992). "A National Customer Satisfaction Barometer: The Swedish Experience" Journal of Marketing, vol.56 (Jan.), 6-21;

Fornell, Claes, Michael D. Johnson, Eugene W. Anderson, Jaesung Cha and Barbara Everitt Bryant (1996a), "The American Customer Satisfaction Index: Nature, purpose, and Findings" Journal of Marketing, vol.60 (October), 7-18;

Fornell, Claes, Christopher D. Ittner and David F. Larcker (1996b), "The valuation consequences of customer satisfaction" Working Paper, The National Quality Research Center, The University of Michigan Business School;

Fournier Susan (1998), "Consumers and Their Brands: Developing Relationship Theory in Consumer Research" Journal of Consumer Research, vol. 24 , March, 1- 18;

Geyskens Inge, Jan-Benedict E. M. Steenkamp. Lisa K. Scheer and Nirmalya Kumar (1996). "The effects of trust and interdependence on relationship commitment: A Trans-Atlantic study" International Journal of Research in Marketing, vol. 13, 303-317;

Hallowell, Roger (1996), "The relationship of customer satisfaction, customer loyalty, and profitability: an empirical study" International Journal of The Service Industry Management, vol. 7 (4), 27-42;

Hausknecht Douglas (1988), "Emotional measures of satisfaction/dissatisfaction" Journal of Consumer Satisfaction, Dissatisfaction and Complaining Behavior, vol.1, 25-33;

Hunt H. Keith (1977). "CS/DBoverview and future research directions" In Conceptualization and Measurement of Customer Satisfaction and Dissatisfaction, H. Keith Hunt (ed.), Cambridge, MA: Marketing Science Institute, 455-488;

Jones Thomas O. and W. Earl Sasser, Jr. (1995). "Why Satisfied Customers Defect?" Harvard Business Review, November-December, 88-99;

Mendelsohn James D. (1998), "Moving from Satisfaction to 100% Guest Delight-A Look at The Olive Garden’s Use of Guest Satisfaction Research to Improve the Dining Experience" paper presented at the 10th Annual Customer Satisfaction and Quality Measurement Conference, Maximizing Results: From Data to Action , Atlanta, GA (March)

Mittal Vikas, William T. Ross Jr., and Patrick M. Baldasare (1998), "The asymmetric Impact of Negative and Positive Attribute-Level Performance on Overall Satisfaction and Repurchase Intentions" Journal of Marketing, vol. 62 (January), 33-47;

Narayandas, Narakesari (1996). "The link between customer satisfaction and customer loyalty: An empirical investigation" Working Paper, Division of Research, Harvard Business School.

Ngobo, Paul V. (1997) "The Standards issue: An Accessibility-Diagnosticity Perspective", Journal of Consumer Satisfaction, Dissatisfaction and Complaining Behavior, 10, 61-79;

Oliva, Terence A., Oliver, Richard L. and Ian C. Macmillan (1992), "A catastrophe Model for Developing Service Satisfaction Strategies" Journal of Marketing, vol. 56 (3), 83-98;

Oliver, Richard L. (1997), Satisfaction: A behavioral perspective, New York: MacGraw-Hill,

Oliver, Richard L., Roland T. Rust and Sajeev Varki (1997). "Customer Delight: Foundations, Findings and Managerial Insight" Journal of Retailing, vol. 73 (3), 311-336;

Sambandam, Rajan and Kenneth R. Lord (1995). "Switching behavior in automobile markets: A consideration-sets model" Journal of the Academy of Marketing Science, vol. 23 (1), 57-65.

----------------------------------------