An Examination of Social Influence on Shopper Behavior Using Video Tracking Data

Abstract

This research investigates how the social elements of a retail store visit affect shoppers' product interaction and purchase likelihood. The research uses a bivariate model of the shopping process, implemented in a hierarchical Bayes framework, which models the customer and contextual factors driving product touch and purchase simultaneously. A unique video tracking database captures each shopper's path and activities during the store visit. The findings reveal that interactive social influences (e.g., salesperson contact, shopper conversations) tend to slow the shopper down, encourage a longer store visit, and increase product interaction and purchase. When shoppers are part of a larger group, they are influenced more by discussions with companions and less by third parties. Stores with customers present encourage product interaction up to a point, beyond which the density of shoppers interferes with the shopping process. The effects of social influence vary by the salesperson's demographic similarity to the shopper and the type of product category being shopped. Several behavioral cues signal when shoppers are in a potentially high need state and may be good sales prospects.

Full text

24
Journal of Marketing
Vol. 78 (September 2014), 24 –41
© 2014, American Marketing Association
ISSN: 0022-2429 (print), 1547-7185 (electronic)
Xiaoling Zhang, Shibo Li, Raymond R. Burke, & Alex Leykin
An Examination of Social Influence
on Shopper Behavior Using Video
Tracking Data
This research investigates how the social elements of a retail store visit affect shoppers’ product interaction and
purchase likelihood. The research uses a bivariate model of the shopping process, implemented in a hierarchical
Bayes framework, which models the customer and contextual factors driving product touch and purchase
simultaneously. A unique video tracking database captures each shopper’s path and activities during the store visit.
The findings reveal that interactive social influences (e.g., salesperson contact, shopper conversations) tend to
slow the shopper down, encourage a longer store visit, and increase product interaction and purchase. When
shoppers are part of a larger group, they are influenced more by discussions with companions and less by third
parties. Stores with customers present encourage product interaction up to a point, beyond which the density of
shoppers interferes with the shopping process. The effects of social influence vary by the salesperson’s
demographic similarity to the shopper and the type of product category being shopped. Several behavioral cues
signal when shoppers are in a potentially high need state and may be good sales prospects.
Keywords: social influence, video tracking, shopper marketing, path analysis, hierarchical Bayes model
Online Supplement: http://dx.doi.org/10.1509/jm.12.0106
Xiaoling Zhang is Assistant Professor of Marketing, Nanyang Business
School, and Research Fellow, Institute on Asian Consumer Insight,
Nanyang Technological University (e-mail: XLZhang@ntu.edu.sg). Shibo
Li is Associate Professor of Marketing and Weimer Faculty Fellow (e-mail:
shili@ indiana. edu), and Raymond R. Burke is the E.W. Kelley Professor of
Business Administration and Director of the Customer Interface Labora-
tory (e-mail: rayburke@indiana.edu), Kelley School of Business, Indiana
University. Alex Leykin is a Research Associate, Kelley School of Busi-
ness, and Adjunct Research Scientist, School of Informatics, Indiana Uni-
versity (e-mail: oleykin@ indiana.edu). V. Kumar served as area editor for
this article.
Social influence plays an important role in the retail
shopping process. It can affect the time shoppers
spend in the store, their attitudes toward the merchan-
dise, and the specific products they pick up, try on, and pur-
chase (Underhill 1999). Although it may seem that retailers
have little control over this process, social influence can
indeed be a critical element of the firm’s selling strategy.
Retailers can exert direct control over the social influence
process by managing how sales associates interact with cus-
tomers (Pennington 1968; Weitz 1981). Salespeople and
customer service representatives can be instructed to engage
shoppers in conversations, offer assistance, provide product
suggestions and recommendations, and give feedback to
shoppers on their product selections.
The retailer can also train employees to adapt their per-
sonal selling strategies to the specific social scenarios that
occur spontaneously during the store’s daily operations
(Franke and Park 2006; Weitz, Sujan, and Sujan 1986).
Whereas product assortments, pricing, and merchandising
may be relatively stable over time, the customer’s in-store
experience changes from minute to minute as people enter
the store, interact with the merchandise, strike up conversa-
tions with acquaintances and sales associates, and talk on
their cell phones. These changes in the size, composition,
and activities of social groups can moderate the influence of
salesperson interactions and other marketing variables.
In addition, the retailer can indirectly manipulate the
social environment through the scheduling of promotions,
product assortment decisions, and the design of the physical
space (e.g., Lam et al. 2001). Appealing promotions and
unique products can create high levels of shopper traffic,
increasing the perceived popularity of the store and encour-
aging shoppers to touch and buy the merchandise. Retailers
need to be cautious not to create too much traffic because
shoppers may become stressed by the crowds and leave
without making a purchase (Harrell, Hutt, and Anderson
1980). Similarly, the store layout can showcase a variety of
attractive products, but if the aisles are too narrow, shoppers
may feel that they do not have sufficient personal space and
may exit the store prematurely (Underhill 1999).
Prior research has provided significant insight into the
effects of the social environment on shopper perceptions
and behavior (e.g., Argo, Dahl, and Manchanda 2005; Argo,
Dahl, and Morales 2008; Dahl, Manchanda, and Argo
2001), but it has tended to focus on individual sources of
social influence rather than their interaction, and studies
typically measure perceptual and behavioral outcomes
rather than the shopping process. In most retail stores, cus-
tomers are subject to multiple social forces simultaneously

Social Influence on Shopper Behavior / 25
(e.g., salesperson contact, group discussions, crowding con-
ditions) as they navigate through the store. For example, a
customer may be greeted by an employee or involved in
discussions with companions while walking through a
crowded store and pausing to examine merchandise. It
would be valuable to know how these different types of
social influence affect shoppers’ touch and purchase behav-
iors, how these forces interact, and which tools are most
effective at moving customers along the path to purchase.
These insights have important implications for sales man-
agement and the design of retail spaces.
The goals of this research are to understand how the
interactions between the various social elements of the store
environment affect the shopping process (shopper engage-
ment, as measured by product touch and interaction, and
purchase conversion) and to develop recommendations for
store managers to tailor the environment to optimize store
productivity. We examine social influence from the per-
spective of social impact theory (SIT; Latané 1981). Using
a bivariate hierarchical Bayes model of the shopping
process, we study the influence of crowding, shopping
group size, customer interaction, and salesperson interac-
tion on product touch and purchase, controlling for in-store
marketing activities, shopping path, and other environmen-
tal factors.
This study uses a unique video tracking data set that
records consumers’ entire shopping trips as well as outcome
information in an apparel store. The video data capture
crowding conditions in the store, consumers’ interpersonal
contact with companions and sales associates, cell phone
conversations, product touch and trial, and other activities
that are important for understanding social influence on
shopper behavior. Consumers and sales associates are also
visually classified into demographic groups (gender, age,
and ethnicity) to measure the impact of demographic simi-
larity. In combination with the transactional data, the video
tracking provides a comprehensive picture of the cus-
tomer’s in-store shopping experience.1
We contribute to the shopper marketing literature in
three ways. Theoretically, we extend SIT to explain the
influence of social factors on shopper touch and purchase
during the course of a retail store visit; the interplay
between interactive and noninteractive social influences;
and the moderating roles of product categories, shopping
group size, and shopper demographics. We discover that
two interactive social influences, group discussion and
salesperson contact, tend to slow the shopper down, encour-
age a longer store visit, and increase product interaction and
purchase. In contrast, the noninteractive influence of
crowding has an inverted U-shaped relationship with prod-
uct touch frequency and a negative impact on customer pur-
chase. The latter effect is much larger than the interactive
influences, contrary to the predictions of SIT.
Substantively, our study provides several novel and
actionable insights for retail store managers to help improve
the customer’s in-store shopping experience and purchase
conversion rate. For example, we find that it is beneficial
for sales associates to target shoppers who are shopping
alone, have similar demographic characteristics, and are
shopping in departments where there may be higher uncer-
tainty, such as the new arrivals or clearance sections. Shop-
pers who carry shopping bags, take a straighter path, walk
more slowly, and shop on the gender-appropriate side of the
store also seem to be promising prospects on the basis of
their propensity to touch and/or purchase items. Further-
more, store managers may run traffic-driving events during
quiet periods to increase social density and stimulate shop-
per interest, especially for new arrivals and clearance items.
Methodologically, we develop a bivariate normal process
model to capture shoppers’ touch frequency and purchases.
This approach treats each shopper’s visit to each department
or zone as a social scenario that can be mined for insights
about what drives shopper engagement and purchase. In
addition, we introduce a new approach for coding and ana-
lyzing customer tracking videos, which can be applied to
the mining of video data collected in other contexts.
The remainder of the article is organized as follows. In
the next section, we review the prior literature on social
influence and the shopping process, summarize our concep-
tual framework, and present a set of hypotheses. Then, we
describe our modeling approach and present the empirical
analysis of the tracking data from an apparel store. We con-
clude with a discussion of the results and managerial impli-
cations of this research.
Background
Prior research has revealed that the social context can have
a significant impact on shopper perceptions and behavior.
In some cases, this influence is a consequence of the direct
interaction between the customer and other shoppers in the
store or with a salesperson or customer service representa-
tive. For example, Kurt, Inman, and Argo (2011) find that
men (but not women) spend more when they shop with a
friend. Luo (2005) reports that when shoppers imagine
being accompanied by peers, this increases impulse buying,
but an imagined trip with family members reduces spend-
ing. Baker, Levy, and Grewal (1992) find that the helpful-
ness of salespeople in a simulated store visit heightens
shopper arousal and willingness to buy.
In other cases, the influence is due to the mere presence
of other shoppers in the store. For example, Argo, Dahl, and
Manchanda (2005) find that a shopper in the presence of
one other person tends to be happier than when shopping
alone, but when the number increases to three, happiness
drops and the shopper becomes annoyed. When the product
category is considered “sensitive” (e.g., condoms), a novice
shopper will be embarrassed by the presence of others
(Dahl, Manchanda, and Argo 2001). Moreover, when shop-
pers see others touching the merchandise, they may per-
ceive that the product is “contaminated” and lower their
evaluations (Argo, Dahl, and Morales 2006). The exception
is when the product is touched by an attractive shopper of
the opposite sex, which can increase product evaluations
and willingness to pay (Argo, Dahl, and Morales 2008).
1A drawback of video tracking data is that the manual coding of
activities and validation of the automated customer tracking can
be time consuming.

Recently, marketers have emphasized the importance of
going beyond perceptual, attitudinal, and behavioral out-
come measures and recording the customer’s actual shop-
ping and purchasing activities in the store (Burke 2006;
Hui, Fader, and Bradlow 2009a). Shoppers are often unable
to report accurately on their behavior and the causal factors
because of forgetting habitual buying, unconscious influ-
ences, and social desirability biases (cf. Nesbitt and Wilson
1977). Observational methods enable the researcher to
record shopping activities over time and space and measure
the impact of social influence at each stage in the shopping
process and physical location in the store. One might expect
that in apparel, for example, the social factors influencing a
shopper’s likelihood of touching a product may be different
from those affecting purchase. Similarly, the social factors
affecting whether a shopper buys a new product may be dif-
ferent from those affecting purchase of a commodity or
clearance item.
Several recent studies have used methodological inno-
vations, such as radio-frequency identification (RFID) tags,
computer vision, wearable video cameras, handheld bar-
code scanners, and clickstream analysis, to record the cus-
tomer’s observable movement in, and interaction with, a
physical retail store, shopping website, or simulated shop-
ping environment (Hui, Fader, and Bradlow 2009a; Hui et
al. 2013; Stilley, Inman, and Wakefield 2010). In contrast
with scanner panel data, the shopping path data encode the
sequence of events leading up to a purchase (Montgomery
et al. 2004). Path tracking research has significantly
enhanced researchers’ understanding of the grocery shop-
ping process. For example, studies have revealed that gro-
cery shoppers tend to become less exploratory and more
purposeful (shopping and buying) as their trip progresses,
they are more likely to dwell in “vice” categories after pur-
chasing “virtue” products, and they are drawn to areas of
the store with high shopper density but spend less time vis-
iting these regions (Hui, Bradlow, and Fader 2009). Shop-
pers pick up their purchased products in an order that is
close to ideal from the perspective of traveling salesman
optimality, but they tend to deviate from the most efficient
point-to-point path (Hui, Fader, and Bradlow 2009b). Shop-
pers are more likely to consider making unplanned pur-
chases in categories that are on promotion, have hedonic
qualities, are refrigerated, and are encountered later in the
trip, whereas purchase conversion is higher for products
encountered earlier (while there is still “slack” in the time
budget) and when shoppers stand closer to the shelf (Hui et
al. 2013).2
Although recent path tracking research has generated
several valuable insights about shopper behavior, it has not
been as helpful in illuminating the social dynamics of retail
shopping. In part, this is due to methodological constraints.
Studies using RFID tags to record the location of shopping
carts can measure whether shoppers and their carts tend to
move toward or away from each other (e.g., attraction and
crowding effects), but they do not reveal the size of the
shopping party, shopper discussions, salesperson contact, or
product interactions. A second challenge is that most of this
research has been conducted in grocery stores. These self-
service environments tend to have more solitary shoppers, a
higher percentage of planned purchases, and fewer
instances of social influence than other retail formats.
The present research investigates both interactive and
noninteractive social influences driving product touch and
purchase using video tracking data. This provides a more
complete picture of the dynamics of consumers’ in-store
behavior and the store environment than self-report (e.g.,
survey, exit interview) or RFID-based tracking studies.
Whereas previous studies have typically used grocery data,
our data are from a specialty apparel retailer in a shopping
mall, which allows the examination of customer–salesperson
interactions, an important interactive social influence inside
a store. In the next section, we introduce a conceptual
framework based on SIT and derive a set of hypotheses,
which we then test using a bivariate model of shopper prod-
uct touch and purchase (Poisson and probit) implemented in
a hierarchical Bayes framework.
Theory and Hypotheses
Latané (1981) developed a theory of social impact that pro-
vides a useful foundation for understanding the influence of
social factors on shopper behavior. He defines “social
impact” as “the great variety of changes in physiological
states and subjective feeling, motives and emotions, cogni-
tions and beliefs, values and behavior, that occur in an indi-
vidual, human or animal, as a result of the real, implied, or
imagined presence or actions of other individuals” (p. 343).
According to SIT, the degree to which a person’s behavior
is influenced by another is a function of the strength of the
source of impact, the immediacy of the event, and the num-
ber of sources exerting an influence on the target person.
Social impact will be greater when the source of influence
has higher status or a closer relationship with the target,
when it has high spatial and/or temporal proximity to the
target, and when there are multiple sources of influence.
Social impact theory predicts that the influence the target
person experiences is a multiplicative function of these
three factors and increases as a power, t, of the number of
sources, where t < 1. Conversely, as the number of different
sources increases, the relative impact of each source on the
target declines; as the number of targets increases, a
source’s impact on each individual target is reduced.
Applications of SIT in psychology have typically exam-
ined the influence of the social context on a person’s judg-
ment or behavior at a specific time for a narrowly pre-
scribed set of conditions (e.g., Latané 1981). However, in a
retail context, shopping takes place over time and space,
and social factors change dynamically as customers enter
the store, navigate the aisles, interact with salespeople, and
shop from the available selection of merchandise. In a rela-
tively short period of time, there may be hundreds of unique
social encounters. Interpersonal factors may play a different
role in different departments and product categories, at dif-
26 / Journal of Marketing, September 2014
2Shopper intercept studies—including Inman, Winer, and Fer-
raro (2009), Stilley, Inman, and Wakefield (2010), and Nordfalt
(2009)— provide additional perspective on the factors driving in-
store shopper behavior.

Social Influence on Shopper Behavior / 27
ferent stages in the decision process, and for different shop-
per profiles.
To understand the dynamics of social influence in a
retail setting, we divide the shopping trip into a series of
department or zone visits and separately model the factors
affecting product touch and purchase for each visit. As
shoppers walk through the store, they will often stop to
touch and interact with products as part of the information
search and evaluation process. This physical interaction
helps reduce the uncertainty and risk associated with making
a choice. Touch enables the shopper to assess a product’s
material properties, such as texture, hardness, temperature,
and weight (Peck and Childers 2003), which are important
sensory features for evaluating a variety of products, such
as apparel, electronics, and automobiles (Underhill 1999).
The act of touch can also increase shopper engagement,
increase the perceived “ownership” of the object (Peck and
Shu 2009), and stimulate impulse purchasing (Peck and
Childers 2006). Retailers often encourage consumers to
touch and/or try out merchandise to heighten involvement.
In this research, we define touch as any type of physical
contact between the shopper and a product, except the final
touch of carrying the product to the checkout counter for
purchase. A customer’s frequency of product touch indi-
cates the seriousness of his or her interest in, or level of
engagement with, the product. Although there will be indi-
vidual differences in the amount of product examination
required before making a purchase decision, in general, the
more often products are touched, the higher the level of
consideration.
After a shopper identifies a desired product and possi-
bly touches it or tries it on, he or she will decide whether to
make a purchase. Purchase conversion occurs when there is
a close match between consumers’ shopping needs and the
available merchandise. In general, we would expect that
social factors will have a greater impact on product touch
than purchase because touch reflects product interest,
unconstrained by the affordability of the item, whereas pur-
chase requires a financial commitment. Both touch (an indi-
cator of shopper engagement) and purchase conversion rate
are important retail performance measures for customer
experience management (Burke 2006; DeHerder and Blatt
2011).
As customers shop the store, they are subject to both
interactive and noninteractive social influences. Interactive
social influence occurs when a shopper holds a conversa-
tion with another person in the store, such as a sales associ-
ate or companion (e.g., Goff and Walters 1995; Leibowitz
2010; Underhill 1999). We anticipate that interactive social
influences will slow down the shopper’s movement through
the store, encourage product interaction, and (assuming a
positive conversation) increase purchase likelihood. Social
influence can also occur without direct interaction, such as
when a shopper sees other customers in the store and
observes their behavior. Again, this social influence may
slow shoppers down, attract their interest, and encourage
them to browse the merchandise. However, crowds may
also produce a negative emotional response (Argo, Dahl,
and Manchanda 2005) and cause shoppers to spend less
time in a department or store (Hui, Bradlow, and Fader
2009).
Extending SIT to the retail shopping context, we predict
that the influence of other people in a store on a target shop-
per’s behavior will increase as (1) the number of other
people increases, (2) the immediacy of their interaction
with the shopper increases (e.g., personal conversations),
and (3) the strength of their relationship increases (e.g.,
family members). At the same time, the influence of a sales-
person on a target shopper will decrease as (1) the number
of other people competing for the salesperson’s attention
increases and (2) the number of other people influencing
the target shopper increases. Figure 1 provides a conceptual
framework summarizing the influence of these factors on
shopper behavior. We present individual hypotheses in the
following sections.
We do not expect that the shape of the relationship
between the number of sources of influence and shopper
behavior will follow a power function, as predicted by SIT.
In a retail store, the presence of other shoppers provides
information about the desirability of products, but it also
creates a physical obstruction that can interfere with navi-
gation, product interaction, and purchase. We predict that
the combined effects of these factors will appear as a down-
ward concave function, as we discuss next. We also expect
that the effects of social influence will depend on the prod-
uct department and the individual characteristics of shop-
pers and sales associates.
In the following subsections, we focus on three of the
most common types of social influence in a retail setting:
shopper density (crowding), salesperson contact, and dis-
cussions with companions (e.g., friends, family). We exam-
ine their impact on consumer touch frequency and purchase
decisions and the moderating roles of group size and product
category. We also explore the interactions between noninter-
active and interactive social influences because customers
are often subject to multiple influences simultaneously.
Shopper Density
Shopper density is the physical density or concentration of
shoppers within a given space. Social impact theory sug-
gests that the larger the number of customers in the shop-
FIGURE 1
Social Influence in a Retail Context
Shopper Density (Crowding)
Shopper Group (Sources/Targets)
TouchSalesperson (Source) Purchase Shopper (Target)

ping area, the greater the number of sources, and thus the
larger the social impact on the focal consumer. Similarly,
social proof theory suggests that the presence of other shop-
pers in a store zone may signal high-quality products, thus
increasing shopper interest (Cialdini and Goldstein 2004)—
the reasoning being that if many others are interested in a
product, it must be good.
Before touching a product, consumers may be uncertain
about the quality or desirability of the products and prices
in the store (Bell and Lattin 1998) and thus learn informa-
tion by observing other shoppers. Customers may follow
the crowd, engaging in “herd behavior,” which describes
how individuals in a group or crowd act together without
planned direction. Herd behavior may be common in every-
day decisions and works on people’s perceptions that large
groups cannot be wrong (Cialdini and Goldstein 2004). In
the marketing literature stream, herd behavior occurs in
Internet marketing (Hanson and Putler 1996). For example,
Dholakia, Basuroy, and Soltysinski (2002) show that buyers
in digital auctions gravitate toward listings with existing
bids and away from listings without bids, and Chen, Wang,
and Xie (2011) find evidence of observational learning
from others’ choices in online retailing. Similarly, the pres-
ence of more customers in a department or category signals
its attractiveness and product quality (Becker 1991). Hui,
Bradlow, and Fader (2009) report that, in general, the pres-
ence of other shoppers leads a consumer to visit a store zone.
However, when the store becomes more crowded, a
state of psychological stress results if a customer’s demand
for space exceeds the supply (Stokols 1972). Higher shop-
per density can be associated with negative feelings and
coping strategies (Argo, Dahl, and Manchanda 2005;
Arnold et al. 2005; Harrell, Hutt, and Anderson 1980). It
can create physical and psychological barriers to shopping,
reducing access to products and causing a lack of privacy.
An overcrowded store area decreases customers’ demand
and interest level. Therefore, we hypothesize that a cus-
tomer will be more likely to touch the products when there
are other shoppers present; however, as crowding increases,
touch frequency will decrease. Formally,
H1a: Shopper density (crowding) has an inverted U-shaped
relationship with touch frequency.
We expect that crowding has a negative impact on pur-
chase because it can obstruct the shopping process and cre-
ate a psychological state of stimulus overload from inappro-
priate or unfamiliar social contacts (Harrell, Hutt, and
Anderson 1980). Milgram’s (1970) theory regarding adap-
tation strategies to overload suggests that, in crowded con-
ditions, consumers will allocate less time to each stimulus
input and thus give less time to each purchase decision.
Eroglu, Machleit, and Barr (2005) demonstrate a significant
relationship between perceptions of crowding and the dis-
ruption of the pursuit of important activities and goals. Hui,
Bradlow, and Fader (2009) also find that although the pres-
ence of other shoppers attracts consumers to a store zone, it
reduces consumers’ tendency to shop there in a grocery
store setting. Thus,
H1b: Crowding discourages consumers from buying.
Within-Group Discussions
Shopping is often a social activity, whereby consumers visit
stores with friends, family, or peers and talk to others in
their shopping group. Inman, Winer, and Ferraro (2009)
find that shopping with others, especially members of the
same household, leads to a higher incidence of need recog-
nition. Shopping companions may make recommendations
by pointing out or suggesting products to the lead customer,
which can extend the shopping trip and result in more
instances of product touch.
We also expect that discussions with fellow shoppers in
the same shopping group will encourage consumers to buy
because interacting with group members provides more
information and reduces the perceived risk of purchase
(Underhill 1999; Willis 2008). Hartmann (2010) finds that
social interactions within groups have a strong impact on
purchase decisions. Underhill (1999, p. 158) reports that if a
store can create an atmosphere that fosters product discus-
sion, the merchandise “begins to sell itself.” Willis (2008)
investigates the role of conversation at or near the time of
purchase and finds that the nature and context of the con-
versation can change shopper behavior. Thus,
H2: A consumer who talks frequently to other shoppers in his
or her shopping group will touch products more fre-
quently and is more likely to make a purchase.
Salesperson Contact
Salespeople can play an important role in the shopping
process (Sharma 2001). They can help customers find the
desired products (Von Riesen 1974) and stimulate interest
in new arrivals (Goff, Bellenger, and Stojack 1994). Sales-
people help convert needs into purchases by addressing
shoppers’ concerns and focusing and reinforcing customers’
desires. They are the most important factor in managing the
customer experience (Smith and Wheeler 2002), and their
interpersonal effort and engagement are especially impor-
tant in discriminating between delightful and terrible shop-
ping experiences (Arnold et al.2005). Underhill (1999)
reports that purchase conversion rates increase by 50%
when salespeople initiate contact. Thus, salesperson contact
should increase the incidence of product touch and purchase.
H3a: A consumer who is approached by and interacts with a
salesperson will touch products more frequently and is
more likely to make a purchase.
When stores become crowded, sales associates have less
opportunity to interact with and influence each shopper,
reducing their impact on product touch and purchase. Retail-
ers attempt to address this issue by “staffing up” during busy
periods, but it can be difficult to accurately predict traffic
levels and sales resource requirements, which can change
on an hourly basis. Therefore, we expect the following:
H3b: Store crowding reduces the impact of salesperson contact
on a consumer’s frequency of product touch and likeli-
hood of purchase.
The Moderating Role of Shopping Group Size
Research on group dynamics has suggested that group for-
mation in crowded environments helps mitigate the nega-
28 / Journal of Marketing, September 2014

Social Influence on Shopper Behavior / 29
tive effects of crowding (Baum, Harpin, and Valins 1975).
Group membership enables people to regulate, control, or
avoid exposure to the harmful effects of crowding by pro-
ducing boundaries that reduce the experience of crowding
(Paulus and Nagar 1989). Cohesive groups can shield group
members from unwanted interactions, insulating the group
from the external world (Willis 2008). Social impact theory
leads to a similar prediction: as the size of the shopping
group increases, the relative impact of crowds on the target
shopper’s behavior will be reduced. Thus,
H4: As the shopping group increases in size, it reduces the
impact of crowding on shoppers’ touch frequency and
purchase.
As we discussed previously, SIT predicts that the degree
of social influence is positively related to the source
strength and the number of people who are the sources of
influence. Because the shopping group typically includes
family members, friends, or peers who have previous
relationships with each other before the shopping trip, the
source strength of these people is high. Furthermore, as the
size of a shopping group increases, there are more sources
of impact. Therefore, we expect that within-group discus-
sions will have a greater impact on consumer touch and
purchase decisions.
H5: The shopping group size strengthens the impact of within-
group discussions on touch frequency and purchase.
According to SIT, when a group is the target of influence,
the social impact will be divided among all the individual
members. People feel less accountable as the number of
group members increases. If a salesperson talks to a shop-
ping party, as the shopping group grows larger, the amount
of persuasion experienced by each shopper in the group will
be smaller. Thus,
H6: The shopping group size mitigates the impact of sales-
person contact on product touch frequency and purchase.
Product Type
In addition to group size, there are several other contextual
and motivational factors that may affect how shoppers
respond to social forces at the point of purchase. One is the
type of product category being shopped. Shoppers have dif-
ferent levels of familiarity with the merchandise sold in
retail stores, and this will affect their sensitivity to interac-
tive and noninteractive social influences. For new items,
such as seasonal apparel and consumer electronics, shop-
pers typically have limited product knowledge and will be
more receptive to social cues from friends, sales associates,
and other shoppers to decide “what’s hot and what’s not.”
Shoppers who are hunting for bargains will also rely on
social cues (e.g., people congregating around a clearance
rack, conversations with salespeople) to spot the best deals.
In contrast, social factors will play less of a role in the pur-
chase of commodity products (e.g., khakis, polo shirts) and
may actually interfere with shopping for highly personal
items, such as fashion accessories and underwear. To
explore these relationships, we incorporate both the main
effects of product category (i.e., new arrivals, accessories,
and clearance items) and the interactions of category with
the key social influence variables (crowding, within-group
discussions, and salesperson contact) on touch frequency
and purchase to measure these effects.
Shopper and Salesperson Demographics
Another important consideration is the demographic simi-
larity of the shopper and sales associate. Shoppers may be
more likely to identify with, and be influenced by, sales-
people who are the same gender, ethnicity, and/or age as
they are (Evans 1963). To investigate this issue, we created
three binary variables—gender congruence, age congru-
ence, and ethnicity congruence—to capture whether the
gender, age, and/or ethnicity of the salesperson matches the
shopper during an interaction.
Shopper Motivation and In-Store Activities
Prior research by Hui, Bradlow, and Fader (2009) and Mont-
gomery et al. (2004) reveals that a customer’s likelihood of
shopping/buying may also be influenced by his or her shop-
ping path through the store. We control for such effects by
incorporating the shopper’s shopping path information:
cumulative sinuosity (curvature of the path), distance covered,
and whether the customer stays on the gender-appropriate
side of the store. Furthermore, Stilley, Inman, and Wake-
field (2010) demonstrate the impact of sales promotion on
in-store decision making, and Inman, Winer, and Ferraro
(2009) find that customer activities and characteristics can
affect in-store unplanned purchases. Thus, we include in the
model the impact of marketing promotions (sidewalk sales)
and the shopper’s in-store activities (e.g., talking on the
phone, visiting the dressing room). Finally, prior research
(Goff and Walters 1995; Montgomery et al. 2004) has sug-
gested that a shopper’s motivation or orientation before the
store visit may affect his or her purchase behavior. There-
fore, we construct two proxy variables—the shopper’s ini-
tial speed of walking into the store and whether the cus-
tomer carries a shopping bag when entering the store—to
capture the shopper’s initial level of motivation. Walking
speed may reflect the intensity of information processing
and/or the urgency of the customer’s need, and the number
of shopping bags may indicate whether a customer is in a
“shopping mode” and thus more likely to buy (see Dhar,
Huber, and Khan 2007).
Model Setup
To test the hypotheses, we propose a bivariate model of
shopper product touch and purchase: a random-effects
model implemented in a hierarchical Bayes framework to
account for individual heterogeneity. We divide each shop-
ping path into a series of zone transitions (Farley and Ring
1966), which summarize a consumer’s movement through a
store. The unit of analysis is consumer i’s visit to a specific
zone or department at visit occasion t, which increases by 1
whenever he or she enters another zone. Therefore, the
shopper’s entire store visit is decomposed into a series of
zone visit occasions (t). The combination of video data and
transaction data enables us to trace exactly when and in
which zone product touch and purchase occur. Therefore,
we are able to model the dynamic influence of social factors

on both touch frequency and purchase probability at the
zone visit occasion level. We denote consumer i’s touch fre-
quency by TFit at zone visit occasion t. We also denote con-
sumer i’s zone purchase decision by Bit, which equals 1 if
he or she buys at occasion t and 0 otherwise.
A consumer’s touch and purchase decisions are
assumed to be determined by his or her latent product
engagement and purchase utility, respectively. The con-
sumer can touch products in the store without buying, and
vice versa.3Furthermore, there may be unobserved environ-
mental factors that drive both touch and purchase. There-
fore, we define a bivariate normal latent process by allow-
ing consumer product engagement (mit) and purchase utility
(Uit) to be correlated. Specifically, we have
where X1and X2are the covariates that affect the con-
sumer’s engagement and purchase utility, respectively,
which we discuss in detail subsequently. biand giare vec-
tors of individual-level coefficients of the covariates to be
estimated. The two error terms (eit and dit) are correlated
with a bivariate normal distribution of BN(0, S), where Sis
a 2 ¥2 variance–covariance matrix. Because of the binary
nature of the purchase decisions, we normalized the vari-
ance of dit to 1 for identification purposes.
Touch Frequency
We assume that a consumer’s touch frequency at a zone
visit occasion follows a Poisson distribution. A hierarchical
Poisson regression is deemed appropriate to model touch
frequency (Breslow 1984). Specifically, we have
As Equation 1 shows, we allow consumer engagement mit to
be a function of covariates (X1) including social influence
and marketing promotions. We also incorporate product
category variables (i.e., new arrivals, accessories, and clear-
ance), shopping path, and in-store activities to control for
their impacts on touch frequency. Formally, we have
(2) P TF n exp
n! .
it it it
n
()
()
== −µ µ
(1) log x
Ux ,
it 1it i it
it 2 it i it
µ=′β+ε
=′γ+δ





(3) log Crowding Crowding
(3) log TalkFreq SalesContact
(3) log Crowding TalkFreq
(3) log Crowding SalesContact
(3) log GroupSize+ Category
(3) log Crowding GroupSize
(3) log TalkFreq GroupSize
(3) log SalesContact GroupSize
(3) log Crowding Category
(
it i0 i1 it i2 it
2
it i3 it i4 it
it i5 it it
it i6 it it
it i7 it i8c it
it i9 it it
it i10 it it
it i11 it it
it i12c it it
i
µ=β+β +β
µ+β +β
µ+β ×
µ+β ×
µ+β β
µ+β ×
µ+β ×
µ+β ×
µ+β ×
µ
where bi0 denotes consumer i’s intrinsic propensity to touch
a product in the store. The coefficient bi1 captures the influ-
ence of crowding on consumer touch frequency, and bi2
models the potential curvilinear relationship with touch
(H1a). The variable TalkFreqit refers to the frequency of the
consumer’s conversations with fellow shoppers, and the
coefficient bi3 captures the main effect of these discussions
on consumer i’s touch frequency (H2).4The coefficient for
SalesContactit, bi4, captures the main effect of salesperson
contact on shopper i’s touch frequency (H3a), and bi6 mea-
sures how crowding conditions interact with salesperson
contact to determine touch frequency (H3b). The parameter
bi9 captures the moderating effect of shopping group size
on crowding (H4), bi10 measures its moderating effect on
discussions (H5), and bi11 captures its moderating effect on
salesperson contact (H6).
The coefficient bi8c captures the impact of product cate-
gories (i.e., new arrivals, accessories, and clearance) on
touch frequency, where c represents different product cate-
gories. Parameters bi12c, bi13c, and bi14c specify the moder-
ating effects of product category on the impact of crowding,
within-group discussions, and salesperson contact on touch
frequency, respectively. The coefficients bi17, bi18, and bi19
capture the effects of gender, ethnicity, and age congruence
between the consumer and the salesperson during instances
of salesperson contact, respectively. The term Pathit
includes the consumer’s cumulative sinuosity of his or her
path (path curvature), distance covered, and whether the
consumer stays on the side of the store matching his or her
gender; the vector of coefficients bi20 estimates the effects.
The coefficient bi21 measures the consumer’s time spent in
cell phone conversations, and bi22 indicates the impact of
the store’s sidewalk sale on a consumer’s touch behavior.
To capture individual heterogeneity, we model the vec-
tor of response coefficients biusing a random effects model
(Rossi, McCulloch, and Allenby 1996):
where bis a vector of aggregate level means of bi. The
unobservable heterogeneity component h1i is normally dis-
tributed with mean 0 and variance–covariance matrix Sb.
(3) log TalkFreq Category
(3) log SalesContact Category
(3) log EntrySpeed Shopbag
(3) log GenderCongruence SalesContact
(3) log EthCongruence SalesContact
(3) log AgeCongruence SalesContact
(3) log Path PhoneTime SideSale ,
i
it i13c it it
it i14c it it
it 15 i 16 i
it i17 it it
it i18 it it
it i19 it it
it i20 it i21 it i22 it it
µ
µ+β ×
µ+β ×
µ+β +β
µ+β ×
µ+β ×
µ+β ×
µ+β +β +β +ε
(4) , ~ MVN 0, ,
i1i1i
()
β=β+η η Σ
β
30 / Journal of Marketing, September 2014
3“Purchase without touch” would be a grab-and-go purchase
with minimal product interaction.
4We also tried modeling the consumer’s talk duration with fellow
shoppers and obtained similar results. Talk duration and frequency
are highly correlated in our data.

Social Influence on Shopper Behavior / 31
Purchase
We assume that the consumer’s purchase decision at a zone
visit occasion is driven by his or her purchase utility (Uit),
which is a function of the same classes of social influence,
marketing, and other control variables used in the previous
model. We build in a recursive structure in which touching
leads to purchase, but not vice versa, to account for the pos-
sibility that touching has a direct influence on purchase in
addition to the other explanatory variables. Therefore, we
add four independent variables—touch frequency during
the current zone visit (TouchFreqit), cumulative touch fre-
quency in the current zone up to last zone visit occasion
(CTF_Sameit – 1), cumulative touch frequency in other
zones than the current zone (CTF_Diffit – 1), and dressing
room visit in the current zone (Dressroomit)—to the pur-
chase utility function to capture the impact of consumer
touch behavior on purchase. Formally, we have
where gi0 denotes consumer i’s intrinsic preference to pur-
chase. The explanatory variables are defined similarly to
the counterparts in the consumer engagement function in
Equation 3. The parameter gi1 captures the impact of crowd-
ing on consumer i’s purchase decision (H1b). The coeffi-
cient gi3 denotes the influence of within-group discussions
on purchase (H2). The parameter gi4 tests the main effect of
salesperson contact (H3a), and gi6 measures the interaction
between crowding and personal selling (H3b). The coeffi-
cient gi9 measures the moderating effect of shopping group
size on crowding (H4), gi10 measures its moderating effect
on group discussions (H5), and gi11 captures its moderating
influence on salesperson contact (H6).
=γ +γ +γ
+γ +γ
+γ ×
+γ ×
+γ γ
+γ ×
+γ ×
+γ ×
+γ ×
+γ ×
+γ ×
+γ +γ
+γ ×
+γ ×
+γ ×
+γ +γ +γ
+γ +γ +γ
+δ
−
−
(5) U Crowding Crowding
(5) U TalkFreq SalesContact
(5) U Crowding TalkFreq
(5) U Crowding SalesContact
(5) U GroupSize+ Category
(5) U Crowding GroupSize
(5) U TalkFreq GroupSize
(5) U SalesContact GroupSize
(5) U Crowding Category
(5) U TalkFreq Category
(5) U SalesContact Category
(5) U EntrySpeed Shopbag
(5) U GenderCongruence SalesContact
(5) U EthCongruence SalesContact
(5) U AgeCongruence SalesContact
(5) U Path TouchFreq CTF_Same
(5) U CTF_Diff Dressroom SideSale
(5) U ,
it i0 i1 it i2 it
2
it i3 it i4 it
it i5 it it
it i6 it it
it i7 it i8c it
it i9 it it
it i10 it i t
it i11 it i t
it i12c it it
it i13c it it
it i14c i t it
it 15 i 16 i
it i17 it i t
it i18 it it
it i19 it it
it i20 it i21 it i22 it 1
it i23 it 1 i 24 it i 25 it
it it
Similar to Equation 4, we also model the vector of the
coefficients giusing a random effects approach:
where gis the aggregate mean of gi. We assume the unob-
served heterogeneity component h2i to follow a multivari-
ate normal distribution of MVN(0, Sg) with mean 0 and
variance–covariance matrix Sg.
Given the binary probit model specification, we have
the following observed purchase decisions:
We use the hierarchical Bayes approach to estimate the
parameters of the model (Rossi, McCulloch, and Allenby
1996). The Web Appendix presents the detailed estimation
procedure and Markov chain Monte Carlo algorithm.
Data
The data were collected in a retail apparel store located in a
suburban shopping mall in the Midwest region of the
United States. Its customer demographics represent a
diverse population on age (67% adults, 31% teenagers, and
2% children) and gender (59% female and 41% male shop-
pers), with somewhat less variance on ethnicity (86% Cau-
casian, 14% others). The test store belongs to a popular
retail chain based in the United States, and all of the chain’s
stores use a standardized floor plan and assortment of mer-
chandise. The store format is typical of other specialty
apparel stores in the United States, with men’s merchandise
organized on one side of the store, women’s on the other
side, a cash wrap area in the center, and fitting rooms in the
back.
A six-lens panoramic video camera was installed in the
ceiling of the store and recorded customer shopping activi-
ties within the store (see Figure 2). We collected and ana-
lyzed the data for three days in 2006: January 10, 12, and
19. We selected these specific dates because there was a
weeklong sidewalk sale beginning on January 12, 2006, and
we wanted to compare shopper behavior during the sale
with behavior immediately before and after the sale (Janu-
ary 10 and 19, respectively). In addition, shopping patterns
and social influence processes during this period are likely
to be more typical of the year than during the holiday shop-
ping season. Each data set is approximately one hour, from
4:00 P.M. to 5:00 P.M., which is the peak period of customer
traffic during the day. A computer program tracked cus-
tomers’ movement and navigation to generate information
on shopping path and store conditions (e.g., crowding).
We manually coded customer demographics and in-
store activities (e.g., touching, trying on, buying products,
talking to fellow shoppers) and cross-checked them at a
later time. We added five time tags as well: the start and end
points of product touch, phone conversations, salesperson
contact, conversations with fellow shoppers, and dressing
room visits. Therefore, we know whether these events
occurred and the exact time and duration of the events.
(7) B 1, if U 0
0, otherwise .
it
it
=>





(6) ,
i2i
γ=γ+η

We define touch as any type of zone-level physical con-
tact between the shopper and a product, except the final
touch of carrying the product to the checkout counter for
purchase.5The measurement of touch frequency is specific
to both the shopper and the visit occasion. We define group
size as the number of members in a shopping group or
party. Talk frequency is measured by the number of distinct
conversations between a shopper and his or her compan-
ions. When there is an apparent pause between two conver-
sations (longer than five seconds), we increased the fre-
quency of customer discussions by 1. We define salesperson
contact as verbal interaction between the shopper and a
sales associate (e.g., offering help and information),
32 / Journal of Marketing, September 2014
FIGURE 2
Retail Store Layout
A: Panoramic Image of Store and Customer Tracking
B: Floor Map
E
N
ZO
1
s
’
n
Me
L
ead
E
E
N
ZO
3
s
’
n
Me
e
l
d
d
Mi
N
ZO
7
s
’
n
Me
e
c
n
a
r
a
e
l
C
/
e
l
Sa
E
N
ZO
0
e
c
n
a
r
t
En
k
l
a
w
e
d
i
(S
)
e
l
Sa
E
N
ZO
4
s
e
i
r
o
s
s
e
c
Ac
E
N
ZO
6
t
u
o
k
c
e
Ch
2
Z
E
N
ZO
s
’
n
e
m
Wo
L
ead
W
M
E
N
ZO
5
s
’
n
e
m
Wo
e
l
d
d
Mi
E
N
ZO
8
s
’
n
e
m
Wo
e
c
n
a
r
a
e
l
C
/
e
l
Sa
5We also exclude very brief (less than .5 second) incidental or
accidental product touch from the measurement.

Social Influence on Shopper Behavior / 33
exclusive of initial greetings. We define crowding as the
number of other shoppers present in a zone 30 seconds
before the consumer enters the zone.
Figure 2, Panel A, provides an example image of video
tracking, with the ellipses representing projected models of
each human body and the numerical labels containing a
unique tracked customer ID. When we combine video data
with transactional data, we can calculate overall store trip–
level purchase conversion rates. It also enables us to iden-
tify exactly which zone and when (in terms of zone visit
occasion) a purchase occurs. From the product placement,
we divided the store into eight zones or spatial clusters after
consulting with the store manager and referring to the store
layout. Figure 2, Panel B, presents the store floor map. The
store has two large rooms divided by wing walls: Room 1
(Zones 1–5) in the front of the store and Room 2 (Zones 6–
8) in the back. At the time of data collection, all of Room 2
featured products on sale or clearance. This was a unisex
store, with men’s merchandise on the left-hand side as cus-
tomers enter the store (Zones 1, 3, and 7) and women’s
products on the right-hand side (Zones 2, 5, and 8). We sep-
arated out the accessories area (Zone 4), which includes a
jewelry tower, sunglass fixture, men’s H-unit fixture (fea-
turing boxers and belts), and the checkout area (Zone 6).
There were no product displays at the checkout counter;
customers simply completed their transactions there. We
traced each purchase to the zone where the product was fea-
tured and picked up by a customer.
After closer inspection of Room 1, we distinguished the
lead zones (Zones 1 and 2) from the middle zones (Zones 3
and 5). A row of lead tables and wardrobes featured new
arrivals displayed in the most conspicuous places. Fixtures
between the lead zones and Room 2 were classified in the
middle zones, and they displayed full-priced general
apparel items, such as khakis, denim jeans, and polo shirts.
We tracked both shopping paths and customer in-store
activities at the zone visit occasion level.
In the three hours of video, there were 1,400 observa-
tions (zone visits) in total, which represent the paths of 169
completely tracked customers, with 63 buyers. The sample
store trip–level purchase conversion rate was 37.3%. For
the three days of observation, there were 19 buyers out of
59 customers on the first day (32.2%), 22 out of 54 on the
second day (40.7%), and 22 out of 56 on the third day
(39.3%). The average zone-level touch frequency was .52,
and the mean zone purchase probability was .07.
We grouped the variables derived from the data into seven
classes: (1) social influence factors: crowding, frequency of
speaking to fellow shoppers in the same shopping group,
salesperson contact, and shopping group size; (2) product
category: new arrivals (lead zone), accessories, clearance,
and regular/middle zone; (3) marketing promotions: side-
walk sales; (4) shopper in-store activities: talking on the
phone, cumulative touch frequency in the current zone,
cumulative touch frequency in other zones, and dressing
room visits; (5) shopping path information: cumulative sin-
uosity, distance covered, and whether the customer stays on
the gender-appropriate side of the store; (6) proxies for
shopper motivation: initial speed of walking into the store
and whether the customer carries a shopping bag when
entering the store; and (7) demographic congruence: the
match of the gender, age, and ethnicity of the customer and
salesperson in the presence of salesperson contact. A list of
zone-level summary statistics for these variables appears in
Table 1. For model validation purposes, we use January 12
and 19 data as the estimation sample and use January 10
data as the holdout sample. In the estimation sample, we
have 740 observations.
Before presenting the results of the models, we first plot
touch frequency and purchases under different social
conditions to determine whether the relationships are in the
hypothesized directions. Panels A, B, and C of Figure 3
show the touch frequency and purchase under different
crowding levels, within-group discussion levels, and
salesperson contact conditions, respectively. We observe an
inverted U-shaped relationship between crowding and touch
frequency, a somewhat linear declining trend of the effect of
crowding on purchase, and positive impacts of within-group
discussions and salesperson contact on both touch frequency
and purchases, which are consistent with our hypotheses.
Results and Analysis
The estimated proposed model fits the data well with the
log-marginal density of –1,596.41 (Chib and Greenberg
1995; Newton and Raftery 1994). For touch frequency, the
mean absolute error is .22 and the root mean square error is
.33 in the estimation sample. In the holdout sample, the
mean absolute error is .62 and the root mean square error is
1.05. For purchase decisions, the hit rate is 96.7% in the
estimation sample and 93.0% in the holdout sample.
To test for potential problems with multicollinearity, we
checked the correlations among the independent variables and
the variance inflation factors for the two equations (see the
Web Appendix). After dropping the TalkFreq ¥Accessories
variable (highly correlated with other discussion variables),
the highest correlation for the two equations is –.49, which
is low. Furthermore, the variance inflation factor is 4.17 for
the touch equation and 3.99 for the purchase equation, both
of which are much less than 10. Therefore, we conclude
that multicollinearity is not an issue in the study (Belsley,
Kuh, and Welsch 1980).
Estimation results from the proposed model appear in
Table 2, and significant estimates appear in bold (i.e., zero
does not lie in the 95% posterior probability interval). Table
3 summarizes the results of the hypothesis testing. Overall,
the hypotheses are largely supported.
Touch Frequency
The touch frequency model reveals that both noninteractive
(crowding) and interactive (within-group discussion and
salesperson contact) social factors have a significant impact
on shopper behavior. All of the hypotheses for touch fre-
quency (H1–H6) are supported, with parameter estimates
significant at the 95% probability level or higher and in the
expected direction (see Table 3).
Shopper density (crowding). The presence of people in
specific departments encourages shoppers to visit and touch

the merchandise in these departments (bi1 = .272). Shoppers
seem to be attracted to the more crowded regions of the
store, but there are decreasing returns for increased shopper
density (bi2 = –.030). The observed inverted U-shaped rela-
tionship between crowding and touch frequency (Figure 3,
Panel A) is consistent with H1a but departs from SIT’s pro-
posed power function (Latané 1981).
Although shoppers seem to be sensitive to the presence
of other patrons, this crowding effect is lower for larger
shopping groups (bi9 = –.049), as H4predicts. Shoppers
who visit the store with friends or family are less likely to
be drawn to crowded departments. Similarly, shoppers who
are engaged in conversations with companions are less
attracted to crowded departments (bi5 = –.076), presumably
because the focal shopper has closer relationships with
companions than with the strangers in the crowd.
We also observed a main effect of shopping group size
on touch frequency. As the size of the group increases, each
of the members is less likely to touch products (bi7 = –.234).
This effect may be a consequence of group heterogeneity.
As the size of the shopping party increases, members are
more likely to have divergent interests, and some may just
be “along for the ride,” not serious buyers.
As we expected, shoppers’ response to crowding varies
across product categories. The presence of other shoppers in
lead zones and clearance areas stimulates product interactions
(bi12new arrival = .162; bi12clearance = .213), but it discourages
people from shopping for accessories (bi12accessories = –.597).
This may be because shoppers are seeking information on
the quality and popularity of new arrivals in the lead zones
and the promotions in the clearance areas, and thus the
herding effect in these areas may be stronger. In contrast,
shoppers may require more private space to evaluate per-
sonal items such as fashion accessories and underwear.
Within-group discussion. The results also reveal that
shoppers are influenced by their direct interactions with
other people in the store. Shoppers who talk to their com-
panions during a store visit are likely to touch more prod-
ucts, as H2predicts (bi3 = .456). This effect was greatest in
the lead zones featuring the latest (and often most expen-
sive) new arrival items (bi13new arrival = .341), whereas dis-
cussions in the clearance section led to a lower incidence of
product touch (bi13clearance = –.216). This could be because
group discussions about new items generate more excite-
ment and interest, and therefore more touching, than discus-
sions about clearance items. As the size of the group
increases, these discussions have an even greater positive
impact on touch frequency (bi10 = .129), as H5predicts.
34 / Journal of Marketing, September 2014
TABLE 1
Zone-Level Summary Statistics
Class Variable (Definition) M (SD) Min (Max)
Dependent variables Touch frequency (number of times handling the product) .51 (.99) 0 (6)
Purchase (1 if shopper buys item; 0 otherwise) .07 (.25) 0 (1)
Social influence Crowding (number of people in a zone 30 seconds before the .70 (.98) 0 (6)
shopper enters)
Talk frequency (number of times a shopper talks to companions in .08 (.35) 0 (3)
the same shopping group)
Sales contact (1 if there is a shopper–salesperson interaction other .04 (.21) 0 (1)
than at checkout; 0 otherwise)
Group size (number of members in a shopping group) 2.30 (1.22) 1 (5)
Product category New arrival (1 if current zone is lead zone, featuring new arrivals; .33 (.47) 0 (1)
0 otherwise)
Accessories (1 if accessories zone, featuring fashion accessories; .10 (.31) 0 (1)
0 otherwise)
Clearance (1 if clearance zone featuring sale items; 0 otherwise) .25 (.43) 0 (1)
Marketing promotions Sidewalk sale (1 if sidewalk sale is on; 0 otherwise) .57 (.49) 0 (1)
In-store activities Talking on the phone (time spent by shopper in phone conversations, 1.76 (14.61) 0 (290)
in seconds)
Previous touch frequency in the same zone as the current visit t .47 (1.19) 0 (8)
(until time t – 1)
Cumulative touch frequency in other zones (until time t – 1) 1.96 (2.73) 0 (15)
Dressing room visit (1 if shopper visits dressing room; 0 otherwise) .02 (.15) 0 (1)
Shopping path Cumulative sinuosity (the weighted average turning angles of the 4.97 (3.08) 0 (17)
shopper’s path)
Distance covered (distance the shopper walks, in feet) 52.59 (64.89) 0 (729)
Gender appropriate (1 if a man is in the men’s section or a woman .67 (.47) 0 (1)
is in the women’s section; 0 otherwise)
Shopper motivation Initial speed (the speed with which a shopper walks into the store) 5.12 (3.09) .31 (16.14)
Shopping bag (1 if shopper is carrying a shopping bag when .28 (.45) 0 (1)
entering the store; 0 otherwise)

Social Influence on Shopper Behavior / 35
Salesperson contact. The findings indicate that retailers
can exert a direct influence on shopper behavior by encour-
aging sales associates to approach shoppers. When a sales-
person interacts with a customer, this leads to a higher fre-
quency of product touch (bi4 = 1.923), as H3a predicts.
Salespeople were most effective at increasing shopper
engagement for products in the clearance (bi14clearance =
.573), accessories (bi14accessories = .318), and new arrivals
sections of the store (βi14new arrival = .044), perhaps because
these items were less familiar to shoppers than the com-
modity goods (e.g., khakis, polo shirts, graphic tees) sold in
the center of the store.
However, the influence of salespeople on shopper
engagement is reduced by the presence of other people in the
store. When crowds are present, the salesperson’s interaction
with the customer has less impact on product touch (bi6 =
–.120), in support of H3b. Shoppers may be distracted by
other customers in the store and/or take cues from them about
which products are most desirable. Similarly, when shop-
pers are members of a larger shopping group, this reduces
the salesperson’s ability to encourage product touch (bi11 =
–.720), in support of H6. These findings are consistent with
SIT, which predicts that a source’s influence is reduced
when there are multiple competing targets of influence.
The demographics of the salesperson (gender, age, and
ethnicity) seem to moderate the impact of salesperson con-
tact. When the gender of the salesperson matches the shop-
per, this increases the shopper’s touch frequency (bi17 =
.512), consistent with Evans’s (1963) theory of buyer–seller
similarity. Surprisingly, ethnicity and age congruence had
the opposite effect (bi18 = –.498, bi19 = –.163). When sales
associates approached shoppers of the same ethnicity or
age, shoppers were less likely to touch products. Perhaps
the comments from the demographically similar sales asso-
ciate were more persuasive or relevant, reducing the shop-
per’s need for touch. In any case, a demographic match had
a positive impact on purchase likelihood, as we report in the
next section.
The findings suggest that there are behavioral clues that
the salesperson can use to identify shoppers who have a
higher need state and may be more responsive to a sales
intervention. Shoppers are likely to touch more often if they
walk into the store more slowly (b15entry speed = –.162), walk
in a straighter path (bi20sinu = –.072), and/or enter a zone
that is gender appropriate (i.e., men in the men’s section
and women in the women’s section; bi20gender app. = .613).
Store layout. One might expect that a sidewalk sale
would increase shopper engagement, stimulating customers
to enter the store and interact with the merchandise. How-
ever, we observed the opposite effect: the sidewalk sale was
associated with lower touch frequency inside the store (bi22 =
–.375). This could be due to several factors. Shoppers may
have browsed the sidewalk sale racks and decided there was
nothing of interest, so they never entered the store. Social
influence may have also played a role in this effect. If the
sidewalk sale reduced in-store shopper density, customers
may have been less attracted to the in-store merchandise.
It should be noted that product category has a significant
main effect on shoppers’ frequency of touching products.
0 1 2 3 4 5
1.6
1.4
1.2
1.0
.8
.6
.4
.2
0
Crowding
FIGURE 3
Touch Frequency and Purchase Under Different
Social Conditions
A: Crowding
B: Discussion Frequency
Mean touch frequency
Zone purchase likelihood
C: Salesperson Contact
0 1 2 3
3.5
3.0
2.5
2.0
1.5
1.0
.5
0
Talk Frequency
Mean Touch
Frequency
Zone Purchase
Likelihood
2.5
2.0
1.5
1.0
.5
0
No Yes
Mean touch frequency
Zone purchase likelihood

Customers touch products more often in accessories and
clearance than they touch new arrivals (bi8new arrival = –.463,
bi8clearance = .865, bi8accessories = 1.371), possibly because
the prices of accessories and clearance items are lower and
attract more attention and interest from the shopper com-
pared with the new arrivals and other items in the store.
Accessories may also be smaller and require closer, physi-
cal examination.
Purchase
We also observe support for most of the hypotheses involv-
ing product purchase (see Table 3). Parameter estimates were
significant at the 95% probability level and in the expected
direction.6
Shopper density (crowding). Although crowds can stimu-
late shopper interaction with products, they have the opposite
effect on purchase conversion rate. Shoppers are less likely
to buy merchandise when the store is crowded, consistent
with H1b (gi1 = –1.859). Crowds seem to interfere most with
36 / Journal of Marketing, September 2014
TABLE 2
Estimation Results: Proposed Model
Touch Equation Purchase Equation
Posterior 95% Posterior Posterior 95% Posterior
Variable Parameter Mean Interval Parameter Mean Interval
Intercept bi0 –1.631 (–1.660, –1.590) gi0 –1.633 (–2.698, –.850)
Crowding bi1 .272 (.240, .306) gi1 –1.859 (–3.105, –.011)
Crowding2 bi2 –.030 (–.057, –.008) gi2 .073 (–.273, .484)
Talk frequency bi3 .456 (.432, .475) gi3 .924 (.110, 1.755)
Sales contact bi4 1.923 (1.856, 1.979) gi4 1.110 (.297, 1.964)
Crowding ¥Talk frequency bi5 –.076 (–.122, –.024) gi5 –.438 (–1.512, .315)
Crowding ¥Sales contact bi6 –.120 (–.180, –.075) gi6 1.269 (.229, 2.335)
Group size bi7 –.234 (–.296, –.184) gi7 –.561 (–1.217, –.028)
Lead zone bi8-Lead –.463 (–.519, –.413) gi8-Lead –.139 (–1.015, .621)
Clearance bi8-Clear .865 (.827, .892) gi8-Clear 1.002 (.012, 1.797)
Accessories bi8-Accessory 1.371 (1.323, 1.421) gi8-Accessory .766 (.193, 1.648)
Crowding ¥Group size bi9 –.049 (–.070, –.029) gi9 –.021 (–.521, .432)
Talk frequency ¥Group size bi10 .129 (.086, .186) gi10 .777 (.269, 1.224)
Sales contact ¥Group size bi11 –.720 (–.776, –.680) gi11 –1.002 (–1.742, –.398)
Crowding ¥Lead zone bi12-Lead .162 (.140, .184) gi12-Lead .253 (–.419, .990)
Crowding ¥Clearance bi12-Clear .213 (.186, .241) gi12-Clear .472 (–.761, 1.971)
Crowding ¥Accessories bi12- Accessory –.597 (–.625, –.572) gi12-Accessory –.436 (–1.162, –.016)
Talk frequency ¥Lead zone bi13-Lead .341 (.316, .367) gi13-Lead –.490 (–1.290, .072)
Talk frequency ¥Clearance bi13-Clear –.216 (–.264, –.169) gi13-Clear –.396 (–1.510, .375)
Sales contact ¥Lead zone bi14-Lead .044 (.008, .083) gi14-Lead 1.180 (.173, 2.466)
Sales contact ¥Clearance bi14-Clear .573 (.507, .629) gi14-Clear 1.767 (.292, 3.070)
Sales contact ¥Accessory bi14-Accessory .318 (.293, .345) gi14-Accessory –.555 (–1.876, .667)
Speed of entry b15 –.162 (–.188, –.114) g15 –1.298 (–1.756, –.816)
Shopping bag b16 –.042 (–.109, .015) g16 .716 (.084, 1.627)
Gender congruence ¥ bi17 .512 (.491, .539) gi17 .259 (–1.001, 1.633)
Sales contact
Ethnicity congruence ¥ bi18 –.498 (–.525, –.477) gi18 .804 (.074, 1.432)
Sales contact
Age congruence ¥ bi19 –.163 (–.203, –.133) gi19 1.260 (.253, 2.528)
Sales contact
Cumulative sinuosity bi20-Sinu –.072 (–.123, –.012) gi20-Sinu .111 (–.232, .475)
Distance traveled bi20-Distance .001 (–.006, .009) gi20-Distance –.109 (–.172, –.074)
Gender appropriate bi20-Gen. app. .613 (.576, .640) gi20-Gen. app. .312 (–.531, 1.568)
Talking on the phone bi21 .012 (–.014, .045) — .— .—a
(in seconds)
Touch frequency — .— .— gi21 2.461 (1.747, 3.165)
Cumulative touch in the same — .— .— gi22 .029 (–1.509, .972)
zone until last visit
Cumulative touch frequency — .— .— gi23 –.363 (–.983, –.012)
in other zones
Dressing room visit — .— .— gi24 1.864 (.960, 2.711)
Sidewalk sale bi22 –.375 (–.409, –.350) gi25 –.663 (–1.689, .324)
aDropped because it predicted failure perfectly.
Notes: Boldfaced entries indicate that the 95% posterior probability interval excludes zero.
6From the estimated variance–covariance matrix S, we find that
the variance of eit is significant, with a value of .850. However, the
covariance between eit and dit is insignificant, indicating the insignifi-
cant correlation of the unobserved shocks in the touch and purchase
decisions after controlling for the impact of touch on purchases.

Social Influence on Shopper Behavior / 37
the purchase of accessories (gi12Accessories = –.436), perhaps
because of the personal nature of these products.
The size of the shopper group does not moderate the
negative crowding effect, contrary to the prediction of H4.
These findings run counter to SIT, which would predict that
the negative feelings created by crowds (reported by Argo,
Dahl, and Manchanda [2005] and others) would be diffused
by the shopper’s membership in a larger shopping group. At
a certain point, shopper density becomes so great that it
physically interferes with the customer’s ability to touch
and buy merchandise, independent of group size (see Figure
3, Panel A). Shopping group size did have a main effect on
purchase, mirroring its effect on touch, with larger shopping
groups having a lower likelihood of individual purchase (gi7 =
–.561). As noted previously, we suspect that larger groups
may have more browsers and fewer serious buyers.
Within-group discussion. When shoppers interact with
their friends and family in the shopping group, this
increases their purchase likelihood (gi3 = .924), and the
effect is magnified with larger shopping groups (gi10 =
.777). These findings support the predictions in H2and H5.
Salesperson contact. When a salesperson interacts with
a customer, this increases the shopper’s likelihood of buy-
ing, consistent with H3a (gi4 = 1.110). Sales contact had the
greatest positive impact on the purchase of new arrival and
clearance items (gi14new arrival = 1.180, gi14clearance = 1.767).
When the shopper is accompanied by friends and family,
this reduces the salesperson’s influence on purchase, as pre-
dicted by H6(gi11 = –1.002). Contrary to H3b, there was a
significant positive interaction between salesperson contact
and crowding. It seems that personal selling plays an even
greater role when the store is busy, helping mitigate the
negative effects of crowding on purchase (gi6 = 1.269).
Salespeople may help shoppers overcome the physical and
psychological barriers created by crowds, providing the
desired access to product information and inventory and
thereby “closing the sale.”
As we noted previously, the match of the salesperson’s
gender had a significant positive impact on the shopper’s
touch frequency, but this effect was not significant for pur-
chase. However, a match on ethnicity and age did increase
purchase likelihood (gi18 = .804, gi19 = 1.260). All of the
demographic congruence effects were in the positive direc-
tion, in support of Evans’s (1963) theory of buyer–seller
similarity.
Again, shoppers’ behavior provides clues about who is
the best prospect. Purchase likelihood is higher for shoppers
who walk more slowly (g15 = –1.298) and cover shorter dis-
tances (gi20distance = –.109). By far, the best predictors of
purchase are shoppers’ frequency of touching products dur-
ing their current visit to the department (gi21 = 2.461) and
whether shoppers visit the dressing room (gi24 = 1.864). In
contrast, when shoppers touch more products in other
zones, they are less likely to buy from the current zone (gi23 =
–.363), which indicates a competition effect across the
departments in the store.
Product category. Shoppers were more likely to buy from
clearance and accessories (gi8clearance = 1.002, gi8accessories =
.766), perhaps because these items were lower priced and
accessories were easy to try on and located near the check-
out counter. These categories were also most likely to be
touched.
Discussion
An important goal of retail promotion is to attract qualified
shoppers to the store to drive sales. Our findings suggest
that it can also be beneficial for stores to attract people who
are simply browsing because this will increase shopper den-
sity, stimulating other shoppers to take an interest in and
interact with the products. As Argo, Dahl, and Manchanda
(2005, p. 211) report, “no one likes to be alone in a retail
environment.” Prospective customers may feel more “at
home” in a store with other shoppers, infer product desir-
ability from their interaction with products (the herd effect),
and be tempted to model their behavior. Shoppers may also
be more comfortable shopping in a popular store because
the other patrons serve as a foil to deflect the advances of
aggressive salespeople. New arrivals and clearance items
seem to benefit the most from nearby shopper traffic, per-
TABLE 3
Hypothesis Testing Results
Stage Hypothesis Parameter Expected Sign Results
Touch frequency H1a (crowding) bi1 + Supported
H1a (crowding2) bi2 – Supported
H2(talking to fellow shoppers) bi3 + Supported
H3a (salesperson contact) bi4 + Supported
H3b (crowding × salesperson contact) bi6 – Supported
H4(crowding ¥group size) bi9 – Supported
H5(talk frequency ¥group size) bi10 + Supported
H6(salesperson contact ¥group size) bi11 – Supported
Purchase H1b (crowding) gi1 – Supported
H2(talking to fellow shoppers) gi3 + Supported
H3a (salesperson contact) gi4 + Supported
H3b (crowding ¥salesperson contact) gi6 – Reversed sign
H4(crowding ¥group size) gi9 + Not significant
H5(talk frequency ¥group size) gi10 + Supported
H6(salesperson contact ¥group size ) gi11 – Supported

haps because shoppers are uncertain about the appeal of
these items and are therefore more attentive to social cues.
Although there are compelling reasons to have a busy
store, the research findings indicate that too many cus-
tomers can also have a significant negative impact on the
shopping process. Specifically, we observed that as the
number of people in a specific zone or department
increased, shoppers were less likely to purchase products
from the department, especially for fashion accessories, and
this was the largest effect on purchase observed in the study
(Table 4). These results are consistent with previous find-
ings that perceived crowding has a negative impact on cus-
tomer satisfaction and purchase (Harrell, Hutt, and Ander-
son 1980; Hui and Bateson 1991). The presence of other
people in the store can hamper the shopping process by
reducing the shopper’s personal space, interfering with the
ability to try on merchandise, monopolizing the time of sales
associates, and increasing the wait time at checkout. This
can be a particular problem during the holiday shopping
season, when stores can become very crowded and cause
shoppers to leave without buying (see, e.g., Burke 2006).
A second noninteractive social factor is the size of the
shopping group, and this can also have a negative effect.
Shoppers who entered the store with a larger group were
less likely to touch products and less likely to buy. Shop-
pers’ frequency of touching products was even lower when
they were in a larger shopper group and the store was
crowded. The group size effect is consistent with a recent
study by Point of Purchase Advertising International and
ShopperSense (2011), which indicates that grocery shop-
pers with companions spend approximately 10% less per
trip than those who shop alone.
These findings illustrate the value of modeling shopping
behavior as a dynamic process, in which social influence
can have a differential effect on product engagement and
purchase over time as store conditions change and shoppers
move between different store zones, departments, and prod-
uct categories. The observed inverted U-shaped relationship
between crowding and customer touch frequency, the inter-
active influence of crowding and group size on touch, the
varying effects of shopper density across departments, and
crowding’s negative impact on purchase cannot be fully
explained by SIT. The results indicate that a shopper’s inter-
action with the merchandise and likelihood of purchase are
affected by the information conveyed by the activities of
other shoppers (e.g., herding and imitation effects) as well
as by the physical impact of crowding on the shopper’s
ability to touch products and make a purchase.
Whereas noninteractive social factors often had a nega-
tive impact, interactive social influence had a consistently
positive effect on shoppers’ interaction with the merchan-
dise and purchase. When shoppers talked with the other
members of their shopping party, this communication
increased their likelihood of touching and buying products.
Although it is not known what was said in these conversa-
tions, on average it seems that the content was positive.
Thus, “friends and family” events that encourage group dis-
cussion should help increase product interaction and stimu-
late sales. The size of this conversation effect increased
with the size of the shopping party for both touch and pur-
chase, in support of the prediction that interactive social
influence is greater when there are multiple sources of
influence, consistent with SIT.
In addition, we ran two mediation tests using the shop-
per’s walking speed as a proxy for his or her motivation
level (for details, see the Web Appendix). The results indi-
cate that walking speed mediates the influence of group
shopping on touch and purchase. Thus, group conversations
tended to slow shoppers down, encouraging them to stop,
touch, and purchase the merchandise. Group conversations
were also positively correlated with the distance the shopper
traveled in the store (p< .05), which suggests that this inter-
active social factor encourages greater store penetration.
Another important source of interactive social influence
is the sales staff. Salespeople play a key role in the shop-
ping process by encouraging shoppers to interact with prod-
ucts and complete their purchase transactions. Both of these
effects were statistically significant. Of the variables stud-
ied, sales contact had the greatest impact on touch fre-
quency and was one of the most important factors affecting
purchase (see Table 4). Sales interventions were particularly
effective when the salesperson was the same gender (for
touch) and the same ethnicity and age (for purchase) as the
customer. As with group conversations, we found that sales-
person conversations slowed shoppers down, and this medi-
ated the influence of sales contact on touch and purchase.
However, the influence of salesperson contact on shop-
pers’ interaction with and purchase of products was reduced
when shoppers were accompanied by friends and family.
Similarly, crowds seemed to reduce the salesperson’s effec-
tiveness in encouraging shoppers to touch merchandise.
Again, the findings are consistent with the predictions of
SIT: the impact of a sales intervention is reduced when
there are competing sources of social influence.
Table 4 indicates that salesperson contact and crowding
have the largest impact on shopper touch frequency and
purchase, respectively. This is contrary to the prediction of
SIT, which suggests that within-group discussion, as an
interactive social influence, will have greater strength and
thus a greater influence on behavior than salesperson con-
tact or crowding. One possible explanation is that the con-
tent of group discussions includes a mix of positive and
negative comments, and our measure of group discussion
frequency does not capture this heterogeneity. A potential
research opportunity is to record and code the number and
valence of these comments and include them in a more
comprehensive model of social influence.
The results of the present study reveal several behav-
ioral cues that signal when shoppers are in a particularly
high need state and may be good prospects for a salesperson
to offer assistance: shoppers who enter the store carrying
shopping bags, walk more slowly, walk in a straighter path,
enter a zone that is “gender appropriate,” and are shopping
alone.7Using these cues, salespeople may be able to
38 / Journal of Marketing, September 2014
7The relationship we observed between shopping bags and pur-
chase likelihood suggests that the shopping momentum effect
reported by Dhar, Huber, and Kahn (2007) may carry over from
one store to the next.

Social Influence on Shopper Behavior / 39
improve their productivity by allocating their selling effort
contingently. In a similar vein, Montgomery et al. (2004)
find, through the analysis of clickstream data, that a shop-
per’s online navigational path can signal his or her goals,
which can be helpful in predicting purchase conversion.
A benefit of the video tracking methodology is that it
enables the researcher to measure patterns of behavior at
both the store and department levels. We observed, for
example, that accessories (e.g., sunglasses, jewelry, belts)
were more frequently touched and purchased than other
products, perhaps benefiting from their easy trialability and
their close proximity to checkout. Clearance items were
also popular for bargain-hunting shoppers, despite being
located in the back of the store. Several of the social influ-
ence effects were contingent on the product category.
Higher shopper density encouraged shoppers to interact
with new and clearance items. In these categories, the
apparent popularity of the items may signal their desirabil-
ity. However, crowds had the opposite effect on shoppers’
interaction with accessories, reducing both touch and pur-
chase. These products may have a higher self-image com-
ponent, and shoppers may need more personal space to
evaluate them.
The video methodology also permits the coding of
shopper demographics and other physical and behavioral
characteristics, enabling researchers to explore how these
attributes might moderate the influence of social factors.
One would expect, for example, that some demographic
segments would be more comfortable than others shopping
in crowded environments or with groups of friends. To
investigate these issues, we expanded the touch and pur-
chase equations to include interactions between demo-
graphics (i.e., gender, age, and ethnicity), cell phone usage,
and the crowding and talk frequency variables. Although
the fit of this model was not quite as good as the original
model (log-marginal density of –1,634.89), and most inter-
actions were not significant, the analysis revealed a signifi-
cant interaction between crowding and gender. It seems that
women are less likely than men to make purchases in
crowded conditions.
The findings suggest that retailers need to manage the
levels of store traffic and shopper density carefully. Attrac-
tive store fronts and effective advertising and promotional
campaigns can entice shoppers to visit the store, but the in-
store environment must effectively convert this demand to
purchase. Some level of traffic is important to draw in cus-
tomers and encourage product interaction, but too much
traffic can hurt conversion rates. The key is to make the
store look busy but have sufficient resources (e.g., aisle
width, product inventory, sales staff, fitting rooms, cashiers)
to minimize the negative effects of crowding, giving shop-
pers sufficient opportunity to touch and try on products and
consult with salespeople.
Future Research Directions
Social impact theory, in combination with the video track-
ing data and hierarchical Bayes framework reported herein,
provide a rich set of insights about the influence of social
and other contextual factors on shopper engagement and
purchase in a retail setting. Although these initial findings
are enlightening, they could benefit from experimental vali-
dation with laboratory or field research. In particular, the
implications of the observed interactions between sales-
person contact and other social and environmental factors
could be tested by manipulating whether a salesperson
approached a shopper in various contexts (varying shopper
density, group size, product category, and demographic
similarity) and measuring the shopper’s response elasticity.
In addition to the demographic variables studied herein
(gender, age, and ethnicity), other physical attributes of the
salesperson and customer could be coded, such as general
attractiveness, style of dress, and similarity of appearance
(e.g., Argo, Dahl, and Morales 2008). With the shopper’s
consent, the actual comments of the salesperson could be
recorded, coded, and analyzed to determine their influence
on shopper behavior (see Burke and Leykin 2014).
The strong, lagged effects of shopper density on product
touch and purchase are also provocative and could be con-
firmed through experimental research. One approach would
TABLE 4
Standardized Coefficients for Social Influence
Variables
Variables Coefficients
Touch Frequency
Sales contact .402
Sales contact ¥group size –.291
Group size –.288
Crowding .269
Crowding ¥clearance .163
Crowding ¥accessories –.162
Talk frequency .161
Talk frequency ¥group size .144
Crowding ¥group size –.132
Crowding2
–.107
Crowding ¥new arrivals .099
Sales contact ¥clearance .073
Talk frequency ¥new arrivals .052
Talk frequency ¥clearance –.044
Crowding ¥talk frequency –.028
Sales contact ¥accessories .026
Crowding ¥sales contact –.023
Sales contact ¥new arrivals .005
Purchase
Crowding –7.305
Talk frequency ¥group size 3.437
Group size –2.744
Sales contact ¥group size –1.607
Crowding ¥clearance 1.429
Talk frequency 1.292
Crowding2 1.031
Sales contact .921
Sales contact ¥clearance .898
Crowding ¥talk frequency –.651
Crowding ¥new arrivals .613
Sales contact ¥new arrivals .491
Crowding ¥accessories –.471
Talk frequency ¥clearance –.318
Talk frequency ¥new arrivals –.297
Crowding ¥group size –.225
Sales contact ¥accessories –.183
Crowding ¥sales contact .173

be to have confederates shop in various store departments at
selected times and measure the immediate and delayed
impact on shopper touch frequency and purchase conver-
sion (see Argo, Dahl, and Manchanda 2005). It would also
be worthwhile to examine the impact of social influence
factors on the shopper’s other decisions throughout the
course of the store visit, including timing of walking up to a
product, consideration time, and touch duration, among oth-
ers. The findings will lead to more detailed and robust mod-
els of the social influence process.
It would also be valuable to track shopper behavior over
a longer period of time to measure the time course of these
effects. We observed all of the predicted main effects of
social influence on product touch and purchase during a
single store visit (Table 3), but the shopping process for
durable goods can extend across multiple shopping trips.
For apparel and other infrequently purchased products,
shoppers may visit the store several times before making a
purchase decision, and a conversation with a friendly sales
assistant on one shopping trip may lead to a large basket of
purchases on a future visit. By the same token, a shopper
who is discouraged from shopping due to weekend crowds
may return to buy on a weekday when the store is less busy.
Consequently, sales contact may have a greater impact on
purchase than crowding when observed over a sufficient
time period. As video tracking and other location-sensing
tools become more sophisticated, it may be possible to “rec-
ognize” shoppers who revisit stores and more accurately
estimate the effects of social influence and other contextual
variables on both short- and long-term retail performance.
40 / Journal of Marketing, September 2014
Cialdini, Robert B. and Noah J. Goldstein (2004), “Social Influ-
ence: Conformity and Compliance,” Annual Review of Psy-
chology, 55, 591–621.
Dahl, Darren W., Rajesh V. Manchanda, and Jennifer J. Argo
(2001), “Embarrassment in Consumer Purchase: The Roles of
Social Presence and Purchase Familiarity,” Journal of Con-
sumer Research, 28 (3), 473–81.
DeHerder, Richard and Richard Blatt (2011), Shopper Intimacy: A
Practical Guide to Leveraging Marketing Intelligence to Drive
Retail Success. Upper Saddle River, NJ: FT Press.
Dhar, Ravi, Joel Huber, and Uzma Khan (2007), “The Shopping
Momentum Effect,” Journal of Marketing Research, 44
(August), 370–78.
Dholakia, Utpal M., Suman Basuroy, and Kerry Soltysinski
(2002), “Auction or Agent (or Both)? A Study of Moderators of
the Herding Bias in Digital Auctions,” International Journal of
Research in Marketing, 19 (2), 115–30.
Eroglu, Sevgin A., Karen Machleit, and Terri F. Barr (2005), “Per-
ceived Retail Crowding and Shopping Satisfaction: the Role of
Shopping Values,” Journal of Business Research, 58 (8), 1146–
53.
Evans, F.B. (1963), “Selling as a Dyadic Relationship—A New
Approach,” American Behavioral Scientist, 6 (9), 76–79.
Farley, John U. and L. Winston Ring (1966), “A Stochastic Model
of Supermarket Traffic Flow,” Operations Research, 14 (4),
555–67.
Franke, George R. and Jeong-Eun Park (2006), “Salesperson
Adaptive Selling Behavior and Customer Orientation: A Meta-
Analysis,” Journal of Marketing Research, 43 (November),
693–702.
Goff, Brent G., Danny N. Bellenger, and Carrie Stojack (1994),
“Cues to Consumer Susceptibility to Salesperson Influence:
Implications for Adaptive Retail Selling,” Journal of Personal
Selling & Sales Management, 14 (2), 25–39.
——— and David L. Walters (1995), “Susceptibility to Sales-
person’s Influence and Consumers’ Shopping Orientations,”
Psychological Reports, 76 (3), 915–28.
Hanson, Ward and Daniel Putler (1996), “Hits and Misses: Herd
Behavior and Online Product Popularity,” Marketing Letters,7
(4), 297–305.
Harrell, Gilbert D., Michael D. Hutt, and James C. Anderson
(1980), “Path Analysis of Buyer Behavior Under Conditions of
Crowding,” Journal of Marketing Research, 17 (February),
45–51.
Hartmann, Wesley R. (2010), “Demand Estimation with Social
Interactions and the Implications for Targeted Marketing,”
Marketing Science, 29 (4), 585–601.
REFERENCES
Argo, Jennifer J., Darren W. Dahl, and Rajesh V. Manchanda
(2005), “The Influence of a Mere Social Presence in a Retail
Context,” Journal of Consumer Research, 32 (2), 207–212.
———, ———, and Andrea C. Morales (2006), “Consumer Conta-
mination: How Consumers React to Products Touched by Oth-
ers,” Journal of Marketing, 70 (April), 81–94.
———, ———, and ——— (2008), “Positive Consumer Contagion:
Responses to Attractive Others in a Retail Context,” Journal of
Marketing Research, 45 (December), 690–701.
Arnold, Mark J., Kristy E. Reynolds, Nichole Ponder, and Jason E.
Lueg (2005), “Customer Delight in a Retail Context: Investi-
gating Delightful and Terrible Shopping Experiences,” Journal
of Business Research, 58 (8), 1132–45.
Baker, Julie, Michael Levy, and Dhruv Grewal (1992), “An Experi-
mental Approach to Making Retail Store Environmental Deci-
sions,” Journal of Retailing, 68 (4), 445–60.
Baum, Andrew, R. Edward Harpin, and Stuart Valins (1975), “The
Role of Group Phenomena in the Experience of Crowding,”
Environment and Behavior, 7 (2), 185–98.
Becker, Gary S. (1991), “A Note on Restaurant Pricing and Other
Examples of Social Influence on Price,” Journal of Political
Economy, 99 (5), 1109–1116.
Bell, David R. and James M. Lattin (1998), “Shopper Behavior
and Consumer Preference for Store Price Format: Why ‘Large
Basket’ Shoppers Prefer EDLP,” Marketing Science, 17 (1),
66–88.
Belsley, David A., Edwin Kuh, and Roy E. Welsch (1980), Regres-
sion Diagnostics: Identifying Influential Data and Sources of
Collinearity. New York: John Wiley & Sons.
Breslow, N.E. (1984), “Extra Poisson Variation in Log-Linear
Models,” Applied Statistics, 33 (1), 38–44.
Burke, Raymond R. (2006), “The Third Wave of Marketing Intel-
ligence,” in Retailing in the 21st Century: Current and Future
Trends, Manfred Krafft and Murali Mantrala, eds. New York:
Springer, 113–25.
——— and Alex Leykin (2014), “Identifying the Drivers of Shop-
per Attention, Engagement, and Purchase,” in Review of Mar-
keting Research: Shopper Marketing and the Role of In-Store
Marketing, Vol. 11, Dhruv Grewal, Anne L. Roggeveen, and
Jens Norfalt, eds. Bingley, UK: Emerald Publishing.
Chen, Yubo, Qi Wang, and Jinhong Xie (2011), “Online Social
Interactions: A Natural Experiment on Word of Mouth Versus
Observational Learning,” Journal of Marketing Research, 48
(April), 238–54.
Chib, Siddhartha and Edward Greenberg (1995), “Understanding
the Metropolis Hastings Algorithm,” American Statistician, 49
(4), 327–35.

Social Influence on Shopper Behavior / 41
Hui, Michael K. and John E.G. Bateson (1991), “Perceived Con-
trol and the Effects of Crowding and Consumer Choice on the
Service Experience,” Journal of Consumer Research, 18 (2),
174–84.
Hui, Sam, Eric Bradlow, and Peter Fader (2009), “Testing Behav-
ioral Hypotheses Using an Integrated Model of Grocery Store
Shopping Path and Purchase Behavior,” Journal of Consumer
Research, 36 (3), 478–93.
———, Peter Fader, and Eric Bradlow (2009a), “Path Data in Mar-
keting: An Integrative Framework and Prospectus for Model
Building,” Marketing Science, 28 (2), 320–35.
———, ———, and ——— (2009b), “The Traveling Salesman Goes
Shopping: The Systematic Deviations of Grocery Paths from
TSP-Optimality,” Marketing Science, 28 (3), 566–72.
———, Yanliu Huang, Jacob Suher, and Jeffrey Inman (2013),
“Deconstructing the ‘First Moment of Truth’: Understanding
Unplanned Consideration and Purchase Conversion Using In-
Store Video Tracking,” Journal of Marketing Research, 50
(August), 445–62.
Inman, J. Jeffrey, Russell Winer, and Rosellina Ferraro (2009),
“The Interplay Among Category Characteristics, Customer
Characteristics, and Customer Activities on In-Store Decision
Making,” Journal of Marketing, 73 (September), 19–29.
Kurt, Didem, J. Jeffrey Inman, and Jennifer J. Argo (2011), “How
Shopping with Friends Promotes Consumer Spending,” Jour-
nal of Marketing Research, 38 (August), 741–54.
Lam, Shunyin, Mark Vandenbosch, John Hulland, and Michael
Pearce (2001), “Evaluating Promotions in Shopping Environ-
ments: Decomposing Sales Response into Attraction, Conver-
sion, and Spending Effects,” Marketing Science, 20 (2), 194–
215.
Latané, Bibb (1981), “The Psychology of Social Impact,” Ameri-
can Psychologist, 36 (4), 343–56.
Leibowitz, Josh (2010), “Rediscovering the Art of Selling,”
McKinsey Quarterly, 2, 117–19.
Luo, Xueming (2005), “How Does Shopping with Others Influ-
ence Impulsive Purchasing?” Journal of Consumer Psychol-
ogy, 15 (4), 288–94.
Milgram, Stanley (1970), “The Experience of Living in Cities,”
Science, 167 (3924), 1464–68.
Montgomery, Alan L., Shibo Li, Kannan Srinivasan, and John C.
Liechty (2004), “Modeling Online Browsing and Path Analysis
Using Clickstream Data,” Marketing Science, 23 (4), 579–95.
Nesbitt, Richard E. and T. Decamp Wilson (1977), “Telling More
Than We Can Know: Verbal Reports on Mental Processes,”
Psychological Review, 84 (3), 231–59.
Newton, Michael A. and Adrian E. Raftery (1994), “Approximate
Bayesian Inference with the Weighted Likelihood Bootstrap,”
Journal of the Royal Statistical Society (B), 56 (1), 3–48.
Nordfalt, Jens (2009), “Unplanned Grocery Purchases: The Influ-
ence of the Shopping-Trip Type Revisited,” Journal of Con-
sumer Behaviour, 8 (1), 1–13.
Paulus, Paul B. and Dinesh Nagar (1989), “Environmental Influ-
ences on Groups,” in Psychology of Group Influence, Paul B.
Paulus, ed. Hillsdale, NJ: Lawrence Erlbaum Associates, 111–
40.
Peck, Joann and Terry Childers (2003), “To Have and To Hold:
The Influence of Haptic Information on Product Judgments,”
Journal of Marketing, 67 (April), 35–48.
——— and ——— (2006), “If I Touch It I Have to Have It: Individ-
ual and Environmental Influences on Impulse Purchasing,”
Journal of Business Research, 59 (6), 765–69.
——— and Suzanne B. Shu (2009), “The Effect of Mere Touch on
Perceived Ownership,” Journal of Consumer Research, 36 (3),
434–47.
Pennington, Allan L. (1968), “Customer–Salesman Bargaining
Behavior in Retail Transactions,” Journal of Marketing
Research, 5 (August), 255–62.
Point of Purchase Advertising International and ShopperSense
(2011), “Influence in the Aisle: An Exploration and Analysis of
the Accompanied Shopper Versus the Lone Shopper,” Journal
of Marketing at Retail, 1 (1), 43–50.
Rossi, Peter, Robert McCulloch, and Greg Allenby (1996), “The
Value of Purchase History Data in Target Marketing,” Market-
ing Science, 15 (4), 321–40.
Sharma, Arun (2001), “Consumer Decision-Making, Salespeople’s
Adaptive Selling and Retail Performance,” Journal of Business
Research, 54 (2), 125–29.
Smith, Shaun and Joe Wheeler (2002), Managing the Customer
Experience: Turning Customers into Advocates. Harlow, UK:
Financial Times Prentice Hall.
Stilley, Karen, J. Jeffrey Inman, and Kirk Wakefield (2010),
“Spending on the Fly: Mental Budgets, Promotions, and
Spending Behavior,” Journal of Marketing, 74 (May), 34–47.
Stokols, Daniel (1972), “On the Distinction Between Density and
Crowding: Some Implications for Future Research,” Psycho-
logical Review, 79 (3), 275–77.
Underhill, Paco (1999), Why We Buy: the Science of Shopping.
New York: Simon & Schuster.
Von Riesen, R. Dale (1974), “Toward Staffing Optimality in Retail
Selling,” Journal of Retailing, 49 (4), 37–47.
Weitz, Barton A. (1981), “Effectiveness in Sales Interactions: A
Contingency Framework,” Journal of Marketing, 45 (January),
85–103.
———, Harish Sujan, and Mita Sujan (1986), “Knowledge, Moti-
vation, and Adaptive Behavior: A Framework for Improving
Selling Effectiveness,” Journal of Marketing, 50 (October),
174–91.
Willis, Mike (2008), “Shopping East and Shopping West—Chinese
Consumer Behavior in Two Worlds,” Journal of East-West
Business, 14 (3/4), 271–98.