Determining your share

Editor’s note: Timothy L. Keiningham is senior vice president, head of consulting, Tiffany Perkins-Munn is vice president, Demitry Estrin is senior analyst, and Terry G. Vavra is chairman emeritus, at Ipsos Loyalty, Inc., Paramus, N.J.

Linkage and validation efforts have become increasingly commonplace in customer satisfaction studies and the need continues for the identification of key drivers. All of these initiatives require a criterion variable against which to calibrate the performance variables. In this article we discuss the value and practicality of several commonly used criteria specifically in business-to-business surveys. We conclude that actual share of spending is the most appropriate variable for this purpose.

Overview

Customer satisfaction programs are undertaken in the belief that the information collected is correlated with business results. As a result, the measurement of customer satisfaction has become a basic component in the management of firms’ customer relationships, resulting in a pervasive number of company surveys. Recognizing this intense interest, even industry-wide and national measures have been created. (In particular, the reader may be familiar with the University of Michigan’s and the American Society for Quality’s American Customer Satisfaction Index [ACSI] which is published on a quarterly basis by these two sponsors. The survey tracks American consumers’ satisfaction with specific companies in 11 business sectors.)

This correlation with business results has not been accepted on faith alone; it has been the subject of numerous investigations (e.g., Anderson and Sullivan 1993, among others). And, the relationship has been exploited for three needs we can aggregate under the term “calibration.” These calibration needs include:

  • the identification of key drivers;
  • management’s need to see a satisfaction model validated; and
  • linking the attitudes with magnitudes of business outcomes for ROI allocation of improvement efforts.

The recent history of customer satisfaction has seen an escalation of investigations focusing on these needs. In general, the subsequent investigations have utilized the four different dependent variables described in the accompanying sidebar. These four variables can be divided into two categories: self-reported attitudes and observed behaviors.

Studies that utilize the attitudinal repurchase as a dependent variable should be acknowledged as essentially “closed loop” analyses. That is, in key driver analyses using attitudinal repurchase, the performance measures are selected (as key drivers) based upon their ability to predict (or correlate) with an attitude: repurchase intention. Unfortunately, this intention is generally collected at the same time, in the same questionnaire - a debilitating weakness! To control for multicollinearity (which is obviously present in such closed-system models), the practitioner may impose a hierarchical structure on the outcomes, designating some variables as subordinate and others as supraordinate. Such controls notwithstanding, using a variable internal to the attitudinal dataset is hardly the best practice for determining key drivers.

The second dependent variable, an observation of retention, is far better because it is not an attitude but rather a business behavior. Unfortunately we’ve found it is far from perfect. This is because, in many business-to-business situations, organizations will maintain business relationships with several vendors within a category as a defensive plan. If any one vendor disappoints the organization or comes into financial difficulties, others will be immediately available to make up the difference.

Observations of behavioral spending and share of spending (the third and fourth variables listed in the sidebar) have become increasingly acknowledged in the satisfaction literature. While both are measures of volume, current spend (sales) is less sensitive to business dynamics. Zeithaml (2000) noted “the growing popularity of the concept of share-of-wallet” over sales alone. Others have similarly noted the difficulty of using spend (sales) as a criterion measure. With sales, it’s totally clear when a customer defects, but it is less obvious when a customer shifts some of its purchases to another supplier, or the customer actually increases its category spending, but maintains its level of spending (resulting in a smaller share-of-wallet).

We propose share-of-wallet as the most sensitive and therefore most appropriate dependent variable. This is easiest to advocate in a business-to-business context, since businesses generally keep better data in b2b situations than with consumers. Reasons are numerous: the higher value of each customer; a more intimate relationship with the customer; and larger transactions (and subsequent margins). These conditions all support collecting more information about customers.

While share-of-wallet makes perfect sense as a dependent variable for calibration needs, rigorous proof is lacking. One of the problems has been the inability to actually collect share-of-wallet data. Another, as Zeithaml (2000) noted, is the lack of a definition and metrics for share-of-spending. The little research that does exist relies on the use of self-reported measures of share-of-wallet contained within a satisfaction measurement questionnaire (DeWulf, Odekerken-Schroeder and Iacobucci 2001).

Self-reported share-of-wallet measures are surely not much better than attitudinal repurchase intention. As attitudes about behavior, they are still attitudes. They are further subject both to inaccuracy and to the bias of being administered within the same attitudinal instrument used to collect the satisfaction attitudes. There is need for a disciplined review of the relationship between customers’ level of satisfaction and customers’ actual share-of-wallet. This article reports the findings from one such review. We address many of these issues by examining the relationship between customer satisfaction and share-of-wallet for a large financial institution.

Descriptions of the dataset

A large financial institution provided the dataset used in this analysis. The firm operates in a highly competitive market, with several very large global competitors, and a larger number of smaller, niche-oriented competitors. The vast majority of the firm’s clients are themselves financial institutions.

The sample was composed of individuals at customer organizations who were verified as participating in the decision of whether or not to use the financial institution. The sampled customer organizations represented the largest spenders in terms of total revenues within the product category for the financial institution and its competitors. Responding customers participated in a 20-minute telephone interview. In total, there were 307 completed interviews.

The questionnaire included questions tailored to address specific departmental performance issues and a core set of customer satisfaction questions.

Share-of-wallet defined

The share-of-wallet information used in this study is defined as the ranking of relative spending to other that a customer organization maintained with all financial institutions in the category. The base for this relative measure was the volume of total business in the product category conducted by the client organization within a 12-month period (i.e., 1 denotes the highest percentage, 2 the second-highest percentage, etc.). An independent, third-party source collects this information from the financial institution and its competitors and then provides all institutions with information on their relative share-of-wallet.

Data analysis

In our investigation, we recognize the work of two of the authors (Keining-ham and Vavra 2001) who have written about the relationship between performance inputs and satisfaction outputs. They have proposed, as have other authors (Rust and Zahorik 1993, Mittal and Kamakura 2001, among others) that responses in satisfaction data are best depicted in non-linear, asymmetric relationships. To explore the relationship between satisfaction and share of wallet while accommodating the likely nonlinear relationship between satisfaction and share-of-wallet we used a non-linear (cubic regression) but we also report the results of simple, linear regressions.

Keiningham and Vavra (2001) have suggested dealing with the nonlinearity of satisfaction data by breaking the data into response zones. Consequently an initial analysis was conducted using chi-square automatic interaction detector (CHAID). Overall satisfaction levels were used as a predictor of share-of-wallet (SOW). The results confirmed that the optimal predictive relationship occurred when the satisfaction variable was partitioned into two zones: 1) satisfaction levels <8; and 2) satisfaction levels >9 (see Figure 1). Indeed, for the two zones, SOW was found to increase significantly from 10.8 in Category 1 to 15.4 in Category 2 (adjusted p-value = .00; F = 21.0). These CHAID results indicate that the relationship between satisfaction and SOW is positive, nonlinear and asymmetric.

In addition, a cubic (curvilinear) regression model was applied. The fit accomplished by the cubic-shaped CS-SOW model was based on an adjusted R2 and the Sp criterion of Breiman and Freedman (1983). The linear and cubic models tested are represented as follows:

(1)  SOWij = _ + _CS CSij + _ij

(2)  SOWij = _j + _CS CSij + _CS CS2ij + _CSCS3ij + _ij

where: SOWij = estimated share-of-wallet by customer i at company j, CSij = customer satisfaction rating for customer i at company j, _ = SOW when CS = 0, _ CS = regression coefficient for the corresponding predictor term in the model, and _ = the residual or error term for the model.

Looking across all buyer groups, the correlation between customer satisfaction and SOW increases from r = .19 in the linear model to r = .27 in the cubic model (see Table 1). Furthermore, the fraction of variability (adjusted R2) in SOW that is explained by customer satisfaction doubles from 3 percent in the linear model to 7 percent in the cubic model!

Segmenting the decision makers

Our data consisted of ratings from within customers’ organizations, and therefore from customer representatives in various positions of responsibility. Three major classes were included in this dataset: Customer Group 1 (information managers), Customer Group 2 (product managers), and Customer Group 3 (purchasing agents).

Pearson correlations were examined for each of the customer groups to provide an initial examination of a possible linear relationship (see Table 2). The results revealed that the linear correlation between satisfaction and SOW was not statistically significant for Customer Group 1 (information managers), while the correlations for Customer Group 2 (product managers) and Customer Group 3 (purchasing agents) were statistically different from zero, having correlations of .24 and .23 respectively. The lack of a significant correlation for CG1, versus significant correlations for CG2 and CG3, supports the general belief that within organizations, executives in different roles impact a purchasing decision differently as shown by the different relationships between satisfaction and SOW by customer group.

Implications for practice

The key implication is that managers should not simply strive to improve reported satisfaction levels without an understanding of the relationship to customer behavior measures, in this case share-of-wallet. Given the non-linearity of the relationship, it is critical that managers be certain that efforts designed to improve satisfaction do so in sufficient force so as to reach satisfaction levels that correspond with increasing share-of-wallet levels.

Furthermore, simply treating all buyers as homogenous has the potential to misrepresent the relationship between satisfaction and share-of-wallet. Therefore, it is important that the level of influence of various buyer groups on purchase behavior be uncovered so that resources can be appropriately allocated to those areas providing the greatest impact.

In the case of the firm that provided this data, managerial experience and prior research supported a disproportionate weighting to Customer Group 3 (purchasing agents) because this group actually selects the firm from which to purchase. The results of this research, however, indicated that the satisfaction level of Customer Group 2 (information managers) fits the data equally well as for Customer Group 3. As a result of this and other recent information supporting these findings, the firm has made a concerted effort to dramatically alter the content of the information it provides regarding its product offerings.

Summary

The analysis reported here advances the empirical research regarding the intuitive relationship between customer satisfaction and business outcomes in two key ways. First, our findings indicate that satisfaction is positively related to the share of business a customer conducts with a particular service provider (share-of-wallet), as opposed to simply repurchasing a product or service at some point in the future, or continuing to keep a business relationship with a service provider.

Second, the findings suggest that the relationship between satisfaction and share-of-wallet is nonlinear. This corresponds to similar findings regarding the relationship between satisfaction and repurchase intention, retention, and word of mouth.

The findings also contribute to the research regarding organizational buying. The findings suggest that the relationship between buyer group satisfaction and share-of-wallet is nonlinear, and that the functional form of the relationship varies by segment.

Article Sidebar

Four measures are most frequently used as the dependent variable in the calibration of customer satisfaction studies:

1. A traditional, attitudinal repurchase intention question. Typically satisfaction surveys include a question probing the likelihood of repurchase along with two companion questions: willingness to recommend and overall satisfaction.
2. A behavioral retention observation. Results have been aligned with customer presence or “persistence.” Does the customer continue to transact with the organization? This observation is generally far more available than more demanding sales measures.
3. A behavioral spending observation. Data from company records can be gathered to show how much surveyed customers have spent, are currently spending or spend subsequent to administration of a satisfaction survey.
4. A behavioral share of spending observation. Measures can be created to show the category spending of surveyed customers that is allocated to the surveying organization.

 

References

E.W. Anderson and M.W. Sullivan, The Antecedents and Consequences of Customer Satisfaction for Firms, Marketing Science 12 (Spring 1993): 125-143.

Breiman, L. and D. Freedman (1983), How Many Variables Should Be Entered in a Regression Equation? Journal of the American Statistical Association, 78 (381), 131-136.

De Wulf, Kristof, Gaby Odekerken-Schröder, and Dawn Iacobucci (2001), Investments in Consumer Relationships: A Cross-Country and Cross-Industry Exploration, Journal of Marketing, 65 (October), 33-50.

Keiningham, Timothy L. and Terry G. Vavra (2001), The Customer Delight Principle, New York: McGraw-Hill.

Mittal, Vikas and Wagner Kamakura (2001) Satisfaction, Repurchase Intent and Repurchase Behavior: Investigating the Moderating Effect of Customer Characteristics, Journal of Marketing Research, 38 (February), 131-142.

R.T. Rust and A.J. Zahorik, Customer Satisfaction, Customer Retention, and Market Share, Journal of Retailing 69, no. 2 (1993): 193-215.

Zeithaml, Valarie A. (2000), Service Quality, Profitability, and the Economic Worth of Customers: What We Know and What We Need to Learn, Journal of the Academy of Marketing Science, 28 (1), 67-85.