V. Kumar, J. Andrew Petersen, & Robert P.

Leone

Defining, Measuring, and Managing

Business Reference Value
It is common for business-to-business firms to use references from client firms when trying to influence prospects
to become new customers. In this study, the authors define the concept of business reference value (BRV) as the
ability of a client’s reference to influence prospects to purchase and the degree to which it does so. They develop
a three-step method to compute BRV for a given client using a retrospective reported measure of reference value.
Next, they use data from a financial services and a telecommunications firm to identify and empirically test the
drivers of BRV. These drivers fall into four categories: (1) length of client relationship, (2) client firm size, (3)
reference media format, and (4) reference congruency. Next, the authors empirically show that clients that have the
highest BRV are not the same as the clients that have the highest customer lifetime value. They also show that an
average client that is high on BRV has significantly different characteristics from the average client that is low on
BRV. Finally, they derive implications for managing BRV.

Keywords: client references, business reference value, customer lifetime value, business-to-business marketing

any firms are trying to capitalize on the power of Microsoft to represent what its sales executives believe are

M client reference behavior as part of their general

marketing and sales efforts to encourage new cus-
tomer adoption. For business-to-business (B2B) firms espe-
the best examples of successful implementations of
Microsoft products and services. These case studies serve
as references from current clients that Microsoft uses to
cially, the use of referencing behavior is often the only influence prospects to adopt.
alternative for leveraging the value of current clients on To date, there has been limited research explaining the
new customer adoption through social influences. For B2B social influences present in B2B settings (Libai et al. 2010).
firms, the use of referencing behavior is important because, Therefore, it is not yet clear how seller firms can quantify
unlike for business-to-consumer (B2C) firms, the purchase the value that these references provide, whether seller firms
decision process often does not rely on other social influ- are able to determine which clients are likely to be the most
ences such as word of mouth (WOM) or referrals from valuable references for new customer acquisition, or which
other businesses. For example, Microsoft has created a Cus- reference formats are the most effective at influencing
tomer Reference Program to influence prospects to adopt prospects to adopt. In this study, we focus on further under-
their products and services.1 The firm does so by directing standing the role and value of client references, specifically
prospects to a website that contains case studies and white in a B2B selling context. We aim to answer the following
papers from a sample of current clients, selected by four key research questions:
1. Can we measure the value of a business reference from a
1For an example of Microsoft’s Customer Reference Program, client?
see http://www.microsoft.com/hk/sia/default.mspx. 2. What are the key drivers, both in terms of the referencing
client and the format of reference, of the value of the
reference?
V. Kumar is Professor of Marketing, the Richard and Susan Lenny Distin- 3. Are clients that generate valuable references the same ones
guished Chair in Marketing, the Executive Director of the Center for Excel- that are the most profitable in terms of their own purchasing
lence in Brand & Customer Management, and the Director of the PhD behavior to the seller firm?
Program in Marketing, J. Mack Robinson School of Business, Georgia 4. What types of clients tend to generate high-value references?
State University (e-mail: [email protected]). J. Andrew Petersen is Assistant
Professor of Marketing and the Assistant Director for the Center of Inte- Our first step before we begin to answer these four
grated Marketing and Sales, Kenan-Flagler Business School, University of questions is to determine how business references play a
North Carolina at Chapel Hill (e-mail: [email protected]). Robert role in the adoption of prospects in a B2B setting. We do so
P. Leone is Professor and J. Vaughn and Evelyne H. Wilson Chair of Mar-
keting, Neeley School of Business, Texas Christian University (e-mail:
by leveraging the rich literature in marketing on the buying
[email protected]). The authors thank the anonymous financial and the processes of B2B firms and through qualitative interviews
telecommunication companies for providing access to the data used in with managers from B2B firms who are directly involved in
this study. They thank participants at 2012 New Developments in the prac- making purchase decisions with their respective firms. We
tice of Marketing Science Conference in Washington, DC, for providing use the literature review and the qualitative interviews to
valuable comments on a previous version of this article. They thank Renu develop a conceptual model of the drivers of reference
for copyediting a previous version of the manuscript. Rajdeep Grewal
selection (which client to select as a reference) and refer-
served as area editor for this article.
ence value (how much business the reference will bring in

© 2013, American Marketing Association Journal of Marketing

ISSN: 0022-2429 (print), 1547-7185 (electronic) 68 Volume 77 (January 2013), 68–86
monetary value). In addition, we hypothesize how each of CLVn) in driving BRV. We find that, in general, most firms
these drivers is likely to affect the process of selection and that were acquired during the sales process were highly
value of the reference. influenced by references in making their decisions (high
Then, to empirically answer the four key questions, we Refn). We also find that many of these firms only relied on
use data from a B2B financial services firm and a telecom- a few references in making their decisions (low number of
munications firm that focus on selling products and services firms with positive DOIin). Finally, we find a significant
to mainly small to medium-sized businesses, for which the variance in the profitability of the newly acquired cus-
purchase situations tend to be simple. Examples of these tomers (high variance in CLVn), suggesting that references
products and services include business checking/savings influence low- and high-value customers to adopt.
(financial services) and commercial/business landline tele-
phone service (telecommunications). (For a more detailed
list of product categories, see the Web Appendix at www. Theoretical Foundations
marketingpower.com/jm_webappendix.) We focus specifi-
cally on these two firms’ customer acquisition efforts in Literature Review
which none of the potential client firms had purchased from Business-to-business firms’ decision to purchase, when
the focal selling firms or, likely, from any of the firms’ com- compared with B2C firms, is often more complex and drawn
petitors. This means that the potential client firms may have out. This is because purchases made by organizations tend
some general product category experience, which includes to rely on many different employees within the organization
any personal experience using financial services and (i.e., a buying center) and on budget, cost, and profit consid-
telecommunications products and services, but none of the erations (Webster and Wind 1972). As a result, seller firms’
potential client firms would have any specific product or sales endeavors are often multidimensional and can include
focal firm experience in this context to rely on in the pur-
both personal and impersonal sources along with commer-
chase situation. Given the churn faced by many B2B firms
cial and noncommercial sources (Moriarty and Spekman
that rely on generating new customers, this is a common sit-
1984). Seller firms can influence the organizational buying
uation and makes using references a key part of the potential
decisions of prospects through different sources, timing,
client firm’s decision to adopt. In addition, most of these
and quality of information. Moriarty and Spekman (1984)
small to medium-sized potential client firms likely have few
key decision makers, making it easier to obtain accurate offer empirical evidence showing that the seller firm uses
information regarding the influence of references on adop- personal noncommercial information sources such as social
tion. In addition, the selling firms use some other selling influences (e.g., references) throughout the decision
and marketing efforts, including product brochures, some process. These influences can become most important dur-
general branding efforts on the firms’ websites, and a sales ing the later stages of the buying process and often can be
force that calls on potential client firms, providing some what distinguishes one company from another becoming
potential variance in the effect of references on adoption. the “order winner.” Although not empirically tested in the
Thus, this study makes three key contributions to the literature, this suggests that seller firms that employ refer-
marketing literature. First, we define the concept of busi- ences during the sales process can significantly influence an
ness reference value (BRV), which is the ability of a client’s organizational buyer’s decision to make a purchase.
reference to provide monetary value to the seller firm by During the organizational buying decision process,
influencing a prospect to adopt and the degree to which it there are four key drivers that affect the prospect’s propen-
does so. We measure the BRV of each referencing client sity to adopt: (1) firm-initiated efforts (e.g., direct/mass
and the customer lifetime value (CLV) of each referencing communication), (2) competitor-initiated efforts (e.g., com-
client and newly acquired client from the prospect pool. To petitor direct/mass communication), (3) client-initiated
do this, we develop a new three-step method for computing efforts (e.g., references), and (4) prospect characteristics
BRV using a retrospective reported measure of reference (e.g., demographics) (Prins and Verhoef 2007; Reinartz,
value, which includes information about how much each Thomas, and Kumar 2005; Villanueva, Yoo, and Hanssens
reference influenced each prospect to purchase and the 2008). Because the objective of this study is to investigate
value of the new customer. how client-initiated efforts influence B2B prospects to
Second, we empirically determine the key drivers of make a purchase, we focus specifically on the research that
BRV using data from the two selling firms in this study. We explains the relationship between client-initiated efforts and
find that the characteristics of the referencing client, the customer adoption for organizational buyers, which is rela-
characteristics of the prospect, and the characteristics of the tively scarce (Libai et al. 2010).
reference are all significant predictors of reference selection Wangenheim and Bayón (2007) analyze the linkages
and/or BRV. These key drivers include length of client rela- between customer satisfaction, WOM, and new customer
tionship, client firm size, reference congruency, and refer- acquisition for both B2C and B2B customers in the German
ence media format. These results help us better understand energy-provider market. They find that managers of B2B
which references are likely to be most effective in convert- firms rely on WOM for their purchase decisions, and WOM
ing prospects to clients. affects new customer acquisition. However, it is unclear
Third, we deconstruct the measure of BRV to determine whether the source of the WOM was a manager or a con-
the role of each of the three measures (Refn, DOIin, and sumer, making it difficult for a seller firm to harness the

Defining, Measuring, and Managing Business Reference Value / 69

power of WOM for the purposes of enhancing customer other marketing and sales efforts. Finally, we asked at what
acquisition. point during the purchase process the references had an
Hada, Grewal, and Lilien (2011) show through a mixed- impact on the decision to purchase. We selected these
design experiment how supplier firms make trade-offs executives for the qualitative interviews because many rep-
when selecting references for prospects. They find that resented small to medium-sized businesses similar to the
increasing the similarities between the client and the potential client firms being targeted by the focal firms in
prospect provides significant benefit to the manager’s our empirical application.
evaluation of the referral. Although this study relates The results from our analysis of the qualitative inter-
directly to a B2B context, the outcome these authors are views suggest that references play a key role in the pur-
interested in is manager evaluation of the reference, chase decision process for many of these organizations.
whereas in our study, we are interested in determining the Indeed, our qualitative interviews show that many man-
value of the reference to the seller firm. agers from these businesses rely on references throughout
Stephen and Toubia (2010) analyze a group of online the purchase decision process, especially just before the
social commerce networks, in which the networks involve buying decision is made. As one manager from the buying
individual sellers linking their online marketplace with firm noted,
other sellers, creating a virtual mall. The authors find that The list of references provided by the seller helps us to
allowing sellers to connect with each other provides signifi- shortlist the seller firms. However, to finally decide on
cant economic value, especially to those whose accessibil- whom to buy from, we check at least a couple of refer-
ity is most enhanced by the network. Although this study ences provided by the seller shortly before our purchase
analyzes social connections between sellers, it is focused on decision. We also discuss internally as to how much influ-
ence the references had on our buying decisions.
firms that are trying to become more effective at selling
products and services to end consumers. The results of the interviews with the executives also
Godes (2012) uses an analytical model to uncover the suggested that those references that provided the most value
benefits to early and late adopters as a result of the seller in influencing the purchase decision were those from repu-
firm announcing a referencing program. He finds that one table firms; had the highest similarities to the buyer firm
of the key benefits of announcing a referencing program is (whether it was in terms of products or services purchased,
that early adopters are willing to pay more because of the industry of the referencing client, or job function of the ref-
increase in information transmission and information trans- erencing client); and were provided through richer media
parency—a benefit for late adopters as well. Thus, the refer- formats.
encing program can be viewed as a substitute for an exclu- We also interviewed 12 B2B managers from the sup-
sive use contract. While Godes shows that referencing plier firms about the use of references by their prospective
programs do add value, he does not suggest how seller clients. According to the chief executive officer of Pinnacle
firms should select and use client firm references to maxi- Promotions (who represents the typical sentiments of the
mize profitability. remaining respondents),
Thus, while there have been several studies that gener-
Our clients use the list of references we provide to reduce
ally analyze some aspects of social influence on organiza- the choice of suppliers. Once we get into the decision-
tional purchase decisions, we know of no study that has making stage, our participants ask us [for] specific refer-
specifically investigated how selling firms can understand ences. We provide access to multiple forms of references,
how to use client references to effectively drive customer different types of clients, and different job descriptions of
acquisition and how to value those client references, a key the person providing the reference. Our prospects are free
contribution of this study. to choose whichever reference they like to pursue.
We used the results of these qualitative interviews to help
Qualitative Interviews
guide the development and empirical testing of our concep-
To better understand the role of references in the organiza- tual model.
tional buying process, we conducted qualitative interviews
with 26 executives from small to medium-sized businesses
who play a key role in the purchase decision making for Conceptual Framework and
their respective firms. These executives participated in a Hypothesis Development
sales and marketing workshop organized by an executive Our goal is to understand the key drivers of reference selec-
education forum in the southeast United States over two tion by the seller firm and the drivers of BRV for each client
days. On Day 1, we interviewed 14 executives, and on Day reference. Specifically, we argue that firms strategically
2, we interviewed 12 more executives. Each interview select as references current client firms that are likely to
lasted for approximately 30 minutes and was conducted maximize the profitability from potential client firm pur-
individually. The same person conducted all the interviews. chase behavior. In addition, we argue that potential client
The objective of these qualitative interviews was to deter- firms make decisions to adopt to maximize the benefit they
mine whether the executives believe that references influ- receive from the relationship with the seller firm. Both situ-
ence their decision making. In addition, when they make ations are driven by similar theoretical constructs. Research
purchase decisions, we asked to what extent references suggests that a potential client firm’s decision to adopt (i.e.,
influence the decision to purchase when compared with the build a strategic alliance) is driven by examining the current

70 / Journal of Marketing, January 2013

alliances that the seller firm has built (Gulati and Gargiulo that differences in the referencing client characteristics will
1999). Moreover, the strength of these current alliances affect the reference selection strategy of the seller firm. We
depends on the “trusted informants” (or references) pro- also anticipate that the reference selection strategy, the ref-
vided by the seller firm and the degree to which those infor- erencing client characteristics (in terms of firm size and
mants are embedded in the seller firm. In addition, the abil- relationship length), the reference media format, and refer-
ity of the information transferred from the trusted informant ence congruency with the prospect will drive the BRV.
to the potential client firm to be understood and relatable
depends on the richness of the communication and the Client Firm Size
degree to which the information provided is similar to the To use client references to influence prospect adoption, the
business situation the potential client firm faces. Thus, first step a seller firm must take is to select which firms
seller firms should select trusted informants to pass along from its current client database are the best candidates to be
the most valuable information to potential client firms. references. We anticipate that seller firms select these
From this discussion, we believe that there are four key clients strategically to maximize the impact of the influence
drivers of reference selection and reference value: (1) the of the reference. The seller firm aims to reduce the informa-
degree to which the client firm can be viewed as a trusted tion cost of search and information asymmetry, with the
informant through client firm size, (2) the degree to which goal of making the decision to purchase easier for the
the client firm has built a strategic alliance with the seller prospect (Spence 1973). Reducing information asymmetry
firm through length of client relationship, (3) the ability of by providing information that is not otherwise available
the communication to convey information to the potential (i.e., prospects cannot experience the benefits of the product
client firm through reference media format, and (4) the or service until adoption) at a relatively low cost (e.g.,
degree to which the information provided is relatable to the through a reference) can be an effective signaling strategy
potential client firm through reference congruency. Figure 1 (Connelly et al. 2011). In addition, this information is likely
shows a graphic representation of these drivers. more valuable when the firm is trusted or has a good repu-
During the sales process, it is common for many B2B tation in the marketplace (Morgan and Hunt 1994). More-
firms to use references from clients as a tool to influence a over, large firms tend to receive more public scrutiny than
prospect to adopt a product/service. Thus, the process of small firms (Fombrun and Shanley 1990). However, even if
using references to influence the purchase process breaks this information does not always put the firm in a positive
down into two distinct steps. First, the seller firm selects the view, the mere availability of information about and famil-
clients in its current database to act as references to be used iarity with large firms tends to inflate people’s opinion of
during the sales efforts. Second, the sales force uses the ref- the larger firms’ activities (Tversky and Kahneman 1974).
erences to influence the prospects to adopt. We anticipate As a result, we expect that current client firms whose attri-
that these prospects will evaluate the value of the reference butes are visible in the marketplace and are perceived as
on the basis of several key characteristics. valuable to the marketplace (e.g., significant size) are more
In the next section, we develop the hypotheses related to likely to provide an effective reference to prospects through
the drivers of reference selection and BRV. We anticipate their reputation. Here, we define “client firm size” as the

FIGURE 1
Conceptual Model of the Drivers of BRV

H1a Seller Reference

Client
Firm Size Selection
H5a–c
H1b

H2a
Length of
Client Relationship
H2b

Business
BRV
Reference Reference Value
H3
Media Format ( )

H4
Reference
Congruency

Defining, Measuring, and Managing Business Reference Value / 71

size of the firm in the marketplace, whether it is based on H2a: Seller firms are more likely to select client firms that
labor force or scale of operations. As a result, we expect have a longer relationship with the seller firm to a thresh-
that it is in the best interest of the selling firm to strategi- old (an inverted U shape) to be references than client
firms with a shorter relationship with the seller firm.
cally select larger client firms to be references. Thus:
H1a: Seller firms are more likely to select larger client firms to We also anticipate that the client’s behavior with the
be references than smaller client firms. seller firm in particular, and not just its firm size, can signal
to the prospect about the quality of the seller firm. For
When consumers rely on external information to help in example, Heide and John (1990) find that prospects can
purchase decisions, they often rely on the credibility of the view a closer relationship between a seller firm and current
source as well as the information that is being passed along client as a way to reduce the ambiguity of purchase. How-
(Gershoff, Broniarczyk, and West 2001). The source of the ever, research in the management literature also suggests
reference offers a signal to the prospect about the type of that overembeddedness beyond a certain threshold can be
firms that currently purchase from the selling firm (Herr, problematic because it may indicate to the prospect that the
Kardes, and Kim 1991). Research has shown that the credi- referencing firm is unfamiliar with potential alternatives
bility of the source of the WOM in both B2C and B2B cases (Uzzi 1997). As a result, we anticipate that the longer the
matters (Wangenheim and Bayón 2007). In addition, in relationship between the referencing client and the seller
many instances, especially those such as celebrity endorse- firm, the more positive (to a threshold) the signal will be to
ments (McCracken 1989), the perceived value of the prod- the prospect. Thus:
uct or service is directly tied to the source. Research has
also shown through interviews with managers that source H2b: Among the client firm firms selected as references, those
with a longer relationship with the seller firm to a thresh-
often plays a significant role in the effectiveness of business old (an inverted U shape) are more likely to have a
references (Godes 2008). Moreover, recent research has higher BRV than client firms with a shorter relationship
suggested that the value of the client firm’s reputation is with the seller firm.
often passed along to the seller firm through the reference
(Helm and Salminen 2010), where higher reputations tend Reference Media Format
to be linked to larger firms (Tversky and Kahneman 1974).
Thus, we anticipate that the client featured in the reference Research has shown that the way the content is delivered
provides a valuable signal to the prospect merely through and the quality of the information of any marketing com-
its identity in the marketplace, such that larger current munications can play a significant role in for industrial pur-
clients offer a positive signal to prospects about the quality chasing (Moriarty and Spekman 1984). Thus, the medium
of the products or services and the quality of the selling and specific format of the reference should also play a key
firm. Thus: role in determining the value of the reference. We define
“reference media format” as the type of reference provided
H1b: Among the client firms selected as references, larger firms such that different media formats change (1) the amount of
are more likely to have higher BRV than smaller firms. information that can be conveyed and (2) the amount of
uncertainty of the message content that can be alleviated
Length of Client Relationship (Daft and Lengel 1986). For example, research has shown
We expect seller firms to strategically select client refer- that richer modes of communication are more likely to
ences according to the level of embeddedness the client has influence customers to purchase (Venkatesan and Kumar
with the seller firm, where the embeddedness is often 2004) because they are perceived to have more valuable
stronger when the relationship between the seller and client information content, convey a greater effort by the firm to
firm is longer. Here we define length of client relationship communicate with the consumer, and potentially provide a
as the total expected time of the relationship between the more customizable opinion from the client. We anticipate
seller and client firm, which includes not only the length of that a reference that provides information in a richer media
the past relationship but also the expected length of the format is likely to be more effective in influencing a
future relationship. The structure and quality of ties prospect to purchase. Thus:
between the referencing client and the seller firm, where a H3: The media format of the reference affects the BRV of
more integrated structure and a higher quality relationship client firms such that client firms that provide references
lead to higher expectations of relationship length. Research in a richer media format are more likely to have a higher
has shown that longer relationships between organizations BRV than clients that provide references in a less rich
media format.
increase the likelihood of new product selection (Kaufman,
Jayachandran, and Rose 2006). To relate this research to
reference selection, it suggests that firms with longer rela- Reference Congruency
tionships are more likely to be selected as client references. Another key factor to consider is whether the referencing
However, research on buyer–seller relationships indicates client is congruent with the prospect. Here we define “refer-
that overembeddedness (too much relationship length and ence congruency” as the degree to which there are similari-
depth) actually creates a dark side to relationships (Wuyts ties between the referencing client and the potential client.
and Geyskens 2005), suggesting that seller firms are most For example, we observed from the qualitative interviews
likely to select clients with a moderate level of relationship that prospects may be interested in references from firms
length as references. Thus: from the same industry, firms that purchased similar prod-

72 / Journal of Marketing, January 2013

ucts or services, or even references from people who hold BRV
the same role within their firm (e.g., marketing, operations).
Recent research has shown that sellers can benefit from Measurement of BRV
selecting specific referrals for prospects that have a high To empirically test the hypotheses, we first need to develop
level of congruency (Hada, Grewal, and Lilien 2011). This a measure to compute the value of a business reference. We
homophily between the client providing the reference and propose that the value of a business reference is a function
the prospect is likely to generate trust and reciprocity of three components: (1) the amount of influence that client
(Goldenberg et al. 2009) and strengthen the bond (i.e., tie references (vis-à-vis other marketing elements) had on a
strength) between the two parties (Brown and Reingen prospect’s adoption, (2) the amount of influence that a
1987). As a result, we anticipate that references from clients given client reference (vis-à-vis other client references) had
with greater congruency with the prospect are more influen- on the prospect’s adoption, and (3) the profitability of the
tial in getting the prospect to adopt. Thus: prospect after adoption. We represent this mathematically
as follows:
H4: Client firms that provide references that have a higher
degree of congruency with the prospect firm are more N
Refn × DOIin × CLVn
likely to have a higher BRV than client firms that provide
references that have a lower degree of congruency.
(1) BRVi = ∑
n=1
(1 + r )t n
,

where
The Incremental Benefit of Firm Size
Research on commitment and trust suggest that when both BRVi = Business reference value of client reference i,
commitment and trust are present, “they produce outcomes Refn = Degree to which references affected the prospect
that promote efficiency, productivity, and effectiveness” n’s purchase decision,
(Morgan and Hunt 1994, p. 22). We predict that larger firms DOIin = Degree of influence of client reference i on con-
verted prospect n,
are a more trusted reference in terms of a stronger signal of
CLVn = CLV of converted prospect n,
quality to the prospect. This suggests that not only should
N = Total number of converted prospect,
larger firms send a positive signal to the prospect, which
r = Discount rate (in months), and
serves to reduce the uncertainty of the purchase decision,
tn = month that converted prospect n became a cus-
but they should incrementally strengthen the other drivers
tomer after the first month of the observation
of the value of the reference. Thus, after controlling for the window.
main effect of firm size, we hypothesize the following:
Equation 1 illustrates that we can compute the contribu-
H5a: When client firms are larger, it further strengthens the
tion of BRV from each prospect that chooses to adopt a
positive impact of (a) longer lengths of client relation-
ships on BRV, (b) richer reference formats on BRV, and
product or service by multiplying the degree of influence
(c) higher congruency on BRV. that client references, in general, had on the decision of that
prospect to adopt (Ref) by the degree that the given client’s
We provide a summary of the main hypotheses in Table 1 reference influenced the prospect to adopt (DOI), and by
and the numbers of the hypotheses on the links between the the profitability of the converted prospect (CLV). When we
constructs in our conceptual model in Figure 1. have this information from each converted prospect, we cal-

TABLE 1
Summary of Hypotheses
Hypothesis Expected Effect Rationale
H1a: Client firm size  selection + The larger client firm sends a positive signal,
H1b: Client firm size BRV + which makes it more likely to be selected as a
reference and more likely to have a higher BRV.
H2a: Length of client relationship  selection  Firms with longer relationships with the seller
H2b: Length of client relationship  BRV  firm potentially reduce risk and ambiguity of
purchasing. Overembedded firms (i.e., too long
of a relationship) can be problematic for both
selection and level of BRV.
H3: Reference media format  BRV + Richer media formats send a stronger signal
and more valuable information, leading to a
higher BRV.
H4: Reference congruency  BRV + Homophily between client and prospect
strengthen the value of the information leading
to a higher BRV.
H5a: Firm size relationship length  BRV + When firms are larger, it enhances the strength
H5b: Firm size reference media format  BRV + of the other drivers of BRV. This leads to an
H5c: Firm size reference congruency  BRV + incrementally higher value for BRV.

Defining, Measuring, and Managing Business Reference Value / 73

culate the contributions to each client’s BRV across all the marketing efforts are equal to profit. We convert this calcu-
converted prospects. Now that there is a conceptual under- lation from the aggregate to the customer level:
standing of the process to compute BRV, the next step is to
π ( Purchasesnt ) −
create a method for generating values for the three inputs of T
π ( Operating Costs nt ) − Marketingnt
the BRV equation. (2) CLVn = ∑
t=1 (1 + r )t
,

Computation of BRV
Step 1: Determine whether client references influenced where
adoption. The first step is to determine whether client refer- CLVn = CLV of customer n,
ences in general had any influence on the prospect’s deci- (Purchasesnt) = Profit from purchases of customer
sion to adopt. In this case, we expect that the adoption n in time t,
process is driven by two processes: seller-generated and (Operating Costsnt) = Profit lost from operating costs of
customer-generated marketing. Seller-generated marketing customer n in time t,
efforts tend to include any direct- or mass-marketing efforts Marketingnt = Marketing costs for customer n in
initiated by the seller firm. In this case, client-generated time t,
marketing transpires when the seller firm uses information T = Time horizon of the prediction, and
about the client in influencing the prospect (reference). We r = Discount rate (in months).
compute this value using a constant-sum scale approach. At
Equation 2 shows that the CLV for each new customer n
the time of adoption, we asked the converted prospect to do
is a function of three factors: (1) the profit from purchases,
the following:
(2) the operating costs, and (3) the marketing costs to retain
Allocate a total of 100 points between the following two the customer. In this case, profit from purchases includes
influences: (1) the influence of client references in general the contribution margin from purchases aggregated
and (2) the influence of all other marketing processes (e.g. monthly, profit lost from operating costs includes losses
direct, mass, etc.) and sales initiatives (e.g. personal sell- from product returns and/or costs of customer service
ing).
aggregated monthly, and marketing costs include the costs
We assume that the value given to the influence of client of any marketing efforts the firm expends to the individual
references is, in effect, the percentage of the sale that can be customer n aggregated monthly.
attributed to the influence of client references. We then set For our empirical example, we conducted a field study
the score that was given to the influence of references in during a one-year time period and then observed the pur-
general to the value of Refn. chases, product returns, and marketing efforts for the
remainder of the normal CLV prediction window, over a
Step 2: Determine the influence of each client’s refer- period of three years. In this case, we compute only the
ence. The second step is to determine the influence that actual CLV of converted prospects, not predicting it. In the
each client reference had on the prospect’s decision to case in which the entire prediction window is not observed,
adopt. We compute this value using a similar constant sum we could use the method that Reinartz, Thomas, and Kumar
scale approach as we did for Refn. At the time of adoption, (2005) suggest to predict the CLV of newly acquired cus-
assuming a converted prospect had a value of Refn > 0, we tomers. We then use the resulting value of BRV for each of
asked the converted prospect to do the following: “Allocate the client references as a dependent variable with the
a total of 100 points across each of the client references you antecedents of BRV as independent variables to uncover the
felt influenced your decision to purchase (where more drivers of BRV.
points means the reference had a greater influence).”
All 100 points could be allocated to one client reference
if all other references had no impact, or the points could be An Empirical Example
allocated across any combination of client references To explain the influence and value of client references, we
according to the perceived influence. The value given to provide two empirical examples from two B2B firms, one
each client reference in this step is, in effect, the proportion in the telecommunications industry and another in the
of influence that a given client’s reference had relative to financial services industry. In the following sections, we
the total influence that client references had on a converted describe the data we collected from these two firms, the
prospect’s decision to adopt. We attributed the score given to selection and operationalization of the variables for the con-
structs, and the method used for estimating the model to
the influence of each client’s reference to the value of
empirically determine the drivers of BRV.
DOIin.
Step 3: Compute the CLV of the converted prospect. The Data
third step is to compute the expected future profit, or CLV, The telecommunications firm provides products and ser-
of the converted prospect n. We do so by adapting an equa- vices that include landline phone service, voice over Inter-
tion Petersen and Kumar (2009) use to measure a cus- net Protocol, Internet service, wireless service, maintenance
tomer’s CLV. Petersen and Kumar show at an aggregate contracts, and infrastructure, among other products. The
level how total customer purchases, total operating costs financial services firm provides products and services that
(i.e., product returns/service costs), and total firm-initiated include business checking/savings accounts, payroll deposit

74 / Journal of Marketing, January 2013

services, lines of credit, loans, insurance, and investments June 2010). We used this information for computing the
among other products. The data collection for both firms values of the referencing clients and converted prospects
spanned a period of six years and included a one-year field during the three-year data collection period, which included
study, during which we collected information about the one year of the field test and two years post–field test (i.e.
influence of client references. Figure 2 provides a timeline realized CLV).
of the data collection process for both firms.
In the first three years (from July 2004 to June 2007), Estimation Method
the firms collected historical transaction and firmographic To empirically determine the drivers of BRV, we must
data from the clients that were chosen to be references for accommodate four key statistical issues: sample selection
prospects. In this case, many of the firms that agreed to be bias/truncation, disproportionate stratification, censoring,
used as references were willing to do so without any incen- and unobserved heterogeneity. First, with regard to sample
tive. The only compensation given to the seller firm was selection bias and truncation, we anticipate that the seller
funds to cover any costs incurred while generating the ref- firm strategically selects the clients that are going to be
erence—where the client chose the reference format. There used as references. The telecommunications firm selected
were no rewards (e.g., promotional discounts) for agreeing 88 from a total of 5109 clients, and the financial services
to be used as a reference. There were 88 clients from the firm selected 94 from a total of 5576 clients as references.
telecommunications firm and 94 clients from the financial This sample selection problem results in a truncation prob-
services firm used as references during the one-year field lem because the only clients for which we can observe a
study. The data collected from these clients included value for BRV are those that are selected by the firm and
exchange (e.g., transaction and marketing information) and agreed to be a reference. To accommodate this issue, our
descriptive (e.g., firmographics) information. In the one- modeling framework must be able to uncover the drivers of
year period immediately following the first three years reference client selection by the firm and accommodate this
(from July 2007 to June 2008), both firms conducted a field strategic selection bias from the drivers of BRV.2 To accom-
study in which prospects were shown all the references modate the issue of selection bias, we implement a binary
from the clients that were made available to them during the choice model to uncover the drivers of reference client selec-
sales process. Then, all prospects that made their first pur- tion. We then use the result of this estimation to correct for
chase during this field experiment period filled out a survey, the sample selection bias (the cause of the truncation prob-
answering two questions about the client references as lem) found in the parameter estimates of the drivers of BRV.
described in Steps 1 and 2 from the previous section. Second, we face a problem pertaining to disproportion-
There were 31 converted prospects from the telecom- ate stratification, given that such a small percentage of
munications firm and 37 from the financial services firm
2In this study, both firms already had established customer
that joined during the field experiment period. The firms
then tracked the exchange characteristics, which included bases. However, if a new firm were to apply this framework and
had only a few customers from which to choose, it would also be
the transactions (both purchases and operating costs) and important to understand the type of potential client firm that is
marketing information of the referencing clients and the likely to be an early adopter because the goal of the seller firm is
converted prospects, for two more years (from July 2008 to likely acceleration of diffusion.

FIGURE 2
A Timeline of the Data Collection

6 Years

Begin Collection Field Test End Collection

7/04 6/07 6/08 6/10

The firms collected data The firms conducted a 12- The firms continued to
for three years from the month field test in which they collect data for two years
clients who were used for gathered data on the influence from both the referencing
clients and newly acquired
business references. of references on new customer
customers.
adoption.

Defining, Measuring, and Managing Business Reference Value / 75

clients were selected for being a reference out of the total Client firm size. Research suggests that there are multi-
clients available to the firm. In this case, a little more than ple measures of firm size that rely on both labor force and
1% for both the telecommunications firm (88 of 5109) and scale of operations (Kimberly 1976). Moreover, these mea-
financial services (94 of 5576) firm were selected as refer- sures may not always be highly correlated with one another.
ences. To obtain insights into a firm’s decision to select Thus, we operationalize client firm size on both the labor
clients to be references, we want to maximize the sample force dimension (number of employees) and the scale of
variance in the dependent variable. To do so, we would ide- operations dimension (annual revenue).
ally want an equal proportion of ones and zeros in the sample Length of client relationship. The length of the relation-
(Lancaster and Imbens 1991). However, it is also possible ship between the client and the seller firm can be viewed as
not to lose much information in the coefficients even when both the amount of time the relationship has existed and the
the split is 80/20 rather than 50/50 (Cramer, Franses, and expected time the relationship is likely to exist into the
Slagter 1999). Thus, to maintain a larger sample and still future. This suggests that both the length of time the
maximize the information from the data set, we dispropor- alliance has existed (past relationship duration) and the
tionately sample from the entire data set to create a sample depth of relationship (expected future relationship duration)
with 20% of the clients that are references and 80% of the between the client and seller firms are important factors in
clients that are not used as references. We can then correct the measure of the length of client relationship (Gulati and
for the bias introduced by the disproportionate sample size Gargiulo 1999). To this end, we operationalize length of
straightforwardly by using a choice model with a traditional client relationship as both the length (tenure) and the depth
logit specification and an offset based on the disproportion- (CLV) of the relationship between the seller and the client.
ate stratification (Donkers, Franses, and Verhoef 2003).
Reference media format. We measured reference media
With regard to censoring, of the clients selected as ref-
format using dummy variables coded by format type of the
erences, the value of the BRV observed is greater than or
media used to provide the reference from the client firm to
equal to 0 (given that CLV was positive for new clients). the potential client firm. These formats include video testi-
This occurs because the prospects that adopt decide which monial, audio testimonial, written testimonial, case study/
client references influenced their decision to purchase. We white paper, and “call me” (an invitation to call the reference
observe a positive value for BRV only when at least one directly).
prospect found the reference valuable and 0 when no
prospects indicated that a given reference was influential. Reference congruency. Reference congruency depends
To address this issue, we ran a censored regression model in on the degree of similarity between the referencing firm and
which the dependent variable is the BRV score for the the potential client firms that find the referencing client
clients that were selected to be references and the value of firm’s reference value. Thus, we operationalized reference
BRV is censored at 0. congruency as the ratio of potential client firms that found
To account for the final modeling issue, unobserved the reference valuable that were from the same industry,
heterogeneity, we ran a latent class censored regression bought the same product or service, or had the same role
model to determine the number of potential segments after within their organization as the total number of potential
we account for the variance through the focal variables of client firms that found the reference valuable. Table 2 pro-
this study. For the full details on the sampling steps; selec- vides details of the variables and their operationalizations.
tion, transformation, and censored regression equations;
and the log-likelihood functions, see Appendix A. For Results
details of the latent class segmentation, see Appendix B.
Overall Model Fit
Measures
Table 3 presents the parameter estimates along with the
We need to identify a set of predictor variables to be used in model fit statistics. We found that the full models (both the
our model of the drivers of reference client selection and selection model and the latent class censored regression
BRV. For the selection equation, we include the two key model) for each of the firms provided good overall fit (R2 =
factors the firm is likely to use to select clients to give ref- .71 for the telecommunications firm, and R2 = .68 for the
erences: (1) client firm size and (2) length of client relation- financial services firm) with all variables being significant
ship. For the BRV equation, we include the four key factors predictors of either client selection or BRV. In general, for
that may influence of the value of a client reference to the the telecommunications firm, we find that together client
prospect along with the interaction effects and the selection firm size and length of client relationship (the two con-
bias correction: (1) client firm size, (2) length of client rela- structs based on client characteristics) explain approxi-
tionship, (3) reference media format, and (4) reference con- mately 31% of the variance of BRV, congruency explains
gruency. For the sake of simplicity and generalizability, we approximately 17%, reference media format explains
aggregated the data for each industry, for product/service approximately 14%, and the interaction effects between
category, and for functional area into seven categories to test client firm size and the other three constructs explain
congruency. (For a list of these categories, see Web Appen- approximately 9%. We found similar results for the finan-
dix W1 at www.marketingpower.com/jm_webappendix.) cial services firm. This suggests that references selected on
The following subsections discuss measurement of the four the basis of the client characteristics alone (a potential
key drivers. benchmark model) explain approximately 31% of the vari-

76 / Journal of Marketing, January 2013

 TABLE 2
Variable Operationalization
Variable Operationalization
Client Firm Size
Employees The number of employees in the referencing client
Revenue The annual revenue of the referencing client (in millions of dollars)
Length of Client Relationship
CLV The customer lifetime value of the referencing client (in thousands of dollars)
Tenure The time the referencing client has been a customer (years)
Reference Media Format
Video testimonial 1 if the reference was a video file, 0 if not
Audio testimonial 1 if the reference was an audio file, 0 if not
Written testimonial 1 if the reference was a testimonial, 0 if not
Case study/white paper 1 if the reference was a case study/white paper, 0 if not
“Call me”a 1 if the reference was for a personal call, 0 if not
Reference Congruency
Industry (Number of converted prospects from the same industry that indicated this
client’s reference influenced their purchase)/(Number of converted prospects
that indicated this client’s reference influenced their purchase)
Product/service (Number of converted prospects that purchased the same product/service
that indicated this client’s reference influenced their purchase)/(Number of
converted prospects that indicated this client’s reference influenced their
purchase)
Role (Number of converted clients whose decision maker was in the same role
that indicated this client’s reference influenced their purchase)/(Number of
converted prospects that indicated this client’s reference influenced their
purchase)
Congruency Average number of congruency matches for each converted prospect on
three congruency factors from prospects that indicated this client’s reference
influenced their purchase
Interaction Effectsb
Firm size × relationship length Client firm size length of client relationship
Firm size × reference media format Client firm size reference media format
Firm size × congruency Client firm size congruency
aWe do not include “call me” in the regression because each of the references must be categorized into one of the five categories listed, and
using all five would cause perfect multicollinearity.
bFor the interaction effects, “client firm size” is represented by the revenue variable. We found similar results when we used employees.
“Length of client relationship” is represented by tenure. We found similar results when we used CLV. “Reference media format” was computed
as follows: It is ordered from 5 to 1 according to the richness of the media used for the reference (video = 5, audio = 4, written = 3, case study/
white paper = 2, and “call me” = 1). “Congruency” is represented by the congruency variable.

ance of BRV. By selecting references according to fit (con- .02). This finding suggests that larger clients are more likely
gruency) and reference media format, and taking into to be selected to provide a reference. In addition, we found
account the synergistic value of client firm size, the addi- that revenue has a positive effect on selection for both the
tional variance of BRV that the model explains is approxi- telecommunications firm (.00026, p < .01) and the financial
mately 40%. services firm (.00046, p = .02). This suggests that the larger a
From the results of the latent class analysis, we found that client’s own revenue, the more likely a client will be selected
there is only one latent segment for our model; thus, we only to provide a reference. These findings support H1a, that the
report the results for the single-segment censored regression seller firm is more likely to select larger clients, in terms of
model. (For the results of the null models, along with the both labor force and scale of operations, as a reference.
full model, for the financial services firm, see Web Appen- Length of client relationship. For length of client rela-
dix W2 at www.marketingpower.com/jm_webappendix; tionship in the selection model, first, we found that CLV has
note that we found similar results for the telecommunica- an inverted U-shaped effect on selection for both the
tions firm.) In addition, the parameter estimates from the telecommunications firm (for CLV: .102, p < .01; for CLV2:
models provide several key insights into the effect of each –.00024, p = .02) and the financial services firm (for CLV:
variable on the BRV for each of the referencing clients. .124, p < .01; for CLV2: –.0003, p = .01). This suggests that,
first, the firm is more likely to select referencing clients that
Selection Model
provide more profit and are more likely to have a longer
Client firm size. For client firm size in the selection relationship duration, to a threshold. Second, we find that
model, we found that the number of employees has a posi- tenure has a positive effect on client selection for both the
tive effect on selection for both the telecommunications firm telecommunications firm (.163; p < .01) and the financial
(.062, p < .01) and the financial services firm (.076, p = services firm (.149; p = .02). This suggests that the seller

Defining, Measuring, and Managing Business Reference Value / 77

 TABLE 3
Parameter Estimates
Variable Telecommunications (SD) Financial Services (SD)

A: Selection Model

Intercept .18 (.049) .11 (.041)

Client Firm Size
Number of employees .062 (.021) .076 (.032)
Annual revenue .00026 (.0001) .00046 (.0002)
Length of Client Relationship
CLV .102 (.028) .124 (.044)
(CLV)2 –.00024 (.0001) –.0003 (.00012)
Tenure .163 (.056) .149 (.063)

B: BRV Model

Intercept 3.14 (1.06) 3.04 (1.06)

Client Firm Size
Number of employees .231 (.082) .201 (.082)
Annual revenue .0086 (.003) .019 (.0063)
Length of Client Relationship
CLV .339 (.052) .409 (.052)
(CLV)2 –.0072 (.002) –.0076 (.0015)
Tenure .471 (.091) .394 (.082)
Reference Media Format
Video testimonial .256 (.042) .308 (.059)
Audio testimonial .204 (.033) .246 (.047)
Written testimonial .176 (.057) .208 (.042)
Case study/white paper .112 (.038) .141 (.056)
Reference Congruency
Industry .464 (.087) .374 (.066)
Product/service .626 (.144) .521 (.142)
Role .291 (.083) .238 (.098)
Interaction Effects
Firm size × relationship length .004 (.0016) .008 (.0027)
Firm size × reference media format .0027 (.0008) .006 (.0022)
Firm size × congruency .0052 (.0017) .0081 (.0024)
Sample Selection Correction
Lambda () .42 (.109) .41 (.118)
Model Fit
R2 .71 .68
Log-likelihood –178.61 –169.42
AIC 417.81 401.4
Notes: All variables in the model are significant at p < .05.

firm is more likely to select clients that have had a longer Censored Regression (BRV Model)
prior relationship as a reference. In general, these findings
Client firm size. We found that both variables that
support H2a, which predicts that, to a threshold, seller firms
describe the client firm size of the referencing client for the
are more likely to select clients that have had and are
telecommunications firm (for employees: .231, p < .01; for
expected to have longer relationships with the seller firm as
a reference. revenue: .0086, p < .01) and the financial services firm (for
employees: .201, p < .01; for revenue: .019, p < .01) are
Summary. In general, all these variables and the signifi- positive and significant, in support of H1b. “This finding
cance of the parameter estimates suggest that the firm does suggests that prospects are more likely to value references
not randomly select the clients that provide references. from clients that are larger, in terms of either labor force or
Instead, the firm is probably selecting larger clients with scale of operations.
stronger relationships with the seller firm (to a threshold).
This finding is also supported by the positive and signifi- Length of client relationship. For length of client rela-
cant coefficient on the pseudo inverse Mills ratio for both tionship, we found that CLV had an inverted U-shaped rela-
the telecommunications firm (.42, p < 0.01) and the finan- tionship for both the telecommunications firm (for CLV:
cial services firm (.41, p < .01), further validating the need .339, p < .01; for CLV2: –.0072, p < .01) and the financial
to accommodate for this selection bias. services firm (for CLV: .409, p < .01; for CLV2: –.0076, p <

78 / Journal of Marketing, January 2013

.01). We also found that tenure has a positive relationship increases, so does the BRV of that referencing client. For
for both the telecommunications (.471, p < .01) and finan- the product congruency variable, we found a positive rela-
cial services (.394, p < .01) firms. In this case, the longer tionship for both the telecommunications (.626, p < .01)
the client has been purchasing from the firm and the greater and financial services (.521, p < .01) firms. This suggests
the likelihood of a longer future relationship duration, the that as the percentage of prospects that valued the reference
greater was the influence of the reference from that client and purchased the same product or service increased, the
on a prospect’s decision to adopt. In general, these findings BRV of the referencing client increased. For the role con-
support H2b, suggesting that longer relationships between gruency variable, we found a positive relationship for both
clients and the seller firm (to a threshold) lead to a higher the telecommunications (.291, p < .01) and financial services
BRV. (.238, p = .02) firms. This suggests that as the percentage of
Reference media format. For all reference media format prospects that valued the reference and held the same role
variables, we found a positive relationship for both firms. as the person in the referencing client increased, the BRV of
The specific values are as follows: for the video testimonial the referencing client increased.
variable, telecommunications: .256, p < .01; financial ser- Our findings suggest that increasing the opportunities
vices: .308, p < .01; for the audio testimonial variable, for congruency can add significant value to a client’s refer-
telecommunications: .204, p < .01; financial services: .246, ence (e.g., through strategically having a broad set of refer-
p < .01; for the written testimonial variable, telecommuni- ences from which to choose). However, the type of congru-
cations: .176, p < 0.01; financial services: .208, p < 0.01; ency can add differing levels of value. We find that for both
and for case study/white paper variable, telecommunica- firms, the value provided by different reference congruen-
tions: .112, p < .01; financial services: .141, p = .02. cies follows the same order in terms of importance: Product/
We ordered the five reference media format variables service congruency is the most valuable (when compared
according to the size of the parameter estimates from high- with industry congruency, for the telecommunications firm:
est to lowest: (1) video testimonial, (2) audio testimonial, t = 9.05, p < .01; for the financial services firm: t = 9.07, p <
(3) written testimonial, (4) case study/white paper, and (5) .01); industry congruency is the second most valuable
“call me.”3 In addition, we tested to determine whether (when compared with role congruency, for the telecommu-
there is a significant difference between the levels of each nications firm: t = 13.52, p < .01; for the financial services
of the parameter estimates using a pairwise t-test. We found firm: t = 11.15, p < .01); and role congruency is the third
that in both firms, the parameter estimate for video testimo- most valuable in terms of BRV. Thus, if a prospect is inter-
nial is statistically significantly larger than the parameter ested in a reference, although all congruencies are impor-
estimate for audio testimonial (for telecommunications: t =
tant, the most influential references will be product/service
9.12, p < .01; for financial services: t = 7.95, p < .01). We
related first, industry related second, and role related third.
found that in both firms, the parameter estimate for audio
The finding that product/service and industry (i.e. market)
testimonial is statistically significantly larger than the
are the most important congruency factors supports the
parameter estimate for written testimonial (for financial ser-
management literature that focuses on product–market fit as
vices: t = 4.00, p < .01; for telecommunications for finan-
a surrogate for congruency (Alderson 1965).
cial services: t = 5.85, p < .01). For both cases, we found
that the parameter estimate for written testimonial is statis- Interaction effects. We found each of the interaction
tically significantly larger than the parameter estimate for terms to be positive and significant for both the firms. For
case study/white paper (for telecommunications: t = 8.77, p < client firm size × length of client relationship, we found a
.01; for financial services: t = 9.31, p < .01). For both cases, coefficient of .004 (p = .02) for the telecommunications
we found the parameter estimate for case study/white paper firm and .0008 (p < .01) for the financial services firm. For
is statistically significant, meaning it is larger than the value client firm size × reference media format, we found a coef-
added by “call me,” which can be found in the intercept (for ficient of .0027 (p < .01) for the telecommunications firm
telecommunications: t = 2.96, p < .01; for financial ser- and .006 (p = .01) for the financial services firm. For client
vices: t = 2.87, p < .01). In general, this finding supports firm size × reference congruency, we found a coefficient of
H3, suggesting that client references with richer media con- .0052 (p < .01) for the telecommunications firm and .0081
tent tend to influence prospects more than references with (p < .01) for the financial services firm. These findings sup-
less rich content. port H5a–H5c, suggesting that larger client firms have a syn-
Reference congruency. For the industry congruency ergistic and positive effect on the value of the reference
variable, we found a positive relationship for both the across the other constructs.
telecommunications (.464, p < .01) and financial services Summary. In general, all these variables and the signifi-
(.374, p < .01) firms. This suggests that as the number of
cance of the parameter estimates suggest that we have
prospects in the same industry as the referencing client
uncovered many of the drivers of BRV, or the level of influ-
ence a given reference will have on getting a prospect to
3First, we omitted “call me” from the regression and treated it
adopt. Thus, when accounting for the potential of selection
as the base category for comparison. Second, it could be the case
that simply providing a client list offers some value. Because all bias, seller firms can now determine which references in
the references were provided in each sales situation, we cannot their client reference database are likely to be the most
directly control for simply being on the client list in this context. valuable in a given sales situation.

Defining, Measuring, and Managing Business Reference Value / 79

 Does High BRV = High CLV? medium CLV customers did. In this case, there is no incen-
tive, and the BRV of the clients in the top deciles is still
A key finding from Kumar, Petersen, and Leone (2007,
relatively high. The highest deciles still have a high BRV
2010) is that individual customers with the highest cus-
because of client firm size (i.e., higher client firm size 
tomer referral value (CRV) are not the same customers as
higher BRV). However, the medium CLV clients that give
those with the highest CLV. Thus, it is important to manage
references have the highest BRV due to the length of client
customers on both their CLV and their CRV scores. In the
case of BRV, it is also important to determine whether relationship (i.e., medium relationship duration  higher
clients that have a high CLV are the same as those that pro- BRV) and the congruency effect (higher congruency 
vide the most valuable references. To do this, we ordered higher BRV). With regard to congruency, we found that many
the client reference firms from the telecommunications and of the prospects being targeted for acquisition had the most
financial services firms that were part of the field test by in common with the customers in the medium CLV group.
their realized CLV during the three years at the beginning of We also observe that the average BRV scores in Deciles
the field test. We divided these clients into ten deciles and 1–6 for the telecommunications firm and Deciles 1–7 for
computed the average CLV and BRV for each decile. Table the financial services firm are higher than the average CLV
4 presents the results. scores for the clients in those deciles. This suggests that get-
Table 4 shows that, similar to the findings regarding ting a medium- to high-value business client to provide a
CRV, the clients with the highest BRV are not the same as reference can be potentially much more profitable than
the clients with the highest CLV. For both firms, the clients merely focusing on cross- and up-selling opportunities to
with the highest BRV fall into the third, fourth, and fifth that business client. To provide some additional insights
deciles on CLV (fifth having the highest BRV in both into the type of clients that provide high- versus low-value
firms). We observe that referencing clients in the first two references, we segmented the referencing clients using a
deciles (highest CLV) are also higher on BRV compared median split of BRV (see Table 5).
with the clients in the lowest four deciles (seventh through Table 5 shows significant differences between the pro-
tenth), which have by far the lowest average BRV. files of clients that are high on BRV for both firms. As we
When comparing CLV with CRV, Kumar, Petersen, and expected, the CLV of the high-BRV clients is much higher
Leone (2007, 2010) find that the customers with the highest than that for the low-BRV clients (for the telecommunica-
CLV did not have a relatively high CRV. This was possibly tions firm, 18,400 vs. 5800; for the financial services firm,
the case because of the incentives provided to encourage 18,700 vs. 5200). We observe a significant difference in the
referral behavior. Customers with the highest CLV probably tenure between the two groups: Clients that have a higher
did not value the referral incentive to the degree that the BRV have longer tenure (for the telecommunications firm,

TABLE 4
Decile Analysis for BRV and CLV
Telecommunications Financial Services
(n = 9 for All but Tenth Decile) (n = 9 for All but Tenth Decile)

Decile CLV (in Thousands) BRV (in Thousands) CLV (in Thousands) BRV (in Thousands)
1 30.8 34.6 26.2 31.2
2 25.7 40.8 23.6 33.6
3 20.2 49.6 20.5 41.8
4 17.3 55.8 18.1 59.2
5 14.9 61.2 15.7 66.8
6 12.1 30.2 12.8 36.1
7 9.3 6.2 9.6 10.2
8 6.4 3.1 5.5 4.1
9 3.2 1.8 2.9 2.2
10 .8 .2 .4 .18

TABLE 5
Segment Description for High and Low BRV Clients
Telecommunications Financial Services

Variable High BRV (n = 44) Low BRV (n = 44) High BRV (n = 47) Low BRV (n = 47)
Average CLV (in thousands) 18.4 5.8 18.7 5.2
Average tenure (years) 10.3 4.9 14.7 6.8
Most common media format Video “Call me” Video “Call me”
Average number of employees 2710 468 2158 318
Average annual revenue 59.4 11.2 70.6 18.8
(in millions of dollars)

80 / Journal of Marketing, January 2013

10.3 years vs. 4.9 years; for the financial services firm, 14.7 role in the purchase decisions of prospects for both the
years vs. 6.8 years). We observe that the clients that have financial services and telecommunications firms, with, on
high BRV are more likely to have provided a video refer- average, more than 50% of the purchase influence resulting
ence than a “call me” reference for both the telecommuni- from the client references. Although at first glance, the
cations and financial services firms. Finally, in terms of influence of client references may seem high, this finding is
firm size, the average number of employees (for the not surprising given some recent evidence in the market-
telecommunications firm, 2710 vs. 468; for the financial place. First, a recent report found that more than 90% of
services firm, 2158 vs. 318) and the average annual revenue survey respondents stated that the demand for business ref-
(for the telecommunications firm, $59.4 million vs. $11.2 erences was higher in 2007 than in 2006 (Wood 2007).
million; for the financial services firm, $70.6 million vs. Additional evidence indicates that for three customer refer-
$18.8 million) are higher for the high BRV clients than for ence programs at technology firms, references played an
the low BRV clients. The results from this profile analysis important role for customer acquisition efforts by both the
add more support to the endogenous decision the firm is marketing and sales departments (Godes 2008). Thus, while
already making with regard to selecting clients for refer- traditional marketing and sales force efforts continue to be
ences that generally have a higher CLV, longer tenure, relevant to customer acquisition, references from current
larger employee bases, and higher annual revenues. In addi- business customers are playing an increasingly prominent
tion, it suggests that in the future, the selection of the clients role in influencing new customer adoption. This means that
can be even more targeted to the prospects to increase the it is necessary in most cases not only to develop a usable
probability of customer acquisition and the CLV of con- business reference program but also to identify which
verted prospects. clients have the potential to bring value in customer acqui-
sition efforts.
To shed light on whether prospects value many different
The Impact of Refn, DOIin, and CLVn client references more than only a few or one key client ref-
on BRV erence, we investigated the mean number of client refer-
Our findings show that using client references can add sig- ences for each converted prospect that had DOIin > 0 for
nificant value to the new customer acquisition process. It is both the financial services ((number of references per
also important to understand, at least in a general sense, the newly acquired customer with positive influence [DOIin >
relative impact of each of the three components on BRV. 0]) = 2.1 and
= .5) and telecommunications ((number of
For example, do clients with a high BRV influence the references per newly acquired customer with positive influ-
small number of prospects to a great extent (i.e., high Refn ence [DOIin > 0]) = 1.5 and
= .4) firms. This finding
and DOIin for a given client i with a small frequency of n)? shows that, on average, converted prospects use one or two
Alternatively, do clients with a high BRV provide a small clients as key references to influence their decision to
amount of influence on a large number of prospects (i.e., adopt. However, the particular references selected in each
low Refn and DOIin for a given client i with a large fre- case vary widely across the set of converted prospects.
quency of n)? First, we want to know how well the three Again, this finding is not necessarily surprising given the
determinants of BRV are correlated (see Table 6). empirical results of this study. For example, we found that
Table 6 indicates that Refn, DOIin, and CLVi are posi- reference congruency played an important role in a refer-
tive but weakly correlated. The only significant correlation encing client’s BRV score. This means that prospects put
(p < .10) is between Refn and CLVi (r = .16). This suggests more weight on client references that match with their situ-
that newly acquired client firms that are ultimately the most ation, whether it is based on industry, product/service, or
role congruency. Given that only a few clients have high
profitable are more likely to find references more influential
congruency in each new customer acquisition setting, it
in their adoption than less profitable newly acquired client
makes sense that, on average, only a few client references
firms. However, these results show that, in general, the
mattered in influencing prospect adoption. In addition, it
three determinants of BRV are uncorrelated.
shows the importance of having a variety of references
To understand how references in general affect the deci-
from clients in different industries that purchase different
sion of a prospect to adopt, we investigate the means and
products and services and from people from different func-
standard deviations of Refn across all converted prospects
tional areas within the client’s firm; this variety is necessary
for the financial services ((Refn) = 55 and
(Refn) = 19)
to increase the probability that reference congruency can be
and telecommunications ((Refn) = 62 and
(Refn) = 17)
achieved for a prospect.
firms. These results suggest that client references play a key
Finally, to understand what proportion of the value gen-
TABLE 6 erated by prospects through their CLV was being attributed
Correlation Between Refn, DOIin, and CLVi to BRV, we investigated the mean and standard deviation of
CLV for all the converted prospects for the financial ser-
Refn DOIin CLVi vices ([CLVn] = $14,064 and
[CLVn] = $5,818) and the
Refn 1 telecommunications firm ([CLVn] = $15,469 and
[CLVn] =
DOIin .12 1 $7,963). This finding shows that, on average, the converted
CLVi .16* .08 1 prospects that joined during the one-year time period of the
*Denotes significance at p < .10. (All other pairwise correlations are field study were profitable customers for the financial ser-
not significant.) vices and telecommunications firms. If we compare the

Defining, Measuring, and Managing Business Reference Value / 81

mean values with those in Table 5 (i.e., the mean CLVs of highest effect size). Finally, we find that when larger clients
the referencing clients), we find that, on average, the con- also have a longer relationship with the seller firm, a refer-
verted prospects have a slightly lower CLV than the high- ence in a richer media format, and/or are more congruent
BRV referencing clients but have a much higher CLV than with the prospect the reference has the ability to provide an
the low-BRV referencing clients for both firms. This sug- incrementally larger degree of adoption influence—sug-
gests that business reference programs can be successful in gesting that larger client firms add direct and indirect value
attracting profitable customers, some of which could serve to social influence.
as successful referencing clients in the future, a potential A third key contribution of this study is in shedding
topic for further research. light on the impact of Refn, DOIin, and CLVn on BRV. Our
findings from this additional analysis can provide insights
to firms that aim not only to begin but also to run successful
Implications business reference programs. We show that client refer-
This research yields several key implications for both the ences play an ever-increasing role in the decision for a
marketing literature and practice. First, we introduce the prospect to adopt, making it necessary to have a catalog of
concept of BRV. Here, we define BRV as the ability of a client references ready for the sales team to use during the
client’s reference to provide value to the seller firm and the sales process. However, we show that usually only a few
degree to which it does so by influencing a prospect to client references seem to play a key role in influencing
adopt. We show that there are three components of BRV: (1) prospects to adopt and that the key client references varied
the influence that all client references have on newly significantly across converted prospects. This suggests that
acquired customers compared with other forms of market- sellers should have a portfolio of client references, includ-
ing, (2) the influence each specific client reference has on ing clients with varied characteristics, to match the client to
converted prospects, and (3) the profitability of the con- the prospect successfully and that it might be more effective
verted prospects. We also demonstrate how a firm can com- to bring in only a limited set of “targeted” references in the
pute BRV using a straightforward three-step process that sales process. Finally, we show that running a successful
retrospectively reports a measure of reference value. The business reference program can help in converting prospects
process requires two items: (1) a survey of the newly to clients that are profitable to the firm.
acquired customers at the time of adoption and (2) a mea-
sure of the profitability of the newly acquired customers.4
Second, we identify and empirically test the drivers of Limitations and Further Research
BRV. Identifying the key drivers is critical to select the best We should point out a few limitations and research opportu-
clients to serve as references and also to understand how nities to consider when measuring BRV and attempting to
best to deliver the reference message (format). We show implement a successful reference program. First, we mea-
that there are four key drivers of BRV: (1) client firm size, sured BRV for only two firms. While there is a significant
(2) length of client relationship, (3) reference media format, number of other firms in these industries that could benefit
and (4) reference congruency. We find that the source of the from this study, client references may not be a key driver of
reference matters, in terms of both firm size and length of customer acquisition in some industries, or the drivers of
relationship. Thus, it is important that managers select the reference value may vary in their importance from this con-
right clients to use as references. We also find that the refer- text. For example, in more complex selling environments
ence media format is an important driver of BRV, such that (e.g., marketing and consulting projects, firms selling cus-
richer media formats are more influential. Importantly, we tomized industrial goods and solutions), specific project,
found that all the reference format variables have a larger rather than product, attributes or degree of project success
effect than the “call me” approach, indicating that it is (e.g., 15% return on investment) may be most relevant.5 In
worthwhile to provide even a single client reference beyond addition, other variables such as prior product or category
simply offering the opportunity to call a current client; this experience might help explain the difference the value of a
type of reference contains no immediate value to the reference can have when prospects have low versus high
prospect. In addition, we observed through our qualitative product knowledge. Further research should test this frame-
interviews that managers often do not believe they have the work by including potential additional drivers in other
time to connect with other managers, which offers some industries and general contexts (e.g., business to consumer)
additional external validity to these findings. We find that to determine its generalizability.
reference congruency significantly increases BRV. Thus, it The method we use to track the influence of client refer-
is important when possible to match clients with prospects ences is retrospective and requires the converted prospect to
that purchase the same product or service (highest effect evaluate the impact the client reference had on their deci-
size), are in the same industry (second highest effect size), sion to adopt. There may be a difference between the true
and/or have the same role/position within their firm (third impact of the reference and the self-reported impact for any
number of reasons. For example, if the sales cycle is
4Although there are other methods that could be used to survey extremely long and the client references were introduced at
customers, such as self-explicated conjoint analysis, we believe
that our method provides a simple and straightforward method to 5We thank one of the anonymous reviewers for providing this
obtain an accurate value for BRV. comment.

82 / Journal of Marketing, January 2013

the beginning of the sales process (and not again later), it fective references. Further research could consider how dif-
may be difficult for the converted prospect to remember ferent reference sets at different times during the adoption
which references influenced them or if client references process might have different levels of influence. This might
mattered at all. In some cases, converted prospects may not also help validate recent survey findings, which suggest that
remember the client reference and give it no value, when it different reference content might be influential at different
may be that the client reference was what brought the sales- times during the selling cycle (Godes 2008).
person to the door in the first place (making it critical to the
acquisition). Moreover, although we have no evidence to
suggest a problem, it might be the case that even if the Appendix A
client can remember the reference, the client may be unable Details on the Sampling, Equations,
to accurately report what influences the references had on and Log-Likelihood Function
their decisions and the weights to put on the different refer-
To show how the models are developed, we can begin with
ences. This is a problem with many studies that use self-
the standard logit model:
report measures. Further research could attempt to measure
BRV as a function of observed variables to alleviate this exp ( w′γ
i )
(A1) P ( zi = 1 w i ) = .
problem. 1 + exp ( w′γ
i )
The operationalization of the four key constructs (client
firm size, length of client relationship, reference media for- In this case, zi is 1 when selected and 0 otherwise, wi is a set
mat, and congruency) can be measured in several alterna- of explanatory variables, and  is a vector of parameter esti-
tive ways. For example, constructs such as client firm size mates. Next, we need to correct for the outcome-dependent
and length of relationship could be construed as client repu- disproportionate stratified sampling. We do this by includ-
tation and client embeddedness, respectively. While the ing the probability that a given observation comes from one
operationalizations for the four key constructs we used in of the two different strata—in this case, either those selected
this study were theoretically driven, further research could (1) or those not selected (0) to be a reference. Similar to
add to their richness in the reference value literature by test- Donkers, Franses, and Verhoef (2003), we create two new
ing alternative measures that capture more of the dimen- parameters us0 and us1. In this case, usk represents the per-
sions of the key constructs. centage of stratum k, which was selected into the sample,
Furthermore, the client may feel obligated to find value where usk = nsk/Nsk, where nsk is the number selected from
in the references (e.g., social desirability bias) given that the the stratum and Nsk is the total population of the stratum.
survey involves the influence of client references (Fisher The result is the following pseudolikelihood:
1993). In the current study, the sales cycles for the telecom-
munications and financial services firms were fairly short, ⎡ ⎛ us1 ⎞ ⎤
i + ln ⎜
exp ⎢ w′γ ⎥
on the order of a few months at most, and many of the deci-
⎣ ⎝ us0 ⎟⎠ ⎦
sion makers from the firm suggested that references were (A2) P ( zi = 1 w i , stratum k = s) = .
⎡ ⎛ us1 ⎞ ⎤
often used in the final decision-making process, likely less- 1 + exp ⎢ w′γ
i + ln ⎜ ⎥
⎣ ⎝ u s0 ⎟⎠ ⎦
ening the potential bias present in the self-report client ref-
erence influence measure.
Thus, we obtain the pseudolikelihood of the logit model
It is possible that a client firm could be a good reference
on many dimensions (e.g., larger, moderate relationship by adding a correction of ln(us1/us0) to offset wi for each
duration, high degree of congruency), but the product or observation. This can easily be accomplished in most statis-
service the seller firm offers is not appropriate for the tical software packages by offsetting the intercept estimate
prospect.6 In this case, the low product or service fit with of the standard logit model.
the potential client might overstate the unconditional impact To accommodate these issues simultaneously, we
of the reference. Although the data we collected in this develop a two-stage modeling framework with pseudo
study cannot directly address this issue, it would be a logit-ordinary least squares (OLS) specification. We imple-
worthwhile topic for further research. ment this two-stage framework in a similar way to the well-
Finally, we allowed all prospects to access all the client known Heckman two-stage model with one exception: We
references during the entire sales process. The results of this do not use a probit model specification in Stage 1. We
study indicate, for example, that reference congruency instead chose to use a logit model specification,because the
plays a major role in influencing prospect adoption. Further correction for disproportionate sampling is straightforward.
research, which would help validate the findings of Hada, Then, to link the logit model with the OLS model, we use a
Grewal, and Lilien (2011), could determine whether target- generalized econometric method as Lee (1983) describes.
ing specific prospects with a limited set of client references We can implement this model in the same way as the
would have a greater impact on prospect adoption by Heckman two-stage model with only a few minor modifica-
removing the cluttering effect of many undesirable or inef- tions. Specifically, we use the pseudo logit model in place
of the usual probit model, and we compute a pseudo inverse
6We thank one of the anonymous reviewers for raising this Mills ratio using the output of the pseudo logit model using
potential issue. the following three steps.

Defining, Measuring, and Managing Business Reference Value / 83

 ∑
1. We estimate the logit model with the offset correction for
disproportionate stratification. (A6) LL = {ln [1 − F ( w′γi )]}
2. We compute the pseudo inverse Mills ratio (*) using the { i z i = 0}
following equation:
⎧ ⎛ x′β
i ⎞
⎪⎪ ⎡ ⎛ y − x′β⎞ ⎤ ⎜⎝ ⎟⎠
λ* =
{
φ Φ−1 [ F ( w′γ
i )] }, (A6) LL + ∑ ⎨ ln ⎢φ ⎜⎝
⎣
i i
⎟ − ln σ
σ ⎠ ⎥⎦
σ
ρ2
* ⎪
F ( w′γ
i ) { i z i = 1, y i = y i } ⎪⎩
2. where
is the normal density function and F(.) is the logit ⎛ ⎧ −1 i ⎞ ⎫⎞ ⎫
⎛ y i − x′β
distribution function. ⎜ ⎪⎪ Φ [ F ( w′γ
i )] + ρ ⎜
⎝ σ ⎟⎠ ⎪⎪⎟ ⎪⎪
3. We run a censored regression and include the pseudo (A6) LL + ln ⎜ Φ ⎨ ⎬⎟ ⎬
inverse Mills ratio (*) as an additional variable.7 ⎜ ⎪ 1 − ρ2 ⎪⎟ ⎪
⎜⎝ ⎪ ⎪⎭⎟⎠ ⎪⎭
⎩
First, we select samples from the two strata that will be
⎧ − x ′β
i ⎫
used in our two-stage modeling framework. We select sam- ∞
⎪ ⎪
ples for the ones (reference clients) and zeros (nonreference (A6) LL + ∑ ⎨ln
i z i = 1, y i = 0 ⎪
∫ ∫
−∞
σ
− w ′γ
φ2 ( u, v, ρ) dudv⎬.
⎪⎭
clients) that generate an 80/20 distribution of zeros to ones. ⎩ i

This results in a sample size of 440 (88 ones and 352 zeros)
for the telecommunications firm and 470 (88 ones and 376
zeros) for the financial services firm. In this case, the ones Appendix B
represent 100% of the total stratum population, and the
zeros represent a random sample of approximately 7% of
Latent Class Segmentation
the stratum population for both the telecommunications and The purpose of the latent class segmentation is to uncover
financial services firms. We ran the models multiple times whether there are multiple homogeneous segments of client
with different random samples from the zeros and found no references in the portfolio of references for the telecommu-
significant differences in the model results for either of the nications or financial services firms. To determine if multi-
firms. We then use the following equations for our model- ple segments are present, we use the method Jedidi,
ing framework: Ramaswamy, and DeSarbo (1993) developed.
We carry out the latent class segmentation of clients by
Selection Equation:
setting the number of segments to different numbers and
(A3) z*i = w′γ
i + ui
then determining which segmentation scheme provides the
best results based on the consistent Akaike information cri-
⎧ 1 if z*i > 0 terion (CAIC). We found the following results for the finan-
zi = ⎨ * .
⎪⎩ 0 if z i ≤ 0 cial services firm8:
Number of Segments Log-Likelihood CAIC
Censored Regression Equation:
1 –169.42 416.26
(A4) y*i = x′β
i + e i , if z i = 1; and 2 –166.03 491.45
⎧ y* if y*i > 0 3 –162.71 566.78
yi = ⎨ i * . 4 –159.46 642.25
⎪⎩ 0 if yi ≤ 0
Given that the CAIC is lowest for the one-segment solu-
Transformation Equation: tion, we chose that as the optimal number of segments
(Jedidi, Ramaswami, and DeSarbo 1993), where the CAIC
(A5) u*i = J1 ( u i ) = Φ −1 [ F ( u i )]. is an alternative criteria to the AIC that corrects for the
overestimation bias in AIC by penalizing for overparame-
In these equations, u*i and ei are bivariate normal with
terization. It is defined as follows:
means of 0, standard deviations of 1 and
, and correlation
of , where F(ui) is the pseudo logit model specification as CAICm = −2 × ln ⎣⎡ L (Q m )⎤⎦ + N m ( ln I + 1) ,
described previously. In addition, z is the selection variable,
where if zi = 1 the client was chosen to be a reference and if where
zi = 0 the client was not chosen to be a reference. Finally, yi
is only observed when zi = 1 and yi is censored at 0—where L(Q|m) = The likelihood of the model given segment m,
BRVi = yi. The resulting log-likelihood function must Nm = The effective number of parameters estimated
accommodate the selection/truncation, disproportionate in an m-class solution, and
stratification, and censoring issue. As Maddala (1983) I = The number of observations in the sample.
notes, we get the following log-likelihood function: A one-segment solution means that there is no signifi-
cant benefit to segmenting clients into multiple latent seg-
7Given that the results of the latent class segmentation sug-
ments. However, a one segment solution does not mean that
gested one segment, we do not discuss the integration of the latent
class segmentation in our model estimation steps here. For a dis-
cussion of the latent class segmentation, see Appendix B. 8We find similar results for the telecommunications firm.

84 / Journal of Marketing, January 2013

there is no variation between segments. It only means that ments—both in magnitude and significance.9 However, in
the additional variance explained by dividing customers no cases did we observe the signs of the parameter esti-
into multiple segments is outweighed by the reduction in mates change. As a result, the slight, but not significant,
parsimony from the use of too many parameters. In addi- increases in variance explained when each subsequent seg-
tion, it also suggests that the variables used in the censored ment is added leads to the one-segment solution having the
regression model explained much of the heterogeneity lowest CAIC, likely because the sample size is small (94
for the financial services firm) relative to the number of
among clients used for references.
variables in the model that need to be estimated (17).
The parameter estimates for each of the four latent class
estimations (one, two, three, and four segments) only 9Parameter estimates for the two-, three-, and four-segment
showed some slight differences between each of the seg- solutions are available on request.

REFERENCES
Alderson, Wroe (1965), Dynamic Marketing Behavior. Home- sion: An Accessibility-Diagnosticity Perspective,” Journal of
wood, IL: Richard D. Irwin. Consumer Research, 17 (4), 454–62.
Brown, Jacqueline Johnson and Peter H. Reingen (1987), “Social Jedidi, Kamel, Venkatram Ramaswami, and Wayne S. DeSarbo
Ties and Word-of-Mouth Referral Behavior,” Journal of Con- (1993), “A Maximum Likelihood Method for Latent Class
sumer Research, 14 (3), 350–62. Regression Involving a Censored Dependent Variable,”
Connelly, Brian L., S. Trevis Certo, R. Duane Ireland, and Christo- Pyschometrika, 58 (3), 375–94.
pher R. Reutzel (2011), “Signaling Theory: A Review and Kaufman, Peter, Satish Jayachandran, and Randall L. Rose (2006),
Assessment,” Journal of Management, 37 (1), 39–67. “The Role of Relational Embeddedness in Retail Buyers’
Cramer, Mars, Philip Hans Franses, and Erica Slagter (1999), Selection of New Products,” Journal of Marketing Research,
“Censored Regression Analysis in Large Samples with Many 43 (November), 580–87.
Zero Observations,” Econometric Institute Research Report Kimberly, John R. (1976), “Organizational Size and the Struc-
9939/A, Erasmus University. turalist Perspective: A Review, Critique, and Proposal,” Admin-
Daft, Richard L. and Robert H. Lengel (1986), “Organizational istrative Science Quarterly, 21 (4), 571–97.
Information Requirements, Media Richness and Structural Kumar, V., J. Andrew Petersen, and Robert P. Leone (2007), “How
Design,” Management Science, 32 (5), 554–71. Valuable Is Word of Mouth?” Harvard Business Review, 85
Donkers, Bas, Philip Hans Franses, and Peter C. Verhoef (2003), (10), 139–46.
“Selective Sampling for Binary Choice Models,” Journal of ———, ———, and ——— (2010), “Driving Profitability by
Marketing Research, 40 (November), 492–97. Encouraging Customer Referrals: Who, When, and How,”
Fisher, Robert J. (1993), “Social Desirability Bias and the Validity Journal of Marketing, 74 (September), 1–17.
of Indirect Questioning,” Journal of Consumer Research, 20 Lancaster, Tony and Guido W. Imbens (1991), “Choice Based
(2), 303–315. Sampling: Inference and Optimality,” working paper, Depart-
Fombrun, Charles and Mark Shanley (1990), “What’s in a Name? ment of Economics, Brown University.
Reputation Building and Corporate Strategy,” The Academy of Lee, Lung-Fei (1983), “Generalized Econometric Models with
Management Journal, 33 (2), 233–58. Selectivity,” Econometrica, 51 (2), 507–512.
Gershoff, Andrew D., Susan M. Broniarczyk, and Patricia M. West Libai, Barak, Ruth Bolton, Marnix S. Bügel, Ko de Ruyter, Oliver
(2001), “Recommendation or Evaluation? Task Sensitivity in Götz, Hans Risselada, and Andrew T. Stephen (2010),
Information Source Selection,” Journal of Consumer Research, “Customer-to-Customer Interactions: Broadening the Scope of
28 (3), 418–38. Word of Mouth Research,” Journal of Service Research, 13
Godes, David (2008), “Customer Referencing: Three Programs,” (3), 267–82.
Harvard Business School Note 508-101. Maddala, G.S. (1983), Limited Dependent and Qualitative Variables
——— (2012), “The Strategic Impact of References in Business in Econometrics. New York: Cambridge University Press.
Markets,” Marketing Science, 31 (2), 257–76. McCracken, Grant (1989), “Who Is the Celebrity Endorser? Cul-
Goldenberg, Jacob, Sangman Han, Donald R. Lehmann, and Jae tural Foundations of the Endorsement Process,” Journal of
Weon Hong (2009), “The Role of Hubs in the Adoption Consumer Research, 16 (3), 310–21.
Process,” Journal of Marketing, 73 (March), 1–13. Morgan, Robert M. and Shelby D. Hunt (1994), “The Commitment–
Gulati, Ranjay and Martin Gargiulo (1999), “Where Do Inter- Trust Theory of Relationship Marketing,” Journal of Market-
organizational Networks Come From?” American Journal of ing, 58 (July), 20–38.
Sociology, 104 (5), 1439–93. Moriarty, Rowland T. and Robert E. Spekman (1984), “An
Hada, Mahima, Rajdeep Grewal, and Gary L. Lilien (2011), Empirical Investigation of the Information Sources Used Dur-
“Trade-Offs in Supplier-Initiated Referrals,” working paper, ing the Industrial Buying Process,” Journal of Marketing
The Mary Jean and Frank P. Smeal College of Business, Penn- Research, 21 (May), 137–47.
sylvania State University. Petersen, J. Andrew and V. Kumar (2009), “Are Product Returns a
Heide, Jan B. and George John (1990), “Alliances in Industrial Necessary Evil? Antecedents and Consequences,” Journal of
Purchasing: The Determinants of Joint Action in Buyer–Seller Marketing, 73 (May), 35–51.
Relationships,” Journal of Marketing Research, 27 (February), Prins, Remco and Peter C. Verhoef (2007), “Marketing Communi-
24–36. cation Drivers of Adoption Timing of a New E-Service Among
Helm, Sabrina and Risto T. Salminen (2010), “Basking in Existing Customers,” Journal of Marketing, 71 (April),
Reflected Glory: Using Customer Reference Relationships to 169–83.
Build Reputation in Industrial Markets,” Industrial Marketing Reinartz, Werner J., Jacquelyn S. Thomas, and V. Kumar (2005),
Management, 39 (5), 737–43. “Balancing Acquisition and Retention Resources to Maximize
Herr, Paul M., Frank R. Kardes, and John Kim (1991), “Effects of Customer Profitability,” Journal of Marketing, 69 (January),
Word-of-Mouth and Product-Attribute Information on Persua- 63–79.

Defining, Measuring, and Managing Business Reference Value / 85

Spence, Michael (1973), “Job Market Signaling,” The Quarterly Villanueva, Julian, Shijin Yoo, and Dominique M. Hanssens
Journal of Economics, 87 (3), 355–74. (2008), “The Impact of Marketing-Induced vs. Word-of-Mouth
Stephen, Andrew T. and Oliver Toubia (2010), “Deriving Value Customer Acquisition on Customer Equity Growth,” Journal
from Social Commerce Networks,” Journal of Marketing of Marketing Research, 45 (February), 48–59.
Research, 47 (April), 215–28. Wangenheim, Florian v. and Tomás Bayón (2007), “The Chain
Tversky, Amos and Daniel Kahneman (1974), “Judgment Under from Customer Satisfaction via Word-of-Mouth Referrals to
Uncertainty: Heuristics and Biases,” Science, 185 (4157), New Customer Acquisition,” Journal of the Academy of Mar-
1124–31. keting Science, 35 (2), 233–49.
Uzzi, Brian (1997), “Social Structure and Competition in Interfirm Webster, Frederick E. and Yoram Wind (1972), “A General Model
Networks: The Paradox of Embeddedness,” Administrative for Understanding Organizational Buying Behavior,” Journal
Science Quarterly, 42 (1), 35–67. of Marketing, 36 (April), 12–19.
Venkatesan, Rajkumar and V. Kumar (2004), “A Customer Life- Wood, Whitney (2007), Customer Reference Program Bench-
time Value Framework for Customer Selection and Resource marking Study, 2007 Edition. Sunnyvale, CA: The Phelon
Allocation Strategy,” Journal of Marketing, 68 (October), Group.
106–125. Wuyts, Stefan and Inge Geyskens (2005), “The Formation of
Buyer–Supplier Relationships: Detailed Contract Drafting and
Close Partner Selection,” Journal of Marketing, 69 (October),
103–117.

86 / Journal of Marketing, January 2013

 Defining, Measuring and Managing Business Reference Value

V. Kumar, J. Andrew Petersen, & Robert P. Leone

Web Appendix W1: Categories for Congruency Variables

We aggregated each of the three congruency variables (industry, product/services, and

functional area) into seven categories. This allowed us to have at least one reference that covered
each of the categories – even though we did not have a reference which covered each three-way
category combination. These categories included:

Industry:

(1) Education, (2) Government, (3) Wholesalers/Distributors, (4) Manufacturing, (5)

Computer Technology, (6) Service, and (7) Other

Product/Service:

Financial Services: (1) Business Checking/Savings, (2) Payroll Deposit, (3) Line of
Credit, (4) Loans, (5) Insurance, (6) Investments, and (7) Other

Telecommunications: (1) Landline Phone Service, (2) VOIP (Voice Over IP), (3) Internet
Service, (4) Wireless Service, (5) Maintenance Contracts, (6) Infrastructure, and (7)
Other

Role Functional Area:

Senior-level positions in the following departments within each firm: (1) Marketing, (2)
Finance, (3) Accounting, (4) Information Technology, (5) Human Resources, (6)
Procurement, and (7) Other
Web Appendix W2: Determining the Key Drivers of BRV – Comparison with Null Models

For each of the three null models we use the same selection equation variables. For the censored
regressions we use different set of variables. We estimate the model using the same algorithm as
described in Appendix A of the paper. Results are as follows.
Selection Equation:
Variable Telecommunications
Intercept 0.18 (0.049)
Client Firm Size
Employees 0.062 (0.021)
Revenue 0.00026 (0.0001)
Length of Client Relationship
CLV 0.102 (0.028)
(CLV)2 -0.00024 (0.0001)
Tenure 0.163 (0.056)

Censored Regression Equation:

Variable Null Model 1 Null Model 2 Null Model 3 Full Model
Intercept 2.91 (1.06) 2.89 (0.92) 2.91 (0.91) 3.14 (1.06)
Client Firm Size
Number of Employees 0.238 (0.091) 0.239 (0.088) 0.228 (0.085) 0.231 (0.082)
Annual Revenue 0.018 (0.0042) 0.013 (0.003) 0.010 (0.003) 0.0086 (0.003)
Length of Client Relationship
CLV 0.342 (0.041) 0.321 (0.033) 0.311 (0.032) 0.339 (0.052)
(CLV)2 -0.0066 (0.0015) -0.0062 (0.0013) -0.0057 (0.0012) -0.0072 (0.002)
Tenure 0.491 (0.092) 0.489 (0.091) 0.481 (0.090) 0.471 (0.091)
Reference Congruency
Industry 0.472 (0.081) 0.481 (0.077) 0.464 (0.087)
Product/Service 0.616 (0.132) 0.617 (0.131) 0.626 (0.144)
Role 0.276 (0.091) 0.271 (0.097) 0.291 (0.083)
Reference Media Format
Audio Testimonial 0.203 (0.031) 0.204 (0.033)
Video Testimonial 0.241 (0.031) 0.256 (0.042)
Written Testimonial 0.186 (0.049) 0.176 (0.057)
Case Study/White Paper 0.107 (0.031) 0.112 (0.038)
Interaction Effects
Firm Size x Relationship
0.004 (0.0016)
Length
Firm Size x Reference
0.0027 (0.0008)
Media Format
Firm Size x Congruency 0.0052 (0.0017)
Sample Selection Correction
Lamba* 0.42 (0.126) 0.41 (0.121) 0.39 (0.112) 0.42 (0.109)
Model Fit
R2 0.31 0.48 0.62 0.71
Log-Likelihood -228.91 -213.41 -192.62 -178.61
AIC 486.8 462.4 430.9 417.81
Copyright of Journal of Marketing is the property of American Marketing Association and its
content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.