The Impact of Customer

Satisfaction on Share-of-Wallet in a
Business-to-Business Environment
Timothy L. Keiningham
Tiffany Perkins-Munn
Marketing Metrics

Heather Evans

Research has found that customer satisfaction is positively Both practitioners and academics have accepted the prem-
related to repurchase intention, actual repurchase, market ise that customer satisfaction results in customer behavior
share, and word of mouth. There is growing recognition patterns that positively affect business results (Kotler
among managers of the importance of measuring the share 1994; Rust and Oliver 1994; Vavra 1997). As a result, cus-
of business a customer conducts with a particular service tomer satisfaction measurement has become a basic con-
provider (share-of-wallet) as opposed to simply repur- struct in the management of firms’ customer relationships,
chasing a product or service at some point in the future or resulting in a large number of company, industry, and even
continuing to keep a business relationship with a service national satisfaction measures (Fornell 1992; Fornell et al.
provider. Currently there is very little empirical research 1996).
concerning the relationship between customer satisfac- A number of studies support the validity of such a link-
tion and share-of-wallet. This article examines the rela- age. Research has found that customer satisfaction has a
tionship between satisfaction and actual share-of-wallet measurable impact on purchase intentions (Bolton and
in a business-to-business environment. The authors not Drew 1991), on customer retention (Mittal and Kamakura
only find that there is a positive relationship but that the re- 2001), and on financial performance (Anderson, Fornell
lationship is nonlinear, with the greatest positive impact and Lehmann 1994; Keiningham et al. 1999; Leung, Li,
occurring at the upper extreme of satisfaction levels. and Au 1998; Zahorik and Rust 1992).
In addition to these measures, there is a widely held be-
Keywords: customer satisfaction; share-of-wallet; reten- lief among both researchers (Sheth, Sisdodia, and Sharma
tion; loyalty; consumer behavior 2000; Zeithaml 2000) and managers (Brookes and Stodin
1995; Fred 2002; Wiegran and Koth 1999) that a link exists
between customer satisfaction and “share-of-wallet.”
Sheth, Sisdodia, and Sharma noted that the “effectiveness
How important is customers’ satisfaction with product [of customer-centric marketing] entails the enhancement
and service quality and customers’ repurchase behavior? of customer retention and ‘share-of-wallet.’” In addition,

The authors thank the staff of Marketing Metrics for their review and comments. The authors also thank Dr. Bruce Cooil for his insight
and advice.
Journal of Service Research, Volume 6, No. 1, August 2003 37-50
DOI: 10.1177/1094670503254275
© 2003 Sage Publications

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

38 JOURNAL OF SERVICE RESEARCH / August 2003

for many in the business community, share-of-wallet is OVERVIEW OF THE

thought to better reflect customer retention rates than actu- FIRM-CLIENT RELATIONSHIP
ally measuring repurchase and/or the continued mainte-
nance of a business relationship between customer and A large financial institution provided the data used in
firm. Reichheld (1996) has explained the logic as follows: this analysis. Because of the proprietary nature of the in-
formation, the strategic importance placed on the results,
Defining Defection. Some customer defections are and the structural changes occurring within the organiza-
easier to spot than others. Customers who close their tion because of this analysis, the firm has asked that it re-
accounts and shift all their business to another sup- main anonymous and that its business be described in
plier are clearly defecting. But what about custom- general terms only.
ers who shift some of their purchases to another
The firm operates in a highly competitive market, with
supplier, and what about those who actually buy
more but whose purchases represent a smaller share a few very large global competitors and a larger number of
of their total expenditures (a smaller share-of- smaller, niche-oriented competitors. With some notable
wallet)? (p. 56) exceptions, most “products” offered by the firm can be
thought of as the bundling of information and service com-
Currently, however, the relationship between share-of- ponents to facilitate the purchase of a core product /service
wallet and customer satisfaction is accepted as a truism that is provided by all competitors. Such “products” are
without rigorous empirical research to support the rela- true of a number of service providers—for example, the
tionship. One of the problems has been the inability to ac- bundling of enhancements such as insurance, free checks,
tually collect share-of-wallet data. Another, as Zeithaml and so on to a checking account or credit card.
(2000) noted, despite “the growing popularity of the con- The vast majority of the firm’s clients are themselves fi-
cept of share-of-wallet,” the term itself “requires both defi- nancial institutions. The scope of the firm’s operations,
nition and metrics.” The little research that does exist relies however, allows it to provide sophisticated global products/
on the use of self-reported measures of share-of-wallet services and advanced analytics to clients that would be
contained in a satisfaction measurement questionnaire (De too difficult and too expensive for most clients to bring
Wulf, Odekerken-Schröder, and Iacobucci 2001). in-house.
As such, the self-reported share-of-wallet measures are As is typical of a number of business-to-business rela-
similar to the use of repurchase intentions questions com- tionships, users of the products/services offered by the
monly contained in customer satisfaction questionnaires. firm represent different departments within the client or-
There is a wide body of research investigating the relation- ganization, with very different interests and motivations
ship between repurchase intentions and actual repurchase (Anderson, Chu, and Weitz 1987; Cannon and Homburg
(Bemmaor 1995; Jamieson and Bass 1989; Manski 1990; 2001; Cannon and Perreault 1999; Kohli 1989; Sheth
Morrison 1979; Morwitz, Johnson, and Schmittlein 1993). 1973; Vyas and Woodside 1984; Webster 1978). The three
Although this research tends to support a positive relation- primary client user groups are the following: (a) an infor-
ship between repurchase intentions and actual repurchase, mation department, (b) a product management depart-
the strength of the relationship has been found to vary ment, and (c) a purchasing department. The primary role
widely, depending on such things as product category and of the information department is to provide intelligence to
purchase cycle, among others (Morwitz, Steckel, and the product management department as to which products/
Gupta 1997). Furthermore, one would expect that mea- services appear to represent the best long-term prospects
sures contained within a single questionnaire concerning for the client. The product management department deter-
satisfaction levels, likelihood of repurchase, and estimated mines which products best fit the client’s strategic objec-
share-of-wallet would be positively correlated. tives. The purchasing department receives orders for a
Therefore, the need exists for an examination of the re- particular product from the product management depart-
lationship between customers’ level of satisfaction and ment and selects the firm from which to purchase it.
customers’actual share-of-wallet. This research addresses
this need by examining the relationship between customer
satisfaction and share-of-wallet for clients of a large finan- CLIENT SATISFACTION MEASUREMENT
cial institution (hereafter referred to as “firm”). Share-of-
wallet information is the proportion of the actual share of The data for this study were collected as part of a larger,
business (in dollars) a client allocates to the firm. ongoing telephone customer satisfaction survey of clients

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 39

of the firm. The larger program consists of more than 30 performance (Anderson and Fornell 1999; Rust and
departments, with a range of 50 to approximately 500 re- Zahorik 1993), and word of mouth (Anderson 1998).
spondents per department. Client respondents, for a given The effects of customer satisfaction on customer be-
department, were randomly selected for participation in havior and business results are widely believed to be non-
the study from a larger pool of potential respondents linear and asymmetric (Anderson and Mittal 2000; Coyne
within the department. Each department was administered 1989; Jones and Sasser 1995; Keiningham and Vavra
a customized questionnaire, which included questions tai- 2001; Oliva, Oliver, and MacMillan 1992). The
lored to address the specific departmental issues and a core nonlinearity and asymmetry of the relationship between
set of customer satisfaction questions. Clients participated customer satisfaction and repurchase intentions has been
in a 20-minute telephone interview, which covered topics confirmed in a number of studies (Bloemer, Kasper, and
on various aspects of their satisfaction with products and Lemmink 1990; DeSarbo et al. 1994; Mittal and Baldasare
services provided by the firm. 1996; Mittal et al. 1998; Oliver, Rust, and Varki 1997).
The data presented here represent clients of one product Mittal and Kamakura (2001) found that this nonlinear and
area within the firm. The total of respondents was 348. Cli- asymmetric relationship also held true for customer repur-
ents surveyed represented one of three buyer groups: chase. Likewise, Anderson (1998) found the relationship
Buyer Group 1: Information Management, 110 respon- between customer satisfaction and word of mouth to be
dents; Buyer Group 2: Product Management, 149 respon- asymmetric and nonlinear. Therefore, we would expect
dents; and Buyer Group 3: Purchasing, 83 respondents. the following with regard to nature of the relationship be-
Six respondents could not be uniquely classified into a par- tween customer satisfaction and share-of-wallet:
ticular department. The 348 clients included as part of this
study represent the 55 top-tier accounts of the firm (e.g., Hypothesis 1: The relationship between customer satis-
the 55 client organizations that provided the greatest reve- faction and share-of-wallet will be positive, asym-
nue to the firm). metric, and nonlinear.
The 55 client organizations that comprise the data set
are themselves large financial institutions and as such are Mittal and Kamakura (2001) found that customer char-
largely homogeneous. There is no other demographic in- acteristics have a moderating effect on customer behavior.
formation, such as age, gender, and so on, available in the In all cases, however, they found the relationship between
data set. satisfaction, repurchase intention, and repurchase behav-
ior to be nonlinear. Their research, however, dealt with in-
dividual consumers, as opposed to organizations.
SHARE-OF-WALLET MEASUREMENT With respect to corporate buying behavior, various de-
partments or committees whose members have different
For the purpose of this study, share-of-wallet is defined interests and motivations frequently make organizational
as the percentage of the volume of total business con- purchase decisions (Sheth 1973). As a result, organiza-
ducted with the firm by a client organization within a 12- tional decision making represents a complex process (An-
month period. An independent, third party source collects derson, Chu, and Weitz 1987; Cannon and Perreault 1999;
this information from all financial institutions and then Cannon and Homburg 2001; Vyas and Woodside 1984;
provides these institutions with information on their rela- Webster 1978). A number of studies have examined why
tive share-of-wallet. particular individuals are able to exert more influence in
the purchase decision (Jackson, Keith, and Burdick 1984;
Leigh and Rethans 1984; McQuiston 1989; Spekman and
HYPOTHESIS DEVELOPMENT Stern 1979; Thomas 1982; Weigand 1966). Research has
found that influence in the purchase decision is related to
expertise (Spekman 1979), formal authority (Ronchetto,
Hypotheses Tested
Jutt, and Reingen 1989; Thomas 1982), and stake in the
Customer satisfaction has been found to be positively decision (Patchen 1974). Therefore, we would expect the
associated with customer behaviors that are beneficial to following with regard to the relationship between satisfac-
business outcomes. Research has found that customer sat- tion among the various buyer groups and share-of-wallet:
isfaction is positively associated with purchase intentions
Hypothesis 2: The relationship between customer satis-
(Anderson and Mittal 2000; Oliver 1997), customer reten-
faction and share-of-wallet will differ across buyer
tion (Anderson and Sullivan 1993; Bolton 1998), financial groups.

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

40 JOURNAL OF SERVICE RESEARCH / August 2003

ANALYSIS Barksdale 1980).1 Overall satisfaction levels were used as

a predictor of share-of-wallet (SOW).2 The results indi-
Many authors (e.g., Pedhazur 1997; Cohen and Cohen cated that the optimal predictive relationship occurred
1984; Feldman and Berry 1985) suggest three primary when the satisfaction variable was partitioned into two
ways to detect nonlinearity. The first method is the use of groups: (a) satisfaction levels ≤ 8 and (b) satisfaction levels
theory or previous research to inform current analyses. ≥ 9 (see Figure 1). Indeed, for the two categories, SOW
However, as is often the case, many researchers have disre- was found to increase significantly from 10.8 in Category
garded the possibility of nonlinear relationships. A second 1 to 15.4 in Category 2 (adjusted p-value = .00, F = 21.0).
and more preferable method of detection is examination of The results of the CHAID would indicate that the relation-
residual scatterplots (i.e., plots of the standardized residu- ship between satisfaction and SOW is positive, nonlinear,
als as a function of standardized predicted values). The and asymmetric.
third method of detecting curvilinearity involves routinely In addition, the relationship was tested by comparing
conducting curve estimation procedures to include various the estimation of a linear model versus the estimation of
curvilinear components (e.g., exponential and cubic several curvilinear models (i.e., power, cubic, exponential,
terms). It is important that the curvilinear aspects of the re- etc.).3 The best fit for the data was a cubic-shaped specifi-
lationship be accounted for in order to best assess the rela- cation of the CS-SOW (customer satisfaction—share-of-
tionship between variables (Osborne and Waters 2001). wallet) relationship based on adjusted R2 and the Sp crite-
For the purposes of our analyses, we used all three rion of Breiman and Freedman (1983) (see Table 1).4
methods: The linear and cubic models tested are represented as
follows:
1. During the past few years, researchers have be-
gun to address the nonlinear nature of the rela-
SOWij = α + βCSCSij + εij (1)
tionship between customer satisfaction and
revenue, profitability, and other measures of
company performance (Bernhardt, Donthu, and SOWij = αj + βCSCSij + βCSCS2ij + βCSCS3ij + εij, (2)
Kenneth 2000; Oliver, Rust, and Varki 1997;
Rust and Zahorik 1993). Mittal and Kamakura
(2001) have illustrated a curvilinear relationship where SOWij = estimated share-of-wallet by customer i at
between satisfaction and repurchase intentions. company j, CSij = customer satisfaction rating for cus-
In addition, Hitt, Bierman, Shimizu, and Kochhar tomer i at company j, α = SOW when CS = 0, βCS = regres-
(2001) illustrated the importance of considering sion coefficient for the corresponding predictor term in the
curvilinear models in the assessment of the effect model, and ε = the residual or error term for the model.
of human capital on professional service firm Equation (1) represents the relationship between SOW
performance. and customer satisfaction as a simple linear model. Equa-
2. A residual scatterplot of the relationship between tion (2) represents the relationship as a curvilinear (cubic)
customer satisfaction and share-of-wallet de-
model. Table 2 illustrates the results.
picted a positive association between the two
variables. The relationship, however, did not ap- Looking across all buyer groups, the correlation be-
pear to be linear. tween customer satisfaction and SOW increases from r =
3. Therefore, we tested a number of curvilinear .19 in the linear model to r = .27 in the cubic model. Fur-
models to find a model that offered the best fit to thermore, the fraction of variability (adjusted R2) in SOW
the data. Of the curvilinear models tested, the cu- that is explained by customer satisfaction doubles from
bic regression was the model that best fit the data 3% in the linear model to 7% in the cubic model.5
(based on the Sp criterion of Breiman and Freed-
man [1983]). 1. Both the overall satisfaction and share-of-wallet variables were
treated as continuous.
2. To maintain variability within the customer satisfaction ratings,
Hypothesis 1 data have been analyzed at the individual level of analysis (as opposed to
aggregated by client organization).
Hypothesis 1 predicted a positive, asymmetric, and 3. In particular, logarithmic, inverse, quadratic, cubic, compound,
nonlinear relationship between customer satisfaction and power, S, growth, exponential, and logistic models were tested.
share-of-wallet. 4. Under general conditions, the model that minimizes Sp provides the
best predictions on new data.
An initial exploratory analysis was conducted using
5. Technical note: Although the reader may question such a low level
chi-square automatic interaction detection (CHAID) of explained variance, the authors contend that it is substantial for two
(Babinec 1990; Kass 1980; Magidson 1993; Perreault and reasons. First, countless other criteria exist that are not included in our ad-

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 41

FIGURE 1
Chi-Square Tests for Share-of-Wallet by Overall Satisfaction Level: All Buyer Groups

Share of Wallet
All Satisfaction Levels
Mean 11.7816
Std. Dev. 7.8374
n 348
% 100.00

Share of Wallet Share of Wallet

Sat <= 8 Sat > 8
Mean 10.8160 Mean 15.4192
Std. Dev. 6.5409 Std. Dev. 10.7814
n 275 n 73
% 79.02 % 20.98

Adj. P-value = .0000; F = 21.0496; d.f. = 1, 346

TABLE 1 The results of the analysis support Hypothesis 1. First,

Model Fit Statistics Using Sp Criterion of the linear model shows a statistically significant positive
Breiman and Freedman (1983): All Groups relationship between customer satisfaction and SOW. Sec-
ond, the results of the Sp analysis indicate that a cubic
SP (nonlinear) model provides a better fit than the linear
Cubic .1684 model. As illustrated in Figure 2, there is support for the
Quadratic .1693 proposed asymmetry and nonlinearity of the relationship:
Linear .1734 (a) At points of high satisfaction (i.e., CS = 9 or 10 on a 10-
Logarithmic .1756 point scale), there is a dramatic increase in predicted SOW,
Logistic .1773
Inverse .1779
which was also found to be the case in the exploratory
Compound .1789 CHAID analysis; (b) at low levels of satisfaction (i.e., CS
Growth .1789 = 3 or less), there is a dramatic decrease in predicted SOW.
Exponential .1789
Power .1814
S .1840
Hypothesis 2

Hypothesis 2 predicted that the relationship between

customer satisfaction and SOW would differ across buyer
mittedly simplistic decision model. Second, this data set contains indi- groups.
viduals from all three buyer groups studied, where we would naturally As with the “All Groups” analysis conducted regarding
expect each buyer group to have different levels of influence on the pur-
chase decision and different response functions with regard to share-of- Hypothesis 1, an exploratory analysis was conducted us-
wallet and customer satisfaction. (As such, we would expect the level of ing CHAID, with overall satisfaction levels used as a pre-
variance explained to increase for buyer groups with greater influence in dictor of SOW. In this case, CHAID analysis was
the purchase decision.) Therefore, we believe these findings to be
compelling. conducted for each of the three buyer groups.
The results indicated that each of the groups behaves
differently with regard to the impact of overall satisfaction

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

42 JOURNAL OF SERVICE RESEARCH / August 2003

TABLE 2
Regression Analyses: Customer Satisfaction as a Predictor of Share-of-Wallet
(SOW): Comparison of a Linear and a Cubic Model for All Groups
B SE B β T Significance T R Unadjusted R2 Adjusted R2

Linear model
(SOWij = αj + βCSCSij + εij) .19 .03 .03
Customer satisfaction 0.90 0.26 0.19 3.528 .0005
Constant 5.12 1.93 2.649 .0084
Cubic model
2 3
(SOWij = αj + βCSCSij + βCSCS ij + βCSCS ij + εij) .27 .07 .07
Customer satisfaction 5.76 4.89 1.19 1.180 .2389
2
Customer satisfaction –1.24 0.80 –3.46 –1.553 .1214
3
Customer satisfaction 0.08 0.04 2.53 1.988 .0476
Constant 2.59 9.41 0.276 .7830

FIGURE 2
Cubic and Linear Predictions of Share-of-Wallet by Overall Satisfaction Level: All Buyer Groups
Predicted Share-of-Wallet

25

20

15

10

0
1 2 3 4 5 6 7 8 9 10

Satisfaction Level
Count 2 1 10 8 25 33 78 118 49 24

Cubic Estimation Linear Estimation Mean SOW

on SOW. For Buyer Group 1 (Information Management), Buyer Group 2 and Buyer Group 3 and was unable to find
CHAID was unable to uncover any statistically meaning- any optimal partitions for Buyer Group 1, the results
ful partition based on level of satisfaction that would opti- would indicate that the relationship between satisfaction
mally predict SOW. For Buyer Group 2 (Product and SOW differs by buyer group and lends support that in
Management), however, the results indicated that the opti- the case of Buyer Group 2 and Buyer Group 3, the relation-
mal predictive relationship occurred when the satisfaction ship is nonlinear and asymmetric.
variable was partitioned into two groups: (a) satisfaction Pearson correlations were examined for each of the
levels ≤ 7, and (b) satisfaction levels ≥ 8, having an ad- buyer groups to provide an initial examination of a possi-
justed p-value = .01 and an F = 11.1 (see Figure 3). ble linear relationship (see Table 3). The results revealed
For Buyer Group 3 (Purchasing), the results indicated that the linear correlation between satisfaction and
that the optimal predictive relationship occurred when the SOW was not statistically significant for Buyer Group 1
satisfaction variable was partitioned into two groups: (1) (Information Management), whereas the correlations
satisfaction levels ≤ 8, and (2) satisfaction levels ≥ 9, hav- for Buyer Group 2 (Product Management) and Buyer Group
ing an adjusted p-value = .04 and an F = 7.2 (see Figure 4). 3 (Purchasing) were statistically different from zero, having
As CHAID uncovered different optimal partitions for correlations of .24 and .23, respectively. The lack of a sig-

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 43

FIGURE 3
Chi-Square Tests for Share-of-Wallet by Overall Satisfaction Level:
Buyer Group 2 (Product Management)

Share-of-Wallet
All Satisfaction Levels
Mean 13.5034
Std. Dev. 9.4293
n 149
% 100.00

Share-of-Wallet Share-of-Wallet
Sat <= 7 Sat > 7
Mean 10.6754 Mean 15.6917
Std. Dev. 7.6848 Std. Dev. 10.0943
n 65 n 84
% 43.62 % 56.38

Adj. P-value = .0055; F = 11.0768; d.f. = 1, 147

nificant correlation for Buyer Group 1, versus significant groups. The results confirmed the results of the hierarchi-
correlations for Buyer Group 2 and Buyer Group 3 also cal regression. Cubic models provided the best fit for
supports Hypothesis 2 that the relationship between satis- Buyer Group 1 and Buyer Group 3, but a quadratic model
faction and SOW differs by buyer group. provided the best fit for Buyer Group 2 (see Table 5).
To estimate the nonlinear relationship between cus- For Buyer Group 1 (Information Management), the
tomer satisfaction and SOW on individual buyer groups, correlation between customer satisfaction and SOW in-
the curvilinear (cubic) regression models—supported by creases from r = .06 in the linear model to r = .34 in the cu-
Hypothesis 1—were tested on each buyer group. To ensure bic model, and the fraction of variability (adjusted R2) in
that the cubic term contributed significantly to the models, SOW that is explained by customer satisfaction increases
a hierarchical regression analysis was conducted on the from 0% in the linear model to 9% in the cubic model (see
data for each buyer group, sequentially entering the linear, Table 6). Similarly, for Buyer Group 3 (Purchasing), the
quadratic, and cubic terms (Bosker and Snijders 1999; correlation between customer satisfaction and SOW in-
Raudenbush and Byrk 2002). Entry of the cubic term was creases from r = .23 in the linear model to r = .33 in the cu-
found to provide statistically significant information (over bic model, and adjusted R2 doubles from 4% in the linear
and above the linear and quadratic terms) for Buyer Group model to 8% in the cubic model (see Table 7).
1 (significance F change = .00) and Buyer Group 3 (signif- For Buyer Group 2 (Product Management), however,
icance F change = .10), but not for Buyer Group 2, where the benefits of the quadratic model over the linear model
only entry of the quadratic term was significant over and are less pronounced. The correlation between customer
above the linear term (significance F change = .08) (see satisfaction and SOW increases from r = .24 in the linear
Table 4). model to r = .28 in the cubic model, and adjusted R2 in-
In addition, Sp analysis was conducted on the linear, creases from 5% in the linear model to 6% in the cubic
quadratic, and cubic functions for each of the buyer model (see Table 7). Nonetheless, the results of the Sp anal-

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

44 JOURNAL OF SERVICE RESEARCH / August 2003

FIGURE 4
Chi-Square Tests for Share-of-Wallet by Overall Satisfaction Level: Buyer Group 3 (Purchasing)

Share-of-Wallet
All Satisfaction Levels
Mean 10.3518
Std. Dev. 5.8179
n 83
% 100.00

Share-of-Wallet Share-of-Wallet
Sat <= 8 Sat > 8
Mean 9.4206 Mean 13.2850
Std. Dev. 3.7718 Std. Dev. 9.3661
n 63 n 20
% 75.90 % 24.10

Adj. P-value = .0441; F = 7.2043; d.f. = 1, 81

TABLE 3
Pearson Correlations: Customer Satisfaction and Share-of-Wallet for Each Buyer Group
Share-of-Wallet

Buyer Group 1 (Information Management) Overall satisfaction Pearson correlation .056

Significance (two-tailed) .561
N 110
Buyer Group 2 (Product Management) Overall satisfaction Pearson correlation .238**
Significance (two-tailed) .003
N 149
Buyer Group 3 (Purchasing) Overall satisfaction Pearson correlation .230*
Significance (two-tailed) .036
N 83

* Correlation is significant at the .05 level (two-tailed). ** Correlation is significant at the .01 level (two-tailed).

ysis indicate that the quadratic model provides a better fit DISCUSSION
to the data than the linear model.
Figures 5, 6, and 7 illustrate the predicted CS-SOW re- The analysis reported here advances the empirical re-
lationship for each buyer group. Again, the results would search regarding the intuitive relationship between cus-
indicate that the relationship between satisfaction and tomer satisfaction and business outcomes in two key ways.
SOW differs by buyer group. In addition, in all cases, the First, our findings indicate that satisfaction is positively re-
best fitting models were found to be nonlinear. lated to the share of business a customer conducts with a

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 45

TABLE 4
Hierarchical Regression Analyses: Customer Satisfaction as a Predictor of
Share-of-Wallet: Comparison of Linear, Quadratic, and Cubic Models: Buyer Groups 1, 2, and 3
Significance
R Unadjusted R2 Adjusted R
2 2
R Change F Change F Change

Group 1 (Information Management)

Linear model .06 .00 –.01 .00 0.34 .56
Quadratic model .20 .04 .02 .04 4.32 .04
Cubic model .34 .12 .09 .07 8.96 .00
Group 2 (Product Management)
Linear model .24 .06 .05 .06 8.86 .00
Quadratic model .28 .08 .06 .02 3.07 .08
Cubic model .28 .08 .06 .00 0.12 .73
Group 3 (Purchasing)
Linear model .23 .05 .04 .05 4.54 .04
Quadratic model .28 .08 .06 .03 2.20 .14
Cubic model .33 .11 .08 .03 2.71 .10

TABLE 5
Model Fit Statistics Using Sp Criterion of Breiman and Freedman (1983):
Comparison of Linear, Quadratic, and Cubic Models: Buyer Groups 1, 2, and 3
SP

Buyer Group 1 (Information Management)

2 3
Cubic SOWij = αj + βCS CSij + βCS CS ij + βCS CS ij + εij .3590
2
Quadratic SOWij = αj + βCS CSij + βCS CS ij + εij .3820
Linear SOWij = αj + βCS CSij + εij .3901
Buyer Group 2 (Product Management)
2 3
Cubic SOWij = αj + βCS CSij + βCS CS ij + βCS CS ij + εij .5857
2
Quadratic SOWij = αj + βCS CSij + βCS CS ij + εij .5781
Linear SOWij = αj + βCS CSij + εij .5823
Buyer Group 3 (Purchasing)
2 3
Cubic SOWij = αj + βCS CSij + βCS CS ij + βCS CS ij + εij .0915
2
Quadratic SOWij = αj + βCS CSij + βCS CS ij + εij .0932
Linear SOWij = αj + βCS CSij + εij .0943

NOTE: SOW = share-of-wallet.

TABLE 6
Regression Analyses: Customer Satisfaction as a Predictor of Share-of-Wallet (SOW)
Within Buyer Group 1 (Information Management): Comparison of a Linear and a Cubic Model
B SE B β T Significance T R Unadjusted R2 Adjusted R2

Linear model
(SOWij = αj + βCSCSij + εij) .06 .00 –.01
Customer satisfaction 0.22 0.39 0.06 0.583 .5613
Constant 8.88 2.82 3.143 .0022
Cubic model
2 3
(SOWij = αj + βCSCSij + βCSCS ij + βCSCS ij + εij ) .34 .12 .09
Customer satisfaction 11.18 5.33 2.78 2.096 .0384
2
Customer satisfaction –2.45 0.93 -7.86 –2.620 .0101
3
Customer satisfaction 0.15 0.05 5.28 2.993 .0034
Constant –2.31 9.70 –0.239 .8119

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

46 JOURNAL OF SERVICE RESEARCH / August 2003

TABLE 7
Regression Analyses: Customer Satisfaction as a Predictor of Share-of-Wallet (SOW)
Within Buyer Group 3 (Purchasing): Comparison of a Linear and a Cubic Model
B SE B β T Significance T R Unadjusted R2 Adjusted R2

Linear model
(SOWij = αj + βCSCSij + εij) .23 .05 .04
Customer satisfaction 0.83 0.39 0.23 2.131 .0361
Constant 4.08 3.01 1.358 .1782
Cubic model
2 3
(SOWij = αj + βCSCSij + βCSCS ij + βCSCS ij + εij) .33 .11 .08
Customer stisfaction 12.18 8.85 3.37 1.376 .1728
2
Customer stisfaction –2.19 1.47 –8.20 –1.491 .1398
3
Customer stisfaction 0.12 0.08 5.21 1.647 .1036
Constant –11.26 16.05 –0.701 .4851

TABLE 8
Regression Analyses: Customer Satisfaction as a Predictor of Share-of-Wallet (SOW)
Within Buyer Group 2 (Product Management): Comparison of a Linear and a Quadratic Model
2 2
B SE B β T Significance T R Unadjusted R Adjusted R

Linear model
(SOWij = αj + βCSCSij + εij) .24 .06 .05
Customer satisfaction 1.37 0.46 0.24 2.977 .0034
Constant 3.31 3.51 0.945 .3461
Quadratic model
2 3
(SOWij = αj + βCSCSij + βCSCS ij + βCSCS ij + εij) .28 .08 .06
Customer satisfaction –3.78 2.98 –0.66 –1.271 .2058
2
Customer satisfaction 0.38 0.21 0.91 1.753 .0816
Constant 19.82 10.04 1.975 .0502

FIGURE 5
Cubic and Linear Predictions of Share-of-Wallet by Overall Satisfaction Level:
Buyer Group 1 (Information Management)
Predicted Share-of-Wallet

25

20

15

10

0
1 2 3 4 5 6 7 8 9 10

Satisfaction Level
Count 2 0 3 2 12 10 27 39 11 4

Cubic Estimation Linear Estimation Mean SOW

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 47

FIIGURE 6
Quadratic and Linear Predictions of Share-of-Wallet by Overall Satisfaction Level:
Buyer Group 2 (Product Management)

Predicted Share-of-Wallet 25

20 Outside
Current Data
Range
15

10

0
1 2 3 4 5 6 7 8 9 10

Satisfaction Level
Count 0 0 4 5 13 12 31 49 23 12

Quadratic Estimation Linear Estimation Mean SOW

FIGURE 7
Cubic and Linear Estimations of Overall Satisfaction and Share-of-Wallet:
Buyer Group 3 (Purchasing)
Predicted Share-of-Wallet

25

20

15 Outside
Current Data
Range
10

0
1 2 3 4 5 6 7 8 9 10

Satisfaction Level
Count 0 1 3 1 0 11 19 28 13 7

Cubic Estimation Linear Estimation Mean SOW

particular service provider (SOW), as opposed to simply Second, the findings suggest that the relationship be-
repurchasing a product or service at some point in the fu- tween satisfaction and SOW is nonlinear. This corre-
ture or continuing to keep a business relationship with a sponds to similar findings regarding the relationship
service provider. between satisfaction and repurchase intention, retention,
and word of mouth.

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

48 JOURNAL OF SERVICE RESEARCH / August 2003

The findings also contribute to the research regarding tential to provide managers with insight into segmentation
organizational buying. The findings suggest that the rela- strategies that could be developed to maximally affect
tionship between buyer group satisfaction and SOW is strategies designed to improve SOW.
nonlinear and that the functional form of the relationship
varies by segment.
CONCLUSION
Implications for Practice
Researchers have found that customer satisfaction is
The key implication is that managers should not simply positively associated with outcomes desired by managers,
strive to improve reported satisfaction levels without an including repurchase, market share, and word of mouth. It
understanding of the relationship to customer behavior has been proposed that satisfaction also results in an in-
measures: in this case, SOW. Given the nonlinearity of the creased share of business a customer conducts with a par-
relationship, it is critical that managers be certain that ef- ticular service provider (SOW). The results of this study
forts designed to improve satisfaction do so in sufficient lend support to such a connection, showing an association
force so as to reach satisfaction levels that correspond with between satisfaction and SOW. The relationship, however,
increasing SOW levels. was found to be nonlinear, with the greatest positive im-
Furthermore, simply treating all buyers as homoge- pact occurring at the upper extreme of satisfaction levels.
neous has the potential to misrepresent the relationship be- As a result, linear analyses of satisfaction and SOW have
tween satisfaction and SOW. Therefore, it is important that the potential to greatly misrepresent the actual relationship.
the level of influence of various buyer groups on purchase
behavior be uncovered so that resources can be appropri-
ately allocated to those areas providing the greatest impact. REFERENCES
In the case of the firm that provided these data, manage-
rial experience and prior research supported a dispropor- Anderson, Erin, Wujin Chu, and Barton Weitz (1987), “Industrial Pur-
tionate weighting to Buyer Group 3 (Purchasing) because chasing: An Empirical Exploration of the Buy-Class Framework,”
this group actually selects the firm from which to pur- Journal of Marketing, 51 (July), 71-86.
Anderson, Eugene W. (1998), “Customer Satisfaction and Word-of-
chase. The results of this research, however, indicated that Mouth,” Journal of Service Research, 1 (1), 1-14.
the satisfaction level of Buyer Group 1 (Information Man-  and Claes Fornell (1999), “The Customer Satisfaction Index as a
agement) fits the data equally well as for Buyer Group 3. Leading Indicator,” in Handbook of Services Marketing and Manage-
ment, Teresa A. Swartz and Dawn Iacobucci, eds. Thousand Oaks,
As a result of this and other recent information supporting CA: Sage, 255-67.
these findings, the firm has made a concerted effort to dra- , , and Donald R. Lehmann (1994), “Customer Satisfac-
matically alter the content of the information it provides tion, Market Share, and Profitability: Findings from Sweden,” Jour-
regarding its product offerings. nal of Marketing, 58 (July), 53-66.
 and Vikas Mittal (2000), “Strengthening the Satisfaction-Profit
Chain,” Journal of Service Research, 3 (2), 107-20.
Limitations and Future Research  and Mary W. Sullivan (1993), “The Antecedents and Conse-
quences of Customer Satisfaction for Firms,” Marketing Science, 12
(Spring), 125-43.
This study used data from a business-to-business rela-
Babinec, T. (1990), “CHAID Response Modeling and Segmentation,”
tionship within a single industry. As such, the results can- Quirk’s Marketing Research Review, June/July, 12-15.
not be immediately applied to different industries or in Bemmaor, Albert C. (1995), “Predicting Behavior from Intention-to-Buy
business-to-consumer relationships. Therefore, additional Measures: The Parametric Case,” Journal of Marketing Research, 32
(May), 176-91.
research needs to be conducted regarding the relationship
Bernhardt, Ken, Naveen Donthu, and Pam Kennett (2000), “Longitudinal
between satisfaction and SOW in a variety of industries. Analysis of Satisfaction and Profitability,” Journal of Business Re-
Similarly, this study focuses exclusively on the rela- search, 47 (2), 161-71.
tionship between overall satisfaction levels and SOW. Bloemer, José, Hans Kasper, and Jos Lemmink (1990), “The Relation-
ship between Overall Dealer Satisfaction, Satisfaction with the At-
More comprehensive models that take into account the re- tributes of Dealer Service, Intended Dealer Loyalty and Intended
lationship between attribute-level satisfaction and SOW Brand Loyalty: A Dutch Automobile Case,” Journal of Satisfaction,
could provide managers with important insights into more Dissatisfaction and Complaining Behavior, 3, 42-47.
tactical measures firms could take to improve the percent- Bolton, Ruth N. (1998), “A Dynamic Model of the Duration of the Cus-
tomer’s Relationship with a Continuous Service Provider: The Role
age of business their customers give to them versus their of Satisfaction,” Marketing Science, 17 (1), 45-65.
competition.  and James H. Drew (1991), “A Longitudinal Analysis of the Im-
In addition, research needs to be conducted into the pact of Service Changes on Customer Attitudes,” Journal of Mar-
keting, 55 (1), 1-10.
moderating effects of customer characteristics on the rela-
tionship between satisfaction and SOW. This has the po-

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

 Keiningham et al. / CUSTOMER SATISFACTION 49

Bosker, Roel J. and Tom A. B. Snijders (1999), Multilevel Analysis: An Kotler, Philip (1994), Marketing Management: Analysis, Planning, Im-
Introduction to Basic and Advanced Multilevel Modeling. Thousand plementation and Control, 8th ed. Englewood Cliffs, NJ: Prentice
Oaks, CA: Sage. Hall.
Breiman, L. and D. Freedman (1983), “How Many Variables Should Be Leigh, Thomas W. and Arno J. Rethans (1984), “A Script-Theoretic
Entered in a Regression Equation?” Journal of the American Statisti- Analysis of Industrial Purchasing Behavior,” Journal of Marketing,
cal Association, 78 (381), 131-36. 48 (Fall), 22-32.
Brookes, Donald and Nicholas Stodin (1995), “The New Service Rela- Leung, Kwok, Wai-Kwan Li, and Yuk-Fai Au (1998), “The Impact of
tionship,” Canadian Business Review, 22 (March 22), 24(4). Customer Service and Product Value on Customer Loyalty and Pur-
Cannon, Joseph P. and Christian Homburg (2001), “Buyer-Seller Rela- chase Behavior,” Journal of Applied Social Psychology, 28 (18),
tionships in Business Markets,” Journal of Marketing, 65 (January), 1731-41.
29-43. Magidson, J. (1993), SPSS for Windows: CHAID, Release 6.0. Chicago:
 and William D. Perreault Jr. (1999), “Buyer-Seller Relationships SPSS.
in Business Markets,” Journal of Marketing Research, 36 (Novem- Manski, Charles F. (1990), “The Use of Intentions Data to Predict Behav-
ber), 439-61. ior: A Best-Case Analysis,” Journal of the American Statistical Asso-
Cohen, Jacob, and Patricia Cohen (1984). Applied Multiple Regression/ ciation, 85 (412), 934-40.
Correlation Analysis for the Behavioral Sciences, 2nd ed. Hillsdale, McQuiston, Daniel H. (1989), “Novelty, Complexity, and Importance as
NJ: Lawrence Erlbaum. Causal Determinants,” Journal of Marketing, 53 (April), 66-79.
Coyne, Kevin (1989), “Beyond Service Fads—Meaningful Strategies for Mittal, Vikas and P. M. Baldasare (1996), “Impact Analysis and the
the Real World,” Sloan Management Review, 30 (Summer), 69-76. Asymmetric Influence of Attribute Performance on Patient Satisfac-
DeSarbo, Wayne S., Lenard Huff, Marcelo M. Rolandelli, and Jungwhan tion,” Journal of Health Care Marketing, 16 (3), 24-31.
Choi (1994), “On the Measurement of Perceived Service Quality: A  and Wagner Kamakura (2001), “Satisfaction, Repurchase Intent
Conjoint Analysis Approach,” in Service Quality: New Directions in and Repurchase Behavior: Investigating the Moderating Effect of
Theory and Practice, Roland T. Rust and Richard L. Oliver, eds. Customer Characteristics,” Journal of Marketing Research, 38 (Feb-
Thousand Oaks, CA: Sage, 199-220. ruary), 131-42.
De Wulf, Kristof, Gaby Odekerken-Schröder, and Dawn Iacobucci , Jerome M. Katrichis, Frank Forkin, and Mark Konkel (1998),
(2001), “Investments in Consumer Relationships: A Cross-Country “The Asymmetric Impact of Negative and Positive Attribute-Level
and Cross-Industry Exploration,” Journal of Marketing, 65 (Octo- Performance on Overall Satisfaction and Repurchase Intentions,”
ber), 33-50. Journal of Marketing, 62 (January), 33-47.
Feldman, Stanley and William Dale Berry (1985), Multiple Regression in Morrison, Donald G. (1979), “Purchase Intentions and Purchase Behav-
Practice. Sage University Paper Series on Quantitative Applications ior,” Journal of Marketing, 43 (Spring), 65-74.
in the Social Sciences, series no. 07-050. Newbury Park, CA: Sage. Morwitz, Vicki G., Eric Johnson, and David Schmittlein (1993), “Does
Fornell, Claes (1992), “A National Customer Satisfaction Barometer: Measuring Intent Change Behavior?” Journal of Consumer Re-
The Swedish Experience,” Journal of Marketing, 56 (January), 6-21. search, 20 (June), 46-61.
, Michael D. Johnson, Eugene W. Anderson, Jaesung Cha, and , Joel H. Steckel, and Alok Gupta (1997), “When Do Repurchase
Barbara Everett Bryant (1996), “The American Customer Satisfac- Intentions Predict Sales?” working paper, Report No. 97-112, Mar-
tion Index: Nature, Purpose, and Findings,” Journal of Marketing, 60 keting Science Institute, Cambridge, MA.
(October), 7-18. Oliva, Terence A., Richard L. Oliver, and Jan C. MacMillan (1992), “A
Fred, Charles L. (2002), “Delivering Value to Customers Fast—The Ulti- Catastrophe Model for Developing Service Satisfaction Strategies,”
mate Competitive Weapon,” ABA Banking Journal White Papers. Re- Journal of Marketing, 58 (July), 83-95.
trieved from http://www.banking.com/aba/whitepapers.asp Oliver, Richard L. (1997), Satisfaction: A Behavioral Perspective on the
Hitt, Michael A., Leonard Bierman, Katsuhiko Shimizu, and Rahul Consumer. New York: McGraw-Hill.
Kochhar (2001), “Direct and Moderating Effects of Human Capital , Roland T. Rust, and Sajeev Varki (1997), “Customer Delight:
on Strategy and Performance in Professional Service Firms: A Re- Foundations, Findings and Managerial Insight,” Journal of Retailing,
source Based Perspective,” Academy of Management Journal, 44, 13- 7 (3), 311-36.
28. Osborne, Jason and Elaine Waters (2001), “Four Assumptions of Multi-
Jackson, Donald W., Jr., Janet E. Keith, and Richard K. Burdick (1984), ple Regression That Researchers Should Always Test,” Practical As-
“Purchasing Agents’ Perceptions of Industrial Buying Center Influ- sessment, Research & Evaluation, 8 (2). Retrieved from http://
ence: A Situational Approach,” Journal of Marketing, 48 (Fall), 75- ericae.net/pare/getvn.asp?v=8&n=2
83. Patchen, Martin (1974), “The Locus and Basis of Influence in Organiza-
Jamieson, Linda F. and Frank M. Bass (1989), “Adjusting Stated Inten- tional Decisions,” Organizational Behavior and Human Perfor-
tion Measures to Predict Trial Purchase of New Products: A Compar- mance, 11 (April), 195-221.
ison of Models and Methods,” Journal of Marketing Research, 26 Pedhazur, Elazar J. (1997), Multiple Regression in Behavioral Research,
(August), 336-45. 3rd ed. Orlando, FL: Harcourt Brace.
Jones, Thomas O. and W. Earl Sasser Jr. (1995), “Why Satisfied Cus- Perreault, W. D. and H. C. Barksdale (1980), “A Model-Free Approach
tomers Defect,” Harvard Business Review, 73 (November/Decem- for Analysis of Complex Contingency Data in Survey Research,”
ber), 88-99. Journal of Marketing Research, 17 (November), 503-15.
Kass, G. (1980), “An Exploratory Technique for Investigating Large Raudenbush, Stephen W. and Anthony S. Byrk (2002), Hierarchical Lin-
Quantities of Categorical Data,” Applied Statistics, 29, 127-29. ear Models: Applications and Data Analysis Methods. Thousand
Keiningham, Timothy L., Melinda K. M. Goddard, Terry G. Vavra, and Oaks, CA: Sage.
Andrew J. Iaci (1999), “Customer Delight and the Bottom Line,” Reichheld, Frederick (1996), “Learning from Customer Defections,”
Marketing Management, 8 (Fall), 57-63. Harvard Business Review, 74 (March/April), 56-69.
 and Terry G. Vavra (2001), The Customer Delight Principle. New Ronchetto, John R., Jr., Michael D. Jutt, and Peter H. Reingen (1989),
York: McGraw-Hill. “Embedded Influence Patterns in Organizational Buying System,”
Kohli, Ajay (1989), “Determinants of Influence in Organizational Journal of Marketing, 53 (October), 51-62.
Buying: A Contingency Approach,” Journal of Marketing, 53 (July), Rust, Roland T. and Richard L. Oliver (1994), “Service Quality: Insights
50-65. and Managerial Implications from the Frontier,” in Service Quality:
New Directions in Theory and Practice, Roland T. Rust and Richard
L. Oliver, eds. Thousand Oaks, CA: Sage.

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016

50 JOURNAL OF SERVICE RESEARCH / August 2003

 and Anthony J. Zahorik (1993), “Customer Satisfaction, Cus- Zahorik, Anthony J. and Roland T. Rust (1992), “Modeling the Impact of
tomer Retention, and Market Share,” Journal of Retailing, 69 (2), Service Quality on Profitability: A Review,” Advances in Services
193-215. Management and Marketing, 1, 247-76.
Sheth, Jagdish N. (1973), “A Model of Industrial Buyer Behavior,” Jour- Zeithaml, Valarie A. (2000), Service Quality, Profitability, and the Eco-
nal of Marketing, 37 (October), 50-56. nomic Worth of Customers: What We Know and What We Need to
, Rajendra S. Sisdodia, and Arun Sharma (2000), “The Anteced- Learn,” Journal of the Academy of Marketing Science, 28 (1), 67-85.
ents and Consequences of Customer-Centric Marketing,” Journal of
the Academy of Marketing Science, 28 (1), 55-66.
Spekman, Robert E. (1979), “Influence and Information: An Exploratory Timothy L. Keiningham is senior vice president of Marketing
Investigation of the Boundary Role Person’s Basis of Power,” Acad- Metrics, Inc. based in Paramus, New Jersey. His research inter-
emy of Management Journal, 22 (1), 104-17. ests include service marketing and linking service improvement
 and Louis W. Stern (1979), “Environmental Uncertainty and efforts to business outcomes. He is coauthor of The Customer
Buying Group Structure: An Empirical Investigation,” Journal of Delight Principle, published by McGraw-Hill/American Mar-
Marketing, 43 (Spring), 54-64.
keting Association.
Thomas, Robert J. (1982), “Correlates of Interpersonal Purchase Influ-
ence in Organizations,” Journal of Consumer Research, 9 (Septem-
ber), 171-82.
Tiffany Perkins-Munn is vice president of Marketing Metrics,
Vavra, Terry G. (1997), Improving Your Measurement of Customer Satis- Inc. Her research interest is on examining the underpinnings of
faction. Milwaukee, WI: ASQC Quality Press. social psychology and behavioral science in marketing informa-
Vyas, Niren and Arch G. Woodside (1984), “An Inductive Model of In- tion systems (i.e., internal reporting, marketing intelligence, and
dustrial Supplier Choice Processes,” Journal of Marketing, 48 (Win- decision support systems) and marketing research using a mix of
ter), 30-45. both qualitative and quantitative methods.
Webster, Frederick E., Jr. (1978), “Management Science in Industrial
Marketing,” Journal of Marketing, 42 (January), 21-27. Heather Evans is a managing director at a large financial institu-
Weigand, Robert E. (1966), “Identifying Industrial Buying Responsibil- tion. She runs the institutional securities group’s Client Feedback
ity,” Journal of Marketing Research, 3 (February), 81-84.
Program, the first large-scale, proprietary client satisfaction mea-
Wiegran, Gaby and Hardy Koth (1999), “Customer Retention in On-Line
Retail,” Journal of Internet Banking and Commerce, 4 (October). surement program at an institutional securities firm.
Retrieved from http://www.arraydev.com/commerce/jibc/9909-07.htm

Downloaded from jsr.sagepub.com at PENNSYLVANIA STATE UNIV on March 5, 2016