“Selling in the Digital Age” © 2019 Michael Ahearne, Zachary Hall, Partha Krishnamurthy,
and Mohsen Pourmasoudi
MSI working papers are distributed for the benefit of MSI corporate and academic members
and the general public. Reports are not to be reproduced or published in any form or by any
means, electronic or mechanical, without written permission.
Marketing Science Institute Working Paper Series 2019 Report No. 19-123
Selling in the Digital Age
Michael Ahearne, Zachary Hall, Partha Krishnamurthy, and Mohsen Pourmasoudi
Selling in the Digital Age
Michael Ahearne, Zachary Hall, Partha Krishnamurthy, Mohsen Pourmasoudi
Michael Ahearne*
C.T. Bauer Professor of Marketing
Department of Marketing
C.T. Bauer College of Business
University of Houston
Phone: 713 743 4155
Address: MH 375D, 4750 Calhoun Road, Houston, Texas 77204
Email: [email protected]
&
Zachary Hall
Associate Professor of Marketing
Marketing Department
Neeley School of Business
Texas Christian University
Phone: 817 257 5068
Address: TOCB 102, Neeley School of Business, Texas Christian University, 2900 Lubbock, Fort Worth, Texas
76109
Email: [email protected]
&
Partha Krishnamurthy
Professor of Marketing
Department of Marketing
C.T. Bauer College of Business
University of Houston
Phone: 713 743 4576
Address: MH 385H, 4750 Calhoun Road, Houston, Texas 77204
Email: [email protected]
&
Mohsen Pourmasoudi
3rd Year PhD Candidate
Marketing Science Institute Working Paper Series 1
Department of Marketing
C.T. Bauer College of Business
University of Houston
Phone: (1) 713 743 4577
Address: MH 375L, 4750 Calhoun Road, Houston, Texas 77204
Email: [email protected]
*The corresponding author
Date of Submission: April 2019
Marketing Science Institute Working Paper Series 2
Selling in the Digital Age
Current sales tactics have lost their efficacy in closing a sale because of the mismatch between
the classic sales model and buyer expectations. According to HubSpot (2016), “it’s about time
that sales catches up with the consumer with a sales process that matches the buyer’s
preferences. If salespeople cater to what buyers want, sales won’t be so hard after all.” Advances
in technology have revolutionized consumer buying behavior. Notably, the Internet has provided
customers with a cheap source of information that can help them make informed buying
decisions. Many customers extensively research a product/service before they decide on what to
buy and spend extensive time on the Internet evaluating different sellers’ online content, peer
reviews, and reviews from other third-party sources. For example, an Ecommerce (2017) study
shows that, globally, 47% of customers check sellers’ websites and 55% of customers consult
online reviews before purchasing. These numbers are even larger in developed countries; for
example, 88% of U.S. consumers engage in online research before making a purchase
(Ecommerce 2017). Furthermore, this behavior is not limited to business-to-consumer sales, as
business-to-business (B2B) buyers also engage in the same type of research before contacting a
salesperson. For example, according to Accenture Interactive (2014), 94% of B2B buyers
conduct online research before engaging in the buying process.
However, while customers are searching for and making decisions about products/services
before meeting a salesperson, many salespeople are still relying on the classic model of selling,
which is based on the assumption that customers are uninformed, uncertain, and undecided when
they meet a salesperson. Building on this premise, the classic sales model suggests that
salespeople, as knowledgble gatekeepers of information, should guide customers through the
Marketing Science Institute Working Paper Series 3
sales process. Thus, current selling textbooks divide the role of a salesperson into four stages:
understanding customers’ needs by asking questions, creating a value proposition,
communicating the value proposition by educating customers and challenging their mindsets,
and, finally, delivering the value proposition (e.g., Manning, Reece, and Ahearne 2018;
Castleberry and Tanner 2018).
Similarly, best-selling practitioner guidebooks on selling divide the sales process into two
broad stages: (1) understanding customers’ needs and (2) educating and challenging customers’
mindsets. For example, under the assumption that customers reach out to a salesperson when
they have no solution at hand, “solution selling” suggests that the salesperson should diagnose
their needs and then recommend the right products/services to fulfill those needs (Eades 2004;
Bosworth 2002). Another example is the influential “challenger sales model,” which “these days,
almost every new hire in sales is told to read [about]” (HubSpot 2018). The challenger sales
model suggests that a seller should actively teach customers, tailor the sales process to them, and
take control of the customer conversation (Dixon and Adamson 2011). Given the assumption that
salespeople are better aware than customers themselves of the potential products/services that
satisfy their needs, Bryon (2018) asserts that “the best sales organizations today increase
business by challenging customers, delivering customer interactions specifically designed to
[both] disrupt their current thinking and teach them something new.”
However, considering the massive amount of information available to customers, the
assumption that customers are uniformed, uncertain, and undecided when meeting with a
salesperson no longer holds. This has resulted in an incongruity between the classic sales model
employed by salespeople and the expectations of customers. For example, a study shows that
80% of buyers contact a salesperson only after they have done their research and have narrowed
Marketing Science Institute Working Paper Series 4
down their consideration set, significantly; in other words, they have a fairly good idea of what
they want to buy (HubSpot 2016). Nevertheless, current literature suggests that salespeople
should challenge, push, and debate with customers (e.g., Dixon and Adamson 2011). Similarly,
while salespeople are trained to “educate” their customers (e.g., Dixon and Adamson 2011;
Sheth and Sharma 2008; Thaichon et al. 2018), a study from HubSpot (2016) reports that more
than 80% of B2B buyers self-educate about products and firms before ever contacting a
salesperson. These customers do not want to contact a salesperson prior to this point, and when
they do meet a salesperson, many feel confident in their knowledge of those products and firms
they are considering, and have very specific expectations. As a result of these types of
informational changes in the environment, nowadays, instead of salespeople looking for
customers, customers are looking for salespeople, but only after doing their due diligence.
The incongruity between customers’ expectations of salespeople and salespeople’s sales
tactics hurts both buyers and sellers. A well-established point in the literature is that adapting
sales tactics to different buyers is an important factor in sales performance (Spiro and Weitz
1990; Weitz 1981; Weitz, Sujan, and Sujan 1986). Furthermore, selling effectively requires that
salespeople make accurate judgments about their customers and adopt an appropriate sales
strategy (Weitz 1981). Thus, it is critical to assess how the classic sales model works in an
environment in which the customer may have greater or lesser preference certainty induced by
accessing information. Against this backdrop, we examine the consequences of the lack of
congruency between contemporary customers’ buying behavior and salespeople’s sales models.
In particular, in this article, we aim to answer the following question: How does employing the
classic sales model with customers with greater and lesser preference certainty affect the sales
interaction outcomes of (1) purchase probability, (2) sales revenue, and (3) customer satisfaction.
Marketing Science Institute Working Paper Series 5
To answer this question, we collected data from 356 individual sales interactions in 15
different stores of a U.S.-based retailer of durable goods over four months. This retailer provides
an ideal context to investigate our question because its selling context involves close
salesperson–customer interactions. Drawing on cognitive dissonance theory (Festinger 1962;
Harmon-Jones and Harmon-Jones 2012; Cooper 2011; Harmon-Jones and Mills 1999), we
hypothesize that using the classic sales model of questioning and challenging customers with
those who have greater preference certainty reduces their probability of purchasing; decreases
revenue from the sales interaction; and reduces customer satisfaction.
We contribute to both the practice and theory of sales in three ways. First, our study offers
direct instructions on how to adapt to the shift in customer buying behavior and also delineates
when the classic sales model is inefficient. We demonstrate that using the classic sales model on
a certain segment of customers reduces sales effectiveness, resulting in lower purchase
probability, sales revenue, and customer satisfaction. This use also hurts customers by reducing
their probability of purchasing, making them spend more time in the buying situation, and
lowering their satisfaction with their purchase. Second, although access to information has
caused a dramatic shift in the customer decision-making process, academic research has yet to
examine the impact of this shift on the practice of personal selling. Furthermore, adaptive selling
contends that adapting sales tactics to different buyers is an important factor for sales
performance (Weitz 1981; Weitz 1978). Thus, we extend the literature on adaptive selling by
showing the importance of a single variable—namely, customer preference certainty—in the
adoption of a suitable sales model. Third, we contribute to extant literature on cognitive
dissonance theory (Festinger 1962; Festinger 1964; Cooper 2011) by examining its predictions in
a sales interaction setting.
Marketing Science Institute Working Paper Series 6
We begin by briefly outlining how advances in technology have affected customers’
buying behavior. Then, we draw on the theoretical frameworks of decision conflict and cognitive
dissonance theory to explain our arguments for the hypothesized effects. Next, we introduce our
empirical setting and test our hypotheses. Finally, we discuss the results and explicate our
contributions.
Theoretical Background and Hypotheses
Shifts in Customer Buying Behavior
Before the introduction of the Internet and other digital technology, distribution of information
was asymmetric, with salespeople having access to more information than buyers. Specifically,
they had more information about the products/services, their features, competitive offerings, and
the comparative advantages and disadvantages, not to mention pricing information. Salespeople
played the role of gatekeepers of information, and they exercised this power of information
asymmetry to provide customers with information required for decision making, thus shaping
their preferences and setting prices accordingly.
The current information environment is no longer conducive to such information
asymmetry. Consumers can easily obtain information about product/service options and prices
and also access peer and third-party opinions on these options, all from the comfort of their own
couch. For example, the information environment for car buyers is now significantly richer, with
websites such as Kelley Blue Book, Edmunds.com, TrueCar, CarGurus, Autotrader, and Carfax
all providing customers with valuable information on makes, models, invoice prices, reliability
statistics, and so on.
Marketing Science Institute Working Paper Series 7
Not surprisingly, many consumers now turn to the Internet before engaging in any physical
shopping activity. A joint study by Google and Shopper Sciences, which covers a variety of
industries, reports that customers consult more than ten sources before they even begin to think
about purchasing (Lecinksi 2011). This ease of access also means that customers often begin
gathering information about their purchases months in advance.
Consumers who have access to information are likely to have more crystalized preferences
than those who do not have such access. In other words, as the Google and Shopper Sciences
study suggests, many customers are likely to be further along in their decision process, meaning
that they have more concrete preferences at the time they reach out to a salesperson. This
increased preference certainty appears to translate to decision making as well; 61% of shoppers
in the consumer goods industry and 97% of automobile shoppers make their decisions online, a
situation that Google calls the “zero moment of truth” (Lecinski 2011). This acceleration in
preference formation and decision making is not limited to the consumer decision context; for
example, a study of 1,500 business leaders involved in key purchases for 22 top B2B
organizations shows that buyers complete approximately 60% of the buying decision before they
ever contact a sales rep (CEB Marketing Leadership Council 2012).
Notwithstanding this shift in decision making induced by increased information access,
selling strategies have remained stagnant, premised on significant consumer–salesperson
information asymmetry and the assumption that consumers do not have well-formed preferences
before the interaction. As noted, the classic sales model suggests that salespeople should inquire
about customers’ problems, act as their information source and education agents, provide them
with a set of alternatives to choose from, and challenge them (Bosworth 2002; Castleberry and
Tanner 2018; Dixon and Adamson 2011; Eades 2004; Manning, Reece, and Ahearne 2018;
Marketing Science Institute Working Paper Series 8
Rackham, Kalomeer, and Rapkin 1988; Sheth and Sharma 2008); however, some customers may
have already passed these stages when they meet the salesperson. Therefore, the central question
we address herein is how preference certainty combines with the classic sales model to influence
key outcomes of the salesperson–customer interaction.
We argue that for customers with greater preference certainty, at the point of sale the aim
of the salesperson should be to close the deal and reinforce customer preferences rather than
attempt to educate or otherwise shape what those preferences should be. In the next section, we
draw on the theoretical frameworks of decision conflict and cognitive dissonance to develop
hypotheses on how the level of preference certainty of a customer before the salesperson meeting
interacts with the classic sales model to determine the likelihood of a sale, the amount of the sale,
and the customer’s satisfaction with the sales process. Our hypotheses are based on three
premises. First, as noted previously, consumers now have greater access to information about the
products/services they want to purchase. Second, some seek out and process the information
before the salesperson encounter. Third, the accessed information is often complex, having both
advantages and disadvantages, and consumers emerge from this process with varying degrees of
preference certainty.
In this discussion, we define “preference certainty” as the extent to which a customer is
clear and specific about the attributes he or she is looking for when shopping for a
product/service. For example, when shopping for an automobile, the buyer may have a clear
preference for the model, make, price, speed, options, and so on. Thus, a buyer with greater
preference certainty has clearly defined preference points along the attributes he or she considers
important. Separately, the classic sales models are broadly based on the notion that salespeople
should become education agents who work to shape customer preferences (Sheth and Sharma
Marketing Science Institute Working Paper Series 9
2008; Thaichon et al. 2018) regardless of the specific sales model used. For example, the
challenger sales model suggests that salespeople should offer insights to customers, educate
them, and challenge their preferences (Dixon and Adamson 2011). The customer-centric selling
model suggests that salespeople should become collaborative consultants to elicit and modify
customer preferences (Bosworth, Holland, and Visgatis 2004). Finally, the inbound selling
model suggests that salespeople should act as consultants and advise customers on what their
preferences should be (HubSpot 2017). Given these premises, we raise the question of how
greater or lesser preference certainty interacts with the classic sales models to influence sales
outcomes.
Preference Certainty, Decision Conflict, and Purchase Probability
Imagine a customer with greater preference certainty interacting with a salesperson. If the
salesperson proceeds to deploy the classic sales model, he or she will ask questions to discern
what the customer wants and why and then will try to shape the preferences to fit the preferred
product that he or she sees as more appropriate for the customer or would rather sell to the
customer. This method requires the salesperson to question the customer about his or her beliefs
about product/service attributes that he or she might value and other attributes he or she might
not be considering. When attribute dimensions and preferences thereof are challenged, the
decision maker experiences a high degree of decision conflict. This conflict can take several
forms, such as approach–approach conflict, if the challenge instigated by the salesperson makes
an ignored or discounted attribute appear equally attractive (e.g., the Samsung Galaxy gives
more value for the money, but the Apple iPhone is more durable both physically and
softwarewise), or approach–avoidance conflict, if the challenge highlights the unattractive
features of preferred attributes (e.g., a curved glass screen is nice but is also prone to cracking on
Marketing Science Institute Working Paper Series 10
the slightest impact, leading to costly repairs). Well known in the decision-making literature is
that decision conflict induces decision deferral (Iyengar and Lepper 2000; Dhar 1997). Thus, a
customer with greater preference certainty is likely to experience decision conflict when his or
her preferences are challenged by the salesperson. By contrast, a customer with lesser preference
certainty is likely to be more receptive to what the salesperson suggests. Furthermore, as decision
conflict precipitates decision deferral, customers with greater preference certainty might react to
the classic sales model by deferring the decision (i.e., reducing the probability of purchasing).
We can make an identical prediction when considering the classic sales model from the
perspective of cognitive dissonance. When customers with greater preference certainty interact
with salespeople who challenge their preferences, they now have an opposing opinion to keep in
mind as they navigate the decision process. This causes cognitive dissonance (Festinger 1962), a
negatively valenced state that motivates people to seek resolution. One way to reduce cognitive
dissonance is to adjust their preferences to resolve the conflict between their own preferences
and the salesperson’s challenge. Customers with greater preference certainty will likely find
doing so more difficult than those with lesser preference certainty. Another way to resolve
cognitive dissonance is to withdraw from the decision-making process to avoid it altogether.
Giving more information to or challenging the choice of a customer with greater preference
certainty can also result in choice procrastination. For example, Miller (1944) reports that
relinquishing an attractive option to obtain another leads to procrastination. By contrast, for
customers with lesser preference certainty, providing additional information or challenging their
preferences serves to help clarify their decisions and guide them to closure of the decision-
making process. Thus, building on the decision conflict and cognitive dissonance literature
Marketing Science Institute Working Paper Series 11
streams, we propose the following with regard to the probability of the sales interaction leading
to a sale:
H1: For customers with greater preference certainty, the probability of purchasing will
be lower when the sale technique is based on challenging customer preferences
than when the technique does not involve a challenge. This effect will be less
pronounced for customers with lesser preference certainty.
Effect on Revenue from the Sales Interaction
The aforementioned arguments should hold true for revenue as well because a lower purchase
probability means lower revenue from the sales interaction. Thus, we similarly hypothesize the
following:
H2: For customers with greater preference certainty, the revenue from the sale will be
lower when the sale technique is based on challenging customer preferences than
when the technique does not involve a challenge. This effect will be less
pronounced for customers with lesser preference certainty.
Effect on Customer Satisfaction from Choice
The other repercussion of the change in the customer decision process relevant to our study is
cognitive dissonance after the decision, which elicits less satisfaction. Related to this, Levin
(1951) and Festinger (1957) propose that choices among attractive but mutually exclusive
alternatives lead to conflicts that people try to avoid or eliminate. At this point, challenging
customers taxes their coping mechanism, leading to less satisfaction with their decision. For
example, in Iyengar and Lepper’s (2000) study, participants reported greater subsequent
satisfaction with their selections when their original set of options had been limited. In another
Marketing Science Institute Working Paper Series 12
study, Iyengar, Wells, and Schwartz (2006) find that people felt worse when they tried to look
for the “best” job among a selection of jobs. Thus:
H3: For customers with greater preference certainty, the satisfaction with the sales
process will be lower when the sales technique is based on challenging customer
preferences than when the technique does not involve a challenge. This effect will
be less pronounced for customers with lesser preference certainty.
Overview of Studies
To test the hypotheses, we conducted two studies. Study 1 is an observational field study that
examines salesperson–customer interactions at a durable goods retailer. Study 2 is a simulated
online-shopping study in which we experimentally vary the predictor variables.
Study 1: Field Study
Study Context
We conducted Study 1 in multiple showrooms of a midsize U.S.-based specialty durable goods
retailer; this retailer offered many different features, thereby providing an ideal context for
testing the hypotheses. First, the retailer carried multiple brands in each store ranging in price
from less than $1,000 up to around $5,000 at the time of the study, thus ensuring sufficient
variety in the product features, brands, and prices. Second, the products are sold primarily from
close salesperson–customer interactions, thus enabling us to characterize both the nature of the
customer preference before the interaction and the type of sales strategy employed during the
interaction, facilitating the testing of the hypotheses. Third, store salespeople are primarily
Marketing Science Institute Working Paper Series 13
incentivized through commission, motivating them to maximize their earning potential by
increasing customers’ purchase probabilities and sales revenue.
Sample and Data Collection
We collected the data in two phases. In the first phase, we carried out prestudy qualitative
interviews to better understand important attributes of the focal product. In the second phase, we
conducted a field study that involved assessing pre- and postinteraction customer surveys and
recording behavioral metrics regarding the salesperson–customer interaction. As part of the
preinteraction survey, we collected information on customers’ preference certainty, including
their research behavior before the survey.
Prestudy (phase 1). To identify the customer preference dimensions, we conducted a series
of qualitative, in-depth interviews. For an average of 30 minutes each, we interviewed more than
40 people, including potential customers, salespeople, store managers, and sales executives.
From these interviews, we created an initial list of attributes. Using this initial list, we
conducted another survey with more than 80 customers, to consolidate the list of customer
preference dimensions by removing redundancies and attributes deemed unimportant. Through
this process, we identified the attributes of price, brand, and attribute X (we blind the name of
this attribute for confidentiality reasons) as the most relevant and important product attributes to
this study. We eliminated other factors, such as financing and product return policy, in this
process because they do not vary across products within a store.
Field study (phase 2). The field study spanned four months at 15 store locations with the
help of two research assistants who each received approximately two weeks of training on how
to administer the survey. Both were blind to the hypotheses. For data collection, at different
times during each week, the interviewers intercepted both customers and salespeople and
Marketing Science Institute Working Paper Series 14
observed the sales interaction, collecting three distinct aspects for the final data set: customer
preinteraction survey, to assess preference certainty and gather demographic data; customer
postinteraction survey, to assess customer satisfaction; and observational data, which included
records of all the products shown to the customer during the sales interaction and whether he or
she purchased or did not purchase. Before completing the survey, customers were informed that
their individual responses were confidential.
Informed consent was obtained from each participant. We received 356 completed surveys.
More than half (52.8%) the customers in our sample purchased a product with an average retail
price of $1,431. The sales interaction took approximately 30 minutes on average.
To confirm that the study is not biased from the customers being intercepted by the
research assistants before the interaction, we gathered data on average purchase probability and
sales revenue from the company. The company indicated that, on average, 50% of store visitors
make a purchase, which is comparable to 52.8% in our sample. The company also reported that
the average purchase value is approximately $1,500, which is also in line with our sample
average of $1,431.
Method
Construct measures. From the customer preinteraction survey, we assessed customers’
preference certainty. We operationalized preference certainty by combining two measured
variables: 1) a dummy variable indicating whether a customer had done research on the products
before the sales interaction and 2) a Likert scale measuring the purchase certainty of the
researched items prior to the interaction from 0 to 10, with 0 being “no chance of purchasing
today” and 10 being “will definitely purchase today”, i.e., the highest level of certainty (coded as
1 if the level of certainty was in the top three boxes [high levels of certainty] and 0 otherwise).1
Marketing Science Institute Working Paper Series 15
For each sales interaction, we operationalized the salesperson’s extent of challenge by taking the
average of three variables: percentage of products shown to a customer that were outside his or
her intended budget, percentage of products that were outside his or her brand preference, and
the percentage of products that were outside his or her initial preference for attribute X.
Outcome measures. To determine the consequences of the classic sales model, we gathered
data by observing the sales interaction, and used objective company data. We analyzed the
effectiveness of the classic sales model with three different outcome measures. First, we
observed whether the customers made a purchase or not to calculate purchase probability.
Second, we captured the dollar amount the customers spent to calculate revenue; if no purchase
was made, we recorded $0 for the revenue. Third, we adapted the measure for customer
satisfaction from the literature to match our context and measured it using a seven-point Likert
scale.
We included customer priors as controls in our framework. We strategically chose to
capture controls that were either intuitively important or relevant to the dependent variables,
according to the literature. For example, several studies show that female customers report
higher satisfaction levels than male customers across all industries (e.g., Bryant and Cha 1996;
Mittal and Kamakura 2001). Bryant and Cha (1996) find that these gender differences in
satisfaction hold across all types of industries, including automobile, apparel, processed foods,
airlines, and restaurants. Therefore, to analyze customer satisfaction, we included gender as a
control. Similarly, to predict the revenue, we also controlled for how much the customer
expected to spend before the sales interaction (customer’s initial budget) and customer income.
We measured the customer’s initial budget by asking for the maximum amount he or she was
willing to spend on the focal product. We also controlled for a set of customer demographic
Marketing Science Institute Working Paper Series 16
variables, including age, marital status, ethnicity, and education, among others, in all models.
Table 1 depicts the correlation matrix for all covariates.
Results: Test of Hypotheses
As we have multiple observations of each salesperson interacting with several different
customers, we initially tested whether a multilevel regression approach was required. To check
whether the data set has a two-level structure, we ran all models with random intercepts for each
dependent variable and nested the salesperson–customer interactions at the salesperson level
(Singer 1998). For the dependent variable purchase probability, the likelihood ratio test of a
logistic random-intercept model versus a logistic regression was nonsignificant (p > .1). Thus, a
two-level hierarchical model was not supported. Similarly, for the dependent variables sales
revenue, and customer satisfaction, likelihood ratio tests of a random intercept model versus a
linear regression model were nonsignificant (both ps >.1), and two-level hierarchical models
were not supported.
Although we proved that fixed-effects and random-effects estimators were nonessential, to
alleviate potential endogeneity concerns arising from fixed characteristics of salespeople (e.g.,
ability), we report the results from both in Appendix 1. We also report the results of a linear
probability model of our dependent variable of purchase probability as a robustness check in
Appendix 1. Moreover, because we have multiple observations of each salesperson, it is highly
probable that the errors of the regression models are correlated within each salesperson.
Therefore, because the usual standard errors of such models are understated in the presence of
serial correlation, we used cluster-robust standard errors at the salesperson level for all our
estimators (Bertrand, Duflo, and Mullainathan 2004; Cameron and Miller 2015).
Marketing Science Institute Working Paper Series 17
Finally, we tested H1 by using binary logistic regression with clustered standard errors at
the salesperson level for the variable purchase probability, and we tested H2 and H3 by using
linear regressions with clustered standard errors at the salesperson level for the dependent
variables sales revenue, and customer satisfaction. Furthermore, as mentioned, we used a set of
controls at the customer level, including budget, income, age, number of children, marital status,
gender, ethnicity, and education. Table 2 reports the results of the analyses. In order to facilitate
interpretation, when applicable, all predictor variables are mean-centered.
In H1, we hypothesized a lower probability of purchasing for customers with greater
preference certainty when the sale technique is based on challenging customer preferences than
when the technique does not involve a challenge. We also hypothesized that this effect would be
less pronounced for customers with lesser preference certainty. The results from Table 2 show a
significant suppressing effect of the extent of challenge on the relationship between customers’
preference certainty and purchase probability, in support of H1 (Model 1: b = –1.636, p < .05; see
Figure 1, Panel A, following References).
H2 predicted a lower amount of revenue for customers with greater preference certainty
when the sale technique is based on challenging customer preferences than when the technique
does not involve a challenge. It also proposed that this effect would be less pronounced for
customers with lesser preference certainty. Model 2 in Table 2 shows a significant, negative
crossover effect of the extent of challenge on the relationship between customer preference
certainty and sales revenue, in support of H2 (Model 2: b = –1158.6, p < .001; ; see Figure 1,
Panel B, following References).
Finally, H3 predicted a lower level of satisfaction for customers with greater preference
certainty when the sale technique is based on challenging customer preferences than when the
Marketing Science Institute Working Paper Series 18
technique does not involve a challenge. It also proposed that this effect would be less
pronounced for customers with lesser preference certainty. As Model 3 in Table 2 shows, the
extent of challenge has a marginally significant and negative effect on the relationship between
customer preference certainty and customer satisfaction, in support of H3 (Model 3: b = –.69, p ≈
.06; see Figure 1, Panel C, following References).
Discussion
In Study 1, we found that customer preference certainty is an important moderating factor of the
effect of the extent of challenge in the classic sales model. First, we showed that customers with
greater preference certainty have a lower purchase probability when the classic sales model is
implemented. As the margins depicted in Panel A of Figure 1 show, for customers at higher
levels of preference certainty, implementing the classic sales model reduces the purchase
probability by 20 percentage points (i.e., from more than 80% to 60%). Second, the revenue from
a sales interaction is reduced when salespeople use the classic sales model on customers with
higher levels of preference certainty. As Panel B of Figure 1 depicts, by not using the classic
sales model, salespeople may be able to increase their revenue from a sales interaction by
approximately 30%. Third, we found a marginally significant effect of customer preference
certainty on the relationship between the challenger technique and customer satisfaction. As
Panel C of Figure 1 shows, when faced with customers who have a good idea of what they want
to buy, salespeople may decrease customer satisfaction by using the classic sales model. All
these results are robust to the choice of estimation method. As Appendix 1 shows, the results
obtained from both the fixed-effects and random-effects models are close to the results presented
here.
Marketing Science Institute Working Paper Series 19
Study 2: A Scenario-Based Experiment
The primary objective of Study 1 was to test the hypotheses that in a sales interaction, customer
preference certainty would have a moderating effect on the relationship between the classic sales
model and key sales interaction outcomes, including purchase probability, sales revenue, and
customer satisfaction. In Study, 2, our objectives were threefold. First, although we showed the
robustness of the findings of Study 1 to different estimation methodologies, in Study 2, we aimed
to corroborate these findings under controlled situations, to further establish the causal
relationship and to rule out any remaining potential endogeneity concerns. Second, although
Study 1 included a combination of attributes on which we based our variable “extent of
challenge,” in Study 2 we aimed to rule out the possibility of price being the main variable
guiding customers’ decisions. Third, we wanted to replicate the findings of Study 1 with a
different product in a different product context.
Method
Design. Study 2 was a 2 × 3 design that manipulated customer preference certainty (high vs. low)
and the extent of challenge (no challenge vs. challenge vs. challenge all attributes except for
price) between-subjects.
Procedure. We recruited 500 U.S.-based participants from an Amazon Mechanical Turk
panel for this study; all received $.40 as compensation for their participation. The study, which
contained three phases, involved a simulated e-bike shopping experience in which the participant
was acting as a shopping agent for another (anonymous) customer. We manipulated participants’
certainty of customer’s preference as either high or low. We adopted this participant-as-
shopping-agent approach to maximize our ability to manipulate preference certainty by isolating
Marketing Science Institute Working Paper Series 20
the participant’s own preference certainty, in line with prior research on preference certainty
(e.g., Syam, Krishnamurthy, and Hess 2008).
In Phase 1, we screened out those who were familiar with e-bikes to ensure that
participants had no prior knowledge about e-bikes. After this, we introduced e-bikes as the focal
product of the study to the participants and described the weight, range, top speed, and price as
the four most important attributes of an electric bike.
In Phase 2, we described the participants’ role as a shopping agent for another customer
and randomly assigned them to the high- or low-preference-certainty scenario. In the high-
preference-certainty scenario, we asked the participants to imagine that they have had several
meetings with the customer regarding what he or she is looking for in an e-bike in terms of the
weight, range, top speed, and price. We also told them to presume that they were “completely
sure” of the preferences of the customer from their extensive interactions. In the low-preference-
certainty scenario, we told the participants that they have had a brief interaction with the
customer and were led to believe that he or she had a range of preferences for the attributes of
weight, range, top speed, and price. Therefore, we told them to presume that they were “not even
slightly sure” about the true preferences of the customer.
Following this, participants saw a range of numbers for weight, range, top speed, and price
of an e-bike and were told that they arrived at these numbers from their discussion with the
customer. The presented numbers were identical in both conditions. Details of the scenarios used
are available in Appendix 2. Having induced different certainty levels, we proceeded to
administer the manipulation check to assess whether the two groups varied on the level of
preference certainty as intended.
Marketing Science Institute Working Paper Series 21
In Phase 3, all participants were told that they would be directed to an online showroom
where a salesperson would present them with e-bikes containing information along the
dimensions of weight, range, top speed, and price. They were also informed that after seeing the
e-bikes, they could (1) purchase that e-bike for the customer; (2) reconfigure an e-bike him- or
herself along the attributes of weight, range, and top speed, with price determined by the levels
of these attributes (we did this to make the experiment more realistic, as in a real sales
interaction, customers have the option to see exactly what they want to purchase); or (3) leave
the showroom and go to another store.
Next, our manipulation involved one of three levels of the extent of challenge, as noted
previously. In the no-challenge condition, the salesperson presented an e-bike with attribute
levels in line with what the participants had seen in Phase 1. In the challenge condition, the
salesperson presented the participants with an e-bike with attribute levels that differed from what
they had seen in Phase 1. In the third condition, challenge all attributes except for price, the price
was kept in line with what the participants had seen in Phase 1, but other attribute levels were
altered. We included this condition to ensure the interaction effect is not solely related to price
by having a nonprice challenge as well. Details of the study procedure, manipulation checks, and
descriptive statistics are available in Appendix 2.
Coding of dependent and independent variables. We measured our dependent variable
purchase probability as a binary response, coded as 0 if the participant chose to leave the store at
any stage and 1 otherwise (whether he or she bought the electric bike that was presented or the
one he or she configured without leaving the store).
We coded the independent variable of preference certainty as 1 for high preference
certainty and 0 for low preference certainty. The independent variable extent of challenge had a
Marketing Science Institute Working Paper Series 22
value of 0 for the no-challenge condition, a value of 1 for the challenge condition, and a value of
2 for the challenge-all-attributes-except-for-price condition.
Results
Manipulation check. To check the effectiveness of the preference certainty manipulation, we
adapted a scale for measuring certainty from the literature (Haas and Kenning 2014; Jain and
Srinivasan 1990) to our context. The results indicated a significant effect of the certainty
condition on the manipulation check for preference certainty (Mcertain = 5.54, SD = 1.14 vs.
Muncertain = 3.94, SD = 1.64, F (1,499) = 159.44, p<.001).
Purchase probability. We analyzed the data collected in this experiment using a logistic
regression model, with the main effects of preference certainty and challenge and their
interaction effect. Table 3 reports the results of the analysis. As predicted, the coefficient of the
interaction of certainty with challenge was significant (b = –1.49, p < .05). In addition, the
coefficient of the interaction of certainty with challenge all attributes except for price was
significant (b = –1.51, p < .05).
Discussion
Study 2 had three main objectives. First, it aimed to corroborate the findings of Study 1 under
more controlled situations, to further establish the causal relationship. The results of Study 2
show that under high levels of preference certainty, challenging the customer results in a lower
likelihood to purchase during the shopping visit. Therefore, preference certainty has a negative
effect on the relationship between the extent of challenge and purchase probability. Second,
Study 2 aimed to check whether the interaction effect is related to price by having a nonprice
challenge as well; the findings indicate that these effects are unrelated to price. The interaction of
certainty with the challenge-all-attributes-except-for-price condition was significant; thus, even
Marketing Science Institute Working Paper Series 23
in absence of a challenge on price, preference certainty had a negative effect on the relationship
between the extent of challenge and purchase probability. Third, we aimed to examine and
replicate the findings from Study 1 in a different product context, and we found consistent results
with Study 1.
General Discussion
Theoretical Contributions
Weitz, Sujan, and Sujan (1986, p. 175) define the practice of adaptive selling as “the altering of
sales behaviors during a customer interaction or across customer interactions based on perceived
information about the nature of the selling situation.” The literature expanding on that study
suggests that rather than using the same tactics, salespeople need to adapt their tactics so that
they suit the buyers with whom they are dealing (e.g., Szymanski 1988; McFarland, Challagalla,
and Shervani 2006). By identifying a shift in the customer decision-making process, we
investigated the consequences of the classic sales model and how it interacts with customers’
decision-making processes. In particular, through an extensive field study and an experiment, we
show that customer preference certainty has a negative impact on the relationship between
employing the classic sales model and three sales interaction outcomes—purchase probability,
sales revenue, and customer satisfaction.
Furthermore, although research has found that the customer decision-making process is
related to a salesperson’s performance (e.g., Weitz 1978), to our knowledge, we are the first to
examine the shift in customer decision making and its relationship to the classic sales model.
Finally, this research contributes to the theory of cognitive dissonance in a sales context by
identifying the interaction of the customer decision-making process and the classic sales model.
Marketing Science Institute Working Paper Series 24
Managerial Contributions
Despite developments in technology and the Internet, personal selling continues to be an
important part of the sales/marketing spend both in complex retail products and B-to-B selling.
Conservative estimates put the amount spent on sales forces in the United States upward of $800
billion (Steenburgh and Ahearne 2012), a staggering number that is greater than the combined
gross domestic product of more than 100 countries.
However, advances in technology have altered the way buyers interact with sellers.
Massive amount of publicly available online information has resulted in buyers who often
involve in self-educating and extensive research before meeting a salesperson. As a result, the
buyer-salesperson interaction has changed, and research on an ideal sales process in face of this
change is needed (HubSpot 2016; CEB Marketing Leadership Council 2012; Accenture
Interactive 2014).
In this article, we showed that the use of the classic sales model, with customers who have
involved in online research and are certain about their purchase decision, hurts the three
managerially important outcomes of purchase probability, revenue, and customer satisfaction.
Nevertheless, in our context, we found that salespeople use a blanket tactic for all customers. In
particular, more than 80% of the products shown to customers were outside their budget, and
approximately 52% of products shown cost more than the maximum amount the customers were
willing to pay. Furthermore, more than 50% of the products shown to the customers deviated
from their brand preferences, and 40% differed from their preferences for another important
product feature. Thus, there is a marked loss of efficiency in sales interactions, which hurts both
selling firms and their customers.
Marketing Science Institute Working Paper Series 25
We offer three ways that managers can benefit from this research, and help their
salespeople move beyond the classic sales model.
Training. In light of the financial significance of salespeople, U.S. companies invest
approximately $2,000 each year on sales force training programs per salesperson (Ingram et al.
2015). However, about half of sales leaders still identify recruiting and training their salespeople
as their most pressing challenges (InsideSales 2016). We argue that training salespeople can be
more efficient if sales tactics take the shift in the customer decision-making process into
consideration. Specifically, we recommend that sales managers teach their salespeople about the
importance of acknowledging the different position of customers along the decision-making
process and train them on the appropriate tactic in each stage.
We found that salespeople switch between a low and high extent of challenge. This means
that salespeople are able to employ both tactic levels, but the inefficiency comes from a lack of
knowledge about when each is appropriate. This is especially important because managers can
train their salespeople to discern when each tactic is appropriate. Managers should also teach
salespeople efficient ways to infer customer preference certainty to enable them to guide
customers through the buying process.
Lead generation. We found that customers are at different stages of preference certainty
when they meet salespeople. Customer relationship management technology now makes it
possible to monitor customers’ research behavior and to index prospects depending on their stage
in the process; therefore, this technology can assist salespeople in adopting the appropriate tactic.
Managers can use the customer relationship management potential to help salespeople navigate
the sales interaction process in a more efficient way. Moreover, insights from the marketing
department can be more valuable when they help salespeople infer the level of customers’
Marketing Science Institute Working Paper Series 26
preference certainty, in turn potentially facilitating the exchange of information between the two
areas.
Development of new sales models. The Internet has revolutionized customer buying
behavior, enabling customers to often know what they want at the time of purchase decision.
Consequently, classic sales models based on asymmetry of information between customers and
salespeople have become inefficient. Salespeople can infer the level of preference certainty of
each customer and tailor their strategy accordingly. Furthermore, we argue that new sales models
that revisit and redefine assumptions of the standard paradigm of selling should be developed.
Limitations and Future Research Directions
This study is not without limitations. The company in Study 1 allowed us to interview customers
before and after the interaction with a salesperson and to observe them during the sales
interaction. However, working with a different company or another product could lead to
different effect sizes. We established the generalizability of our findings with the help of Study
2, which we conducted using another product in a different category. Nevertheless, despite our
use of different contexts, further research that examines the same concepts in other industries is
necessary to expand the generalizability of our results.
Further research could extend our study in several ways. First, research could undertake a
complete audit of customers’ research behavior before the salesperson interaction and assess
their preference certainty for each attribute. Doing so could shed light on how research behavior
and sales technique interact to shape customer decision. Second, research could examine similar
effects in an online setting in which there is no salesperson involved, as websites often show a
set of purchase items to visitors. Some complications might arise in such studies however,
because firms need to determine the customer’s stage of decision making. However, such a study
Marketing Science Institute Working Paper Series 27
could provide firms with a fruitful understanding of how to tailor their offers in line with where
the customer is in the decision-making process. Third, further research could identify ways to
infer customer certainty from the beginning of the sales interaction.
Finally, in this study, we investigated the negative consequences of using the classic sales
model. In turn, research could develop a sales model based on affirmation rather than challenge
to determine how it would perform with regard to the shift that has occurred in the customer
decision-making process.
Marketing Science Institute Working Paper Series 28
APPENDIX 1
Although we proved that fixed-effects and random-effects estimators were nonessential, to
alleviate all endogeneity concerns arising from fixed characteristics of salespeople (e.g., ability),
we report the results from both models in here. First, to choose either fixed-effects or random-
effects estimators, we tested both models with sales interactions nested at the salesperson level.
To test fixed-effects versus random-effects models for each dependent variable, we used the
artificial regression approach of Arellano (1993), and Wooldridge (2002). This method
reestimates a random-effects equation with additional regressors transformed into deviations
from the mean. A rejection of this test means that the fixed-effects model is more appropriate.
We tested all dependent variables with this method using random intercept linear regressions. All
tests were nonsignificant (ps > .4). Nevertheless, we report the results from both estimation
methods here.
Moreover, because the usual standard errors of fixed-effects and random-effects models are
drastically understated in the presence of serial correlation, we used cluster-robust standard
errors at the salesperson level for all our estimators (Bertrand, Esther, and Sendhil 2004;
Cameron and Miller, 2015). As such, we tested H1 by using binary logistic random-intercept
model with clustered standard errors at the salesperson level for the variable of purchase
probability, and we tested H2 and H3 using linear random-intercept regressions with clustered
standard errors at the salesperson level for the dependent variables of revenue, and customer
satisfaction. The results are in Table A1.
Then, we tested H1 by using a binary logistic fixed-effects model with clustered standard
errors at the salesperson level for the variable of purchase probability, and we tested H2 and H3
Marketing Science Institute Working Paper Series 29
using linear fixed-effects regressions with clustered standard errors at the salesperson level for
the dependent variables of revenue, and customer satisfaction. The results are in Table A2.
Marketing Science Institute Working Paper Series 30
31
TABLE A1
Results of the Random-Effects Analyses
DV: Purchase
Probability
DV: Revenue DV: Customer
Satisfaction
Independent Variables
Customer certainty .787*** (.000) 166.6* (.035) -.0461 (.494)
Extent of challenge -1.646* (.039) -113.3 (.747) -.599** (.008)
Customer certainty
× extent of challenge
-1.633* (.025) -1158.6*** (.000) -.693 (.055)
Customer budget -.000388*** (.000) .167** (.004) .0000855 (.096)
Customer income .227** (.001) 121.5*** (.000) -.0224 (.440)
Customer age .0197 (.057) 7.134 (.175) -.00205 (.625)
Time with products .0149 (.159) 9.635 (.168) -.00323 (.624)
No. of children .345** (.007) 115.0* (.016) .0193 (.723)
Marital status (married =1) -.931* (.013) -510.1** (.003) -.0692 (.632)
Customer gender (female = 1) .147 (.603) 42.30 (.730) .420*** (.000)
Ethnicity (white = 1) -.170 (.454) -41.27 (.678) .0640 (.458)
Customer education level -.0585 (.583) -37.41 (.494) -.0246 (.559)
_Cons .459 (.173) 979.0*** (.000) 6.013*** (.000)
Log of the variance
_Cons -4.357
(.452)
N 356 356 356
adj. R2 Notes: p-values are in parentheses. * p < .05. ** p < .01. *** p < .001.
Marketing Science Institute Working Paper Series 31
32
TABLE A2
Results of the Fixed-Effects Analyses
DV: Purchase
Probability
DV: Revenue DV: Customer
Satisfaction
Independent Variables
Customer certainty .876*** (.000)
193.9* (.023)
-.0314 (.661)
Extent of challenge -1.625 (.101)
-8.703 (.981)
-.405 (.104)
Customer certainty
× extent of challenge
-1.397 (.080)
-1036.1*** (.001)
-.675 (.056)
Customer budget -.000389** (.001)
.166** (.006)
.0000511 (.324)
Customer income .255*** (.001)
141.4*** (.000)
.00775 (.784)
Customer age .0172 (.167)
4.950 (.350)
-.00441 (.298)
Time with products .0139 (.400)
10.61 (.193)
-.00146 (.835)
No. of children .285 (.060)
117.6* (.033)
.000425 (.994)
Marital status (married =1) -.797 (.086)
-469.7* (.014)
-.0831 (.581)pp
Customer gender (female=1) .153 (.644)
6.491 (.961)
.366** (.001)
Ethnicity (white=1) -.182 (.474)
-54.33 (.638)
.0975 (.322)
Customer education level -.0386 (.767)
-27.32 (.675)
-.0177 (.710)
_cons
980.6*** (.000)
6.028*** (.000)
N 338 356 356
adj. R2 .177 .031 Notes: p-values are in parentheses. * p < .05. ** p < .01. *** p < .001.
Marketing Science Institute Working Paper Series 32
33
APPENDIX 2
Study 2
Screening Criteria
Amazon Mechanical Turk (MTurk) samples can help researchers gather a quality and diverse
sample (Goodman and Paolacci 2017). However, participant misrepresentation in online
experiments is an issue of concern for some researchers (Sharpe Wessling, Huber, and Netzer
2017). Therefore, to obtain quality participants, we followed guidelines in recent literature
(Sharpe Wessling, Huber, and Netzer 2017; Goodman and Paolacci 2017) and implemented the
following strategies:
1. We limited our participant pool to the American panel to ensure that participants were
able to read and understand written English.
2. We also limited our pool to participants with an average rating of 95% or above on
MTurk.
3. We screened out anyone who was familiar with e-bikes. We asked participants two
questions to infer their familiarity. First, we asked them whether they or anyone in their
family has an e-bike. Second, we asked them whether they had done any prior research
on e-bikes. We screened out those who answered “yes” to either question.
4. We placed several attention questions in the experiment and screened out those who gave
the wrong answer to any of those questions.
Manipulation Check
We adapted two questions from Haas and Kenning (2014) .
“With regards to picking the right e-bike for the Nicholson family, I feel that (Likert scale)
…”
Marketing Science Institute Working Paper Series 33
34
1. I am certain what my choice should be.
2. I am confident that I will be able to pick the right e-bike for the Nicholson family.
Scenarios
Read the following very carefully. You will be asked questions based on this material:
Weight, range, top speed, and price are the most important factors for customers when
purchasing an electric bicycle (e-bike). Range is the maximum distance one can ride an e-
bike without a need for recharge, and top speed is the maximum speed of an e-bike. As range
and speed increase, e-bikes are considered better but get more expensive. Weight is an e-
bike's weight, and lighter e-bikes are better but more expensive.
Low certainty. Now, imagine that you want to purchase an e-bike on behalf of the Nicholson
family. You have had one very brief meeting with the Nicholson family and in that meeting,
they explained very briefly what they are looking for in an e-bike. After this short
conversation, you are not even slightly sure about their preferences. You guess that an e-bike
with the following features would suit their preferences.
High certainty. Now, imagine that you want to purchase an e-bike on behalf of the Nicholson
family. You have had several meetings with the Nicholson family and in those meetings, they
have explained in detail what they are looking for in an e-bike. After those long
conversations, you are completely sure about their preferences. You conclude that an e-bike
with the following features would best suit their preferences.
Marketing Science Institute Working Paper Series 34
35
References
Accenture Interactive (2014), "2014 State of B2B Procurement Study: Uncovering the Shifting
Landscape in B2B Commerce " (accessed 7 Oct, 2018), [available at
https://www.accenture.com/t20150624T211502__w__/us-en/_acnmedia/Accenture/Conversion-
Assets/DotCom/Documents/Global/PDF/Industries_15/Accenture-B2B-Procurement-Study.pdf].
Arellano, Manuel (1993), "On the testing of correlated effects with panel data," Journal of
econometrics, 59 (1-2), 87-97.
Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan (2004), "How much should we trust
differences-in-differences estimates?," The Quarterly journal of economics, 119 (1), 249-75.
Bosworth, Michael T (2002), Solution selling: Creating buyers in difficult selling markets:
McGraw-Hill, Inc.
Bosworth, Michael T, John R Holland, and Frank Visgatis (2004), CustomerCentric selling:
McGraw-Hill New York.
Bryant, Barbara Everitt and Jaesung Cha (1996), "Crossing the threshold," Marketing Research,
8 (4), 20.
Bryon, Jordan (2018), "The Power of the Challenger Sales Model," [available at
https://www.gartner.com/smarterwithgartner/power-challenger-sales-model/].
Cameron, A Colin and Douglas L Miller (2015), "A practitioner’s guide to cluster-robust
inference," Journal of Human Resources, 50 (2), 317-72.
Castleberry, Stephen Bryon and John F Tanner (2018), "Selling: building partnerships."
CEB Marketing Leadership Council (2012), "The Digital evolution in B2B Marketing,"
[available at https://www.cebglobal.com/content/dam/cebglobal/us/EN/best-practices-decision-
support/marketing-communications/pdfs/CEB-Mktg-B2B-Digital-Evolution.pdf].
Cooper, Joel (2011), "Cognitive dissonance theory," Handbook of theories of social psychology,
1, 377-98.
Dhar, Ravi (1997), "Consumer preference for a no-choice option," Journal of consumer
research, 24 (2), 215-31.
Marketing Science Institute Working Paper Series 35
36
Dixon, Matthew and Brent Adamson (2011), The challenger sale: Taking control of the customer
conversation: Penguin.
Eades, Keith M (2004), The new solution selling: The revolutionary sales process that is
changing the way people sell: McGraw-Hill New York, NY.
Ecommerce (2017), "United States B2C Ecommerce Country Report 2017," [available at
https://www.ecommercewiki.org/reports/541/united-states-b2c-ecommerce-country-report-2017-
free].
Festinger, Leon (1964), "Conflict, decision, and dissonance."
---- (1962), A theory of cognitive dissonance: Stanford university press.
Goodman, Joseph K and Gabriele Paolacci (2017), "Crowdsourcing consumer research," Journal
of Consumer Research, 44 (1), 196-210.
Haas, Alexander and Peter Kenning (2014), "Utilitarian and hedonic motivators of shoppers’
decision to consult with salespeople," Journal of retailing, 90 (3), 428-41.
a
Haas, Alexander and Peter Kenning (2014), "Utilitarian and hedonic motivators of shoppers’
decision to consult with salespeople," Journal of retailing, 90 (3), 428-41.
Harmon-Jones, Eddie and Cindy Harmon-Jones (2012), "Cognitive dissonance theory,"
Handbook of motivation science, 71-83.
Harmon-Jones, Eddie and Judson Mills (1999), "An introduction to cognitive dissonance theory
and an overview of current perspectives on the theory."
HubSpot (2018), "A 5-Minute Summary Of "The Challenger Sale" Book Your Boss Told You
To Read," [available at https://blog.hubspot.com/sales/challenger-sale-summary].
---- (2016), "Buyers Speak Out: How Sales Needs To Evolve," [available at
https://research.hubspot.com/buyers-speak-out-how-sales-needs-to-evolve].
Marketing Science Institute Working Paper Series 36
37
---- (2017), "The Inbound Sales Methodology," 2018), [available at
https://blog.hubspot.com/sales/inbound-sales-methodology].
Ingram, Thomas N, Raymond W LaForge, Michael R Williams, and Charles H Schwepker Jr
(2015), Sales management: Analysis and decision making: Routledge.
InsideSales (2016), "Inside Sales Top Challenges: New Research," 2018), [available at
https://blog.insidesales.com/research/inside-sales-challenges-research-2016/].
Iyengar, Sheena S and Mark R Lepper (2000), "When choice is demotivating: Can one desire too
much of a good thing?," Journal of personality and social psychology, 79 (6), 995.
Jain, Kapil and Narasimhan Srinivasan (1990), "An empirical assessment of multiple
operationalizations of involvement," ACR North American Advances.
Lecinski, Jim (2011), Winning the Zero Moment of Truth: ZMOT: Zero Moment of Truth.
Manning, Gerald L, Barry L Reece, and Michael Ahearne (2018), Selling today: Partnering to
create value: Pearson Education.
McFarland, Richard G, Goutam N Challagalla, and Tasadduq A Shervani (2006), "Influence
tactics for effective adaptive selling," Journal of Marketing, 70 (4), 103-17.
Mittal, Vikas and Wagner A Kamakura (2001), "Satisfaction, repurchase intent, and repurchase
behavior: Investigating the moderating effect of customer characteristics," Journal of marketing
research, 38 (1), 131-42.
Sharpe Wessling, Kathryn, Joel Huber, and Oded Netzer (2017), "MTurk character
misrepresentation: Assessment and solutions," Journal of Consumer Research, 44 (1), 211-30.
Sheth, Jagdish N and Arun Sharma (2008), "The impact of the product to service shift in
industrial markets and the evolution of the sales organization," Industrial Marketing
Management, 37 (3), 260-69.
Singer, Judith D (1998), "Using SAS PROC MIXED to fit multilevel models, hierarchical
models, and individual growth models," Journal of educational and behavioral statistics, 23 (4),
323-55.
Marketing Science Institute Working Paper Series 37
38
Spiro, Rosann L and Barton A Weitz (1990), "Adaptive selling: Conceptualization,
measurement, and nomological validity," Journal of marketing Research, 61-69.
Steenburgh, Thomas and Michael Ahearne (2012), "Motivating salespeople: what really works,"
Harvard Business Review, 90 (7-8), 70-5, 160.
Syam, Niladri, Partha Krishnamurthy, and James D Hess (2008), "That's what I thought I
wanted? Miswanting and regret for a standard good in a mass-customized world," Marketing
Science, 27 (3), 379-97.
Szymanski, David M (1988), "Determinants of selling effectiveness: the importance of
declarative knowledge to the personal selling concept," The Journal of Marketing, 64-77.
Thaichon, Park, Jiraporn Surachartkumtonkun, Sara Quach, Scott Weaven, and Robert W
Palmatier (2018), "Hybrid sales structures in the age of e-commerce," Journal of Personal
Selling & Sales Management, 1-26.
Weitz, Barton A (1981), "Effectiveness in sales interactions: a contingency framework," The
Journal of marketing, 85-103.
---- (1978), "Relationship between salesperson performance and understanding of customer
decision making," Journal of Marketing Research, 501-16.
Weitz, Barton A, Harish Sujan, and Mita Sujan (1986), "Knowledge, motivation, and adaptive
behavior: A framework for improving selling effectiveness," The Journal of marketing, 174-91.
Wooldridge, Jeffrey (2002), "Econometric Analysis of Cross Section and Panel Data," The MIT
Press 0.
Marketing Science Institute Working Paper Series 38
39
Footnote
1 Our results are robust to the choice of the top box and the top two boxes as well.
Marketing Science Institute Working Paper Series 39
40
TABLE 1
Study 1: Intercorrelation Matrix for Covariates
Certainty Extent of
Challenge
Budget Income Age Time with
Products
No. of
children
Marital
Status
Gender Ethnicity Education
Certainty 1
Extent of challenge -.0114 1
Budget -.00873 -.222*** 1
Income .144** -.101 .336*** 1
Age -.0248 .0497 .0659 .175*** 1
Time with products -.0116 .0504 .252*** .140** .0816 1
No. of children .0713 .0189 .0136 .236*** .0424 -.0371 1
Marital status .0325 .0217 .0450 .336*** .297*** .0320 .466*** 1
Gender .0513 .0469 .0174 -.0680 -.00263 .00566 .0643 .0390 1
Ethnicity .0892 -.0729 .0661 .197*** .0734 .0563 -.0915 -.0314 -.0704 1
Education .0414 -.0836 .176*** .350*** .0756 .164** .0160 .0417 -.122* .0594 1
N 356 Notes: t-statistics are in parentheses. * p < .05. ** p < .01. *** p < .001.
Marketing Science Institute Working Paper Series 40
41
TABLE 2
Study 1: Results of the Regression Analyses
DV: Purchase
Probability
DV: Revenue DV: Customer
Satisfaction
Independent Variables
Customer certainty .786*** (.000) 166.6* (.042) -.0461 (.498)
Extent of challenge -1.647* (.037) -113.3 (.749) -.599* (.011)
Customer certainty
× extent of challenge
-1.636* (.023) -1158.6*** (.000) -.693 (.061)
Customer budget -.000387*** (.000) .167** (.006) .0000855 (.103)
Customer income .226** (.002) 121.5*** (.001) -.0224 (.445)
Customer age .0197 (.056) 7.134 (.182) -.00205 (.627)
Time with products .0149 (.156) 9.635 (.175) -.00323 (.626)
No. of children .344** (.007) 115.0* (.020) .0193 (.724)
Marital status (married =1) -.928* (.014) -510.1** (.006) -.0692 (.634)
Customer gender (female = 1) .148 (.600) 42.30 (.731) .420*** (.000)
Ethnicity (white = 1) -.172 (.448) -41.27 (.680) .0640 (.462)
Customer education level -.0607 (.545) -37.41 (.498) -.0246 (.562)
_Cons .459 (.173) 979.0*** (.000) 6.013*** (.000)
N 356 356 356
adj. R2 .174 .057 Notes: p-values are in parentheses. * p < .05. ** p < .01. *** p < .001.
Marketing Science Institute Working Paper Series 41
42
TABLE 3
Study 2: Results of the Logistic Regression Analysis
DV: Purchase Probability
Challenge -.908* (.010)
Challenge all attributes except for
price
-.302 (.410)
High certainty 1.309** (.009)
High certainty × challenge -1.499* (.012)
High certainty × challenge all
attributes except for price
-1.515* (.013)
Constant 1.294*** (.000)
N 500
adj. R2
Notes: p-values are in parentheses. * p < .05. ** p < .01. *** p < .001.
Marketing Science Institute Working Paper Series 42