Editor’s note: Bill Etter is vice president, director of research at Rockwood Research, St. Paul.
Customer satisfaction research continues to be a hot topic. Conferences and seminars focus on it. Marketing research journals devote entire issues to it. Research firms claim expertise in it. But there is increasing awareness that measuring customer satisfaction is not enough (see for example, Jones and Sasser (1995), and Pruden, et al (1996)). The concepts of loyalty and brand equity are mentioned more and more as important components for customer retention. This article will focus primarily on the concept of loyalty. In particular, we consider loyalty as a continuum and suggest that it is important to look at both ends of the loyalty spectrum. In the process of looking not only at the most loyal customers, but also the least loyal, we define, in a somewhat different fashion than some other researchers, the concept of penalty and reward attributes and introduce the concept of power attributes. These concepts are illustrated with data from a real world study.
Loyal and at-risk customers defined
Researchers often define loyalty using one or some combination of three measures - overall satisfaction, intention to repurchase, and willingness to recommend. The work of Jones and Sasser (1995) is an example using a single measure, intent to repurchase, as a measure of loyalty. We have followed the lead of some others in using all three measures to define loyalty.
A recent article in these pages (Pruden, et al (1996)) suggested that using all three measures is, "... not only simplistic, it also lacks potential diagnostic value." Simplicity (which is related to parsimony) has long been recognized as a virtue in many areas of human endeavor, including analysis, and this article will suggest that there is a good deal of diagnostic power in defining loyalty using these three measures, especially if the measures are collected for key competitors.
Others have argued the relative merits of behavioral versus attitudinal based measures of loyalty (e.g., Jacoby and Chestnut (1978) and Pruden, et al (1996)). We will not review those issues here. What follows is essentially a real-world example of the diagnostic power that can be obtained from data collected on the three above mentioned measures.1 We would suggest that two of these measures are quasi-behavioral (intention to repurchase and willingness to recommend) and one is attitudinal (overall satisfaction).
The concept of loyalty can be thought of as representing a continuum. At one end of this continuum a company can find its most loyal customers; at the other end are the customers who are least loyal. (We do not consider here former customers who might consider being customers again.) We have chosen to give labels to the definitions of two types of customers: loyal customers and at-risk customers. Formally, we define loyal customers as those who rate a company a four or five (top two-boxes) on scales A, B, and C in Figure 1 and at-risk customers as those who rate a company a one or two (bottom two-boxes) on scales A, B, or C.
These measures are collected for all companies in a given respondent’s consideration set2. Our definitions of loyalty and at-risk correspond roughly to the labels loyalist/apostle and defector/terrorist used by Jones and Sasser (1995). Their paper focuses primarily on the link between satisfaction and loyalty. Our focus in this article is primarily on the diagnostic link between perceived levels of performance on individual product/service attributes and levels of loyalty/at-risk.
Loyalty/at-risk assessment by company
Having competitive information allows loyalty and at-risk measures to be compared across competing brands or companies in a given market as shown in Figure 2.
In this category the ratio across companies between highest to lowest loyalty is 2.4 (41/17) and the corresponding at-risk ratio is 2.0 (38/19). Thus in this category there are considerable differences in the proportion of loyal and at-risk customers across companies. Net loyalty, the difference between the percent of loyal customers and the percent of at-risk customers for a given company, is a measure of the commitment of a company’s customer base. Table 1 shows that this commitment varies widely across companies in our example.
Table 1
New Loyal Customers
Company |
Net Loyal % |
||||
A |
13 |
||||
B |
8 |
||||
C |
-16 |
||||
D |
8 |
||||
E |
-15 |
||||
F |
-19 |
||||
G |
-12 |
||||
H |
19 |
While we haven’t established it empirically, we would hypothesize that to the degree a company has a less committed customer base (i.e., has more at-risk customers than loyal customers) that company is vulnerable to an erosion of their customer base through brand switching behavior. We would also hypothesize that the net loyal measure of company commitment in the above table is positively correlated with a company’s brand equity as discussed, for example, by Park and Srinivasan (1994).
The correlation between loyal and at-risk measures is, for this example, -.59 (r2 = .35) suggesting that while there is a negative relationship between the measures as expected, knowing one does not do a very good job of explaining the other.
No company wants to lose customers. The normally higher cost of gaining new customers versus keeping old customers is well understood (for a discussion see Engel et al, 1993, p. 575). At-risk customers represent a class of customers that needs reassurance about the value of remaining a customer. At the same time it is important to know what not to mess up for loyal customers. The idea of penalty and reward attributes as defined below can help a company develop strategies for boosting the confidence of at-risk customers and keeping loyal customers loyal.
The nature of reward, penalty and power
Our motivation is to answer the question, "What is the gain or reward, if any, for moving customers into the highest levels (top two boxes) of satisfaction on a given attribute, and, conversely, what is the loss or penalty, if any, of having customers move into the lowest levels (bottom two boxes) of satisfaction on a given attribute?"
Suppose we look at all customers who say a given brand or company is better than they require (top two boxes)3 on a given attribute and calculate the percent of these customers who meet our loyalty definition. Next, calculate the percent of loyal customers among all customers who say the brand or company is meeting their requirements on the attribute and subtract this percent from the first percent. If this difference in percentages is relatively high we can infer that high performance on this attribute is associated with a relatively large gain in brand or company loyalty. Stated another way, delighting customers on such an attribute offers the reward of increased loyalty.
Conversely, we can look at all customers who say a given brand or company falls short of their requirements (bottom two boxes) on a given attribute and calculate the percentage of these customers who meet our definition of at-risk. Subtract from this percent the percent of at-risk customers among all customers who say the brand or company is meeting their requirements on the attribute. An attribute with a relatively high difference in these percentages suggests that poor performance on this attribute is associated with a relatively large increase in at-risk customers, or, alternatively, that disappointing customers on such an attribute results in a penalty of increased at-risk customers.
Tables 2 and 3 show the reward and penalty calculations, respectively, for one brand from our example4. In this study there were 23 attributes. For the sake of brevity in Tables 2 and 3, we show results of the calculations only for the highest and lowest scoring attributes.
We focus on the Difference columns in Tables 2 and 3. It is clear that attributes where delighting customers (better than requirements) is associated with large loyalty gains are not necessarily the same attributes where disappointing customers (falls short of requirements) is associated with large increases in being at-risk and vice versa. For example, attribute e, communicates customer programs, is fifth highest in Table 2, but last in Table 3. So for this attribute, delighting customers is associated (rewarded) with relatively large gains in loyalty (60 percentage points), but disappointing customers is not associated (penalized) with large increases in the percent of customers being at-risk (6 percentage points). The opposite relationship exits for attribute u, company stands by its products, which is second highest in Table 3 and third lowest in Table 2. However, some attributes like a and b are high in both tables. A quadrant map based on the data in these two tables can give a better picture of the situation.
By plotting data from the Difference columns in Tables 2 and 3 and marking the horizontal and vertical medians, we get the quadrant map shown in Figure 3. The letters on the map correspond to the attributes listed in Tables 2 and 3. Those attributes above the horizontal median we call reward attributes; those to the right of the vertical median we label penalty attributes.
The four quadrants, beginning in the upper left quadrant and proceeding in a clockwise direction, have been labeled, respectively, reward, power, penalty and neutral. Attributes that fall in the power quadrant are those that lead to large increases in customer loyalty when customers experience delight (better than I require) and also lead to large increases in number of at-risk customers when customers experience disappointment (falls short of what I require). Thus these attributes carry significant power or leverage at both ends of the scale; they are both penalty and reward attributes. They have the potential to make (reward) or break (penalize) a company or product. In this example, these attributes are mostly associated with characteristics of the company’s sales reps (a, b, c, f, k) and product performance (d). For example, consider attribute a, rep’s knowledge. Customers who perceive their rep’s knowledge to be better than they require are much more likely to be loyal customers than are customers who perceive their rep’s knowledge to be just meeting requirements, and, conversely, customers who perceive their rep’s knowledge to fall short of their requirements are much more likely to be at-risk customers than are customers who perceive their rep’s knowledge to be meeting requirements. Thus the relationship between rep and customer is key to this company having loyal or at-risk customers. The same rep, of course, could have both loyal and at-risk customers.
Companies with significant numbers of at-risk customers will want to look carefully at the penalty attributes as providing opportunities for shoring-up the confidence of their less-loyal customer base. In particular those penalty attributes that are in the power quadrant would have first priority, everything else being equal (e.g., cost of implementing change on a given attribute); those penalty attributes not in this quadrant (i.e., in the penalty quadrant) would have secondary priority. Not disappointing customers on these attributes will lead to fewer numbers of at-risk customers.
Companies hoping to gain a more loyal customer base will want to concentrate efforts on the reward attributes. Delighting customers on these attributes will lead to increased numbers of loyal customers. Similar to dealing with penalty attributes, reward attributes in the power quadrant would have highest priority everything else equal; secondary consideration would be given to reward attributes in the reward quadrant.
This quadrant map is for a single company. Each competing company in the market will have its own map. Chances are they will be similar, but differences may exist from company to company.
Conclusion and implications
We have demonstrated, by way of an example, the type of diagnostic information contained in the measures we have used to define loyal and at-risk consumers. Certain concepts have been defined and illustrated; for example, the concepts of net loyalty (commitment), penalty, reward and power attributes. Despite the potential diagnostic value of these and similar analyses, one limitation should be mentioned. There is no way to estimate the actual impact, in terms of gain or loss in market share, of the strategies suggested by these types of diagnostic analyses.
For example, in the above discussion of penalty and reward attributes we have not said anything about the relative payoff to the company of focusing efforts on one type of attribute versus the other. Certainly it would seem that companies with significantly more at-risk customers than loyal customers would want to move customers out of the at-risk category and therefore focus on strategies for improving performance on penalty attributes before turning to strategies for increasing the number of loyal customers. However, it might turn out that, in fact, a strategy of increasing the base of loyal customers would have an equal or even higher payoff for the bottom line. With the data available in typical customer satisfaction studies, managers can only hypothesize about the most effective strategy and use the type of diagnostic analyses presented here to increase their confidence that they are taking the best action. Similarly, in situations where it would seem the appropriate strategy is to focus on attributes that we have defined as reward attributes, the reality may be that improved perceptions on one or more penalty attributes may have an equal or better bottom-line payoff. How can managers increase their confidence in the decisions they make about where to allocate product or brand improvement resources?
We have written elsewhere (Etter, 1996) about the advantages of placing the entire customer satisfaction data collection methodology and analysis in the context of choice models. Such a system allows the estimation of the specific bottom-line implications of strategies focusing on penalty attributes versus strategies focusing on attributes suggested by some other analyses. In short, a choice model simulator allows a manager to explore the relative bottom-line impact of multiple customer satisfaction/loyalty improvement strategies regardless of the source of these strategies.
References
Etter, William, "Customer Satisfaction and Choice Modeling: A Marriage," Quirk’s Marketing Research Review, October, 1996, p. 14-16, 64-65.
Engel, James F., Roger D. Blackwell and Paul W. Miniard, Consumer Behavior. (Fort Worth: The Dryden Press, 1993).
Jacoby, Jacob and Robert W. Chestnut, Brand Loyalty Measurement and Management. (New York: Wiley, 1978).
Jones, Thomas O. and Earl W. Sasser, Jr., "Why Satisfied Customers Defect," Harvard Business Review, November-December, 1995.
Park, Chan Su, and V. Srinivasan (1994), "A Survey-Based Method For Measuring and Understanding Brand Equity and Its Extendibility," Journal of Marketing Research, (31) May, 271-288.
Pruden, Douglas R., Ravi Sankar, and Terry G. Vavra, "Customer Loyalty: The Competitive Edge Beyond Satisfaction," Quirk’s Marketing Research Review, April, 1996, p. 24, 49-53.
Notes
1 The data for this example was collected via a mail survey in August 1996. The average number of respondents profiling a particular company is 141.
2 The wording of these scales may have to be modified to fit the particular category under study.
3 We normally use five-point requirement scales in our customer satisfaction work and the ideas discussed here assume a five-point scale, but the type of scale (e.g., requirement, expectation, satisfaction) is not critical. Adjustments to the procedures described here can be made for scales containing fewer or more than five points.
4 An alternative notion of penalty and reward attributes is based on the data in a modified version of Table 2. Suppose that in addition to the difference in column three we had a second difference formed by subtracting from the percentages in column two (meets requirements) the percentage of loyal customers among all customers who perceive company B to fall short of requirements (bottom two boxes). We could then compare these two differences. Roughly speaking, the first difference, when large compared to the second difference for a given attribute, defines a reward attribute; conversely, the second difference, when large compared to the first difference, defines a penalty attribute.