Journal of Retailing and Consumer Services Information 2022

Contents lists available at ScienceDirect
Journal of Retailing and Consumer Services
journal homepage: www.elsevier.com/locate/jretconser
Multichannel personalization: Identifying consumer preferences for product
recommendations in advertisements across different media channels
Timo Schreiner, Alexandra Rese, Daniel Baier⁎
University of Bayreuth, Faculty of Law, Business and Economics, Chair of Marketing & Innovation, Universitätsstraße 30, 95447 Bayreuth, Germany
A R T I C L E I N F O
Keywords:
Personalization
Product recommendation
Multichannel
Avoidance of advertising
Digital natives
Gender differences
A B S T R A C T
Nowadays, many retailers use personalization in advertising to increase customers’ awareness and interest in their offers.
Product recommendations are a common form of personalization used in various communication channels. However,
previous studies have focussed on particular design aspects of product recommendations on a retailer’s website, without
considering other communication channels. Therefore, this study examines the ideal design of personalized product
recommendations in advertisements from a consumer’s perspective by relying on a choice-based conjoint experiment in
the apparel industry. The findings of two studies for young male (n = 170) and female (n = 162) consumers show that
the advertising channel is the most important attribute for determining the participant’s intentions to adopt the re-
spective product recommendations, followed by the number of recommendations. While banner advertising is the least
preferred channel for both samples, males prefer smaller recommendation sets than females. In addition to exploring
consumer preferences, the reasons for rejecting the advertisements are also analysed. Finally, design recommendations
for advertisers and retailers regarding personalized product recommendations are derived.
1. Introduction
During the past decade the advertising industry has faced major
changes due to the rapid and extensive diffusion of the internet. In ad-
dition to traditional media, online advertising such as display ads, search
engine or social media advertising are increasingly used for promotional
purposes (Danaher, 2017). As a result, consumers are constantly exposed
to a multitude of advertising messages in different online and offline
channels (Baek and Morimoto, 2012). Quite often the consumers’ reac-
tion to such ‘advertising clutter’ is characterized by behaviors of adver-
tising avoidance (Cho and Cheon, 2004; Ha and McCann, 2008). For
instance, 11% of the global Internet users already rely on ‘adblock’
technologies to suppress advertisements on websites (PageFair, 2017).
In order to increase advertising effectiveness, many advertisers make
use of personalization techniques (Bleier and Eisenbeiss, 2015a). Product
recommendations are a widespread type of personalization used in
ecommerce (Arora et al., 2008; Baier and Stüber, 2010; Kaptein and
Parvinen, 2015) and are often integrated directly on a company’s website
as well as in the email communication between a firm and its customers
(Linden et al., 2003). Personalized product recommendations have been
related to information technology from an early stage and charged with
the hope to “shift the focus of traditional mass advertising to more
concentrated and focused audiences” (Pavlou and Stewart, 2000, p. 67).
In addition, experiments have shown that personalized (postal) mailings,
e.g. identifying a “friend” as a sender, increased advertising response
rates (Howard and Kerin, 2004). While it has been expected and pre-
dicted for years that personalized product recommendations would be
used more frequently in offline advertising and retailing (Linden et al.,
2003), to date there have only been a few industry applications. As an
example, a German apparel retailer recently reported an increase of 25%
in its purchase order rates thanks to the implementation of personalized
product recommendations in package inserts (Borchers, 2016).
Due to a growing number of advertising media and almost stag-
nating advertising budgets within firms, it is also becoming increasingly
important for companies to make decisions about the allocation of their
advertising spend across different media channels (Danaher, 2017).
Therefore, it is crucial to investigate the effectiveness of advertising
messages and in particular of personalized advertising across different
media channels in order to derive recommendations for companies on
how to best approach customers.
Accordingly, the main purpose of this study is to examine the ideal
design of personalized product recommendations in advertisements in
terms of the preferred advertising channel, underlying recommender
algorithm as well as other design characteristics from a consumer’s
perspective. While previous studies solely deal with design issues for
product recommendations on a single communication channel, namely
https://doi.org/10.1016/j.jretconser.2019.02.010
Received 22 June 2018; Received in revised form 14 October 2018; Accepted 13 February 2019
⁎ Corresponding author.
E-mail address: [e-mail] (D. Baier).Journal of Retailing and Consumer Services 48 (2019) 87–990969-6989/ © 2019 Elsevier Ltd. All rights reserved. T

a retailer’s website (for a recent literature overview, see e.g.: Jugovac
and Jannach, 2017), the current research aims to investigate both, the
impact of different properties of product recommendations on the
consumers’ willingness to follow the recommendations and a compar-
ison of the effectiveness of three distinct online and offline advertising
channels: package inserts, email advertising and banner advertising.
Therefore, this study enhances and extends current research on perso-
nalization effects with regard to different advertising channels (Bues
et al., 2017; Cheung and To, 2017; Sahni et al., 2018), and provides an
initial connection between personalization research and research on the
ideal design of product recommendations. In addition, gender differ-
ences are taken into account. Previous research has consistently shown
that females are more involved with clothing than males (Millan and
Wright, 2018; Workman and Studak, 2006). This also holds with regard
to advertising involvement, e.g. paying attention to ads about clothing
(O’Cass, 2000) or gathering information before purchasing (Jackson
et al., 2011). However, shopping behaviour is changing over the gen-
erations e.g. with millennial males increasingly enjoying shopping
(Funches et al., 2017; Shephard et al., 2016). Research calls for a
“deeper understanding of each generation” (Funches et al., 2017, p.
101) as well as “understanding the mechanisms of change” (Shephard
et al., 2016, p. 5). Therefore, two separate studies – one for men and
one for women – are conducted and analysed in comparison.
The article is structured as follows: First, related literature regarding
personalization and its effects across different media channels, litera-
ture on product recommendations as well as literature addressing
gender-related issues with regard to fashion and advertising is elabo-
rated. Based on this literature review, several research hypotheses are
developed. Choice-based conjoint analysis (CBC) is used as a metho-
dological approach for data collection and utility estimation of different
attributes (and levels) identified with the help of the literature review.
In addition, the reasons for rejecting the advertisements are explored.
Finally, we present the empirical results, discuss their implications on
retailers and conclude our research by pointing out research limitations
and revealing possibilities for future research.
2. Literature review and hypotheses development
2.1. Personalization in advertising
While nowadays continuous and fast improvements in information
and communication technology increasingly enable companies to pro-
vide personalized product and service offerings (Rust and Huang,
2014), personalization has been applied in direct marketing for decades
(Vesanen, 2007). However, Fan and Poole (2006, p. 183) describe the
concept of personalization as “intuitive but also slippery”. This is be-
cause “(m)odern personalization seems to have different kinds of
meanings“ (Vesanen, 2007, p. 410). Therefore, both references criticize
the lack of a common definition.
While early definitions of personalization mainly refer to a context of
brick-and-mortar stores (Surprenant and Solomon, 1987), more recently
the term is often described in an online context. According to the lit-
erature, one possible goal of web personalization is the provision of the
right content to the right person at the right time (Ansari and Mela, 2003;
Tam and Ho, 2006). Other authors refer to personalization as the tai-
loring of certain marketing mix instruments to an individual based on
customer data (Arora et al., 2008; Chung and Wedel, 2014; Sundar and
Marathe, 2010). Following the latter definition, personalization is a form
of (firm-initiated) one-to-one marketing relying on “a target segment of
size one” (Arora et al., 2008, p. 306). Another form is 'customization'
which describes customer-initiated practices (Arora et al., 2008).
In the context of marketing communication, personalization relates
to advertising messages that are tailored to an individual’s preferences
and characteristics based on specific information about the respective
customer (Bang and Wojdynski, 2016; White et al., 2008; Yu and Cude,
2009). Various data sources can be used for personalizing advertising
messages, ranging from demographic characteristics and personally
identifiable information, e.g. “including the consumer’s name and her
place of work” (Sahni et al., 2018, p. 236), to “consumers’ most recent
shopping behaviors in the retailer’s online store” (Bleier and Eisenbeiss,
2015a, p. 670). Increasingly, companies deploy the concept of “re-
targeting” by providing personalized product recommendations in
banner advertisements based on their customers’ individual previous
browsing behaviour (Bleier and Eisenbeiss, 2015a; Lambrecht and
Tucker, 2013). Although current research mainly focuses on persona-
lized advertising in online channels, also different types of traditional
media, e.g. postal direct mail or telemarketing (Baek and Morimoto,
2012; Yu and Cude, 2009), can be used for delivering personalized
messages (Bang and Wojdynski, 2016).
2.2. Effects of personalization across different media channels
In the past, several researchers have dealt with personalization effects
on consumer behaviour related to brick-and-mortar stores (Goodwin and
Smith, 1990; Mittal and Lassar, 1996; Surprenant and Solomon, 1987),
and this field of research continues to receive much attention especially
in the context of the Internet and digital technologies (for a literature
overview: Salonen and Karjaluoto, 2016). In many cases, positive effects
of personalization on aspects of the customer relationship are found, such
as increased customer satisfaction and loyalty (Benlian, 2015; Ha and
Janda, 2014; Kim and Gambino, 2016; Kwon and Kim, 2012; Verhagen
et al., 2014; Yoon et al., 2013), greater purchase intentions (Ha and
Janda, 2014; Li and Liu, 2017; Pappas et al., 2014; Sahni et al., 2018),
enhanced click-through-rates for banner or email advertisements
(Aguirre et al., 2015; Bleier and Eisenbeiss, 2015a, 2015b; Sahni et al.,
2018; Tucker, 2014; Wattal et al., 2012) as well as more favourable at-
titudes towards the respective advert (Tran, 2017). However, some stu-
dies also report negative customer reactions to personalization such as
increased privacy concerns (Bleier and Eisenbeiss, 2015b; Song et al.,
2016), feelings of vulnerability (Aguirre et al., 2015), perceived intru-
siveness (van Doorn and Hoekstra, 2013) or even reactance (Bleier and
Eisenbeiss, 2015b; Puzakova et al., 2015).
While most of these studies consider personalization effects solely on a
single communication channel like banner or email advertising, only a few
studies exist that provide cross-channel comparisons of the impacts of
personalization. However, nowadays advertisers are increasingly forced
“to make tough decisions about how to allocate their ad budget across the
many possible media channels” (Danaher, 2017, p. 465). Those decisions
are an essential part and great challenge within ‘multichannel marketing’
which aims to provide customers with information, products, services or
support simultaneously in two or more synchronized channels (Ailawadi
and Farris, 2017; Rangaswamy and van Bruggen, 2005). Whereas in prior
literature the term ‘multichannel’ has been mainly used in the context of
retailing and referred to the “design, deployment, coordination, and eva-
luation of channels to enhance customer value through effective customer
acquisition, retention, and development” (Neslin et al., 2006, p. 96),
multichannel marketing rather incorporates evaluations of aspects like
“customer lifetime value, total spending across channels and cross-selling,
and dynamics among media” (Li and Kannan, 2014, p. 41). Following the
request of Verhoef et al. (2015) to explore “the effect of different mar-
keting mix instruments (i.e., promotions) used across touchpoints and
channels on the performance of channels” (Verhoef et al., 2015, p. 179),
more and more studies now also address issues regarding the attribution of
advertising spend in a multichannel context across different types of di-
gital media (Kireyev et al., 2016; Li and Kannan, 2014) as well as between
traditional and digital channels (Danaher and Dagger, 2013; Dinner et al.,
2014; Zantedeschi et al., 2016). Nevertheless, remarkably few studies do
so for personalized advertising.
The few studies that investigate the impacts of personalized ad-
vertising from a cross-channel approach (see Table 1) show that tra-
ditional print media such as direct mails or letters are perceived more
positively than digital media (Baek and Morimoto, 2012; Yu and Cude,
T. Schreiner, et al. Journal of Retailing and Consumer Services 48 (2019) 87–9988

Table 1
Studies on personalization with a cross-channel perspective.
Study Sample Ad Media Constructs/measurements Main results
Yu and Cude (2009) 192 US college students between 19
and 24 years old
Email, offline mail, phone call General perceptions, actual responses, attitude towards
advertiser, privacy concerns, purchase intentions
There were significant differences with regard to
purchase intentions between ad media.
Personalized offline mail was considered more
favourable.
Personalized advertising had a significantly negative
effect on purchase intentions.
Females evaluated personalized advertising more
negatively than males.
Baek and Morimoto
(2012)
442 US college students between 18
and 31 years, average age 20.4 years,
27.5% male and 72.5% female
Email, offline mail, phone call, wireless text message Ad avoidance - AAV (dependent), ad scepticism - ASK
(mediator), privacy concerns - PVC, ad irritation - IRR,
perceived personalization - PSL (antecedents)
PSL directly decreases ASK and AAV.
ASK has a partially mediating role.
The direct negative effect of PSL on AAV was highest
for emails, but lowest and not significant for wireless
text messages.
The direct negative effect of PSL on ASK was highest
for offline mail and wireless text messages.
ASK has a fully mediating role for PSL on AAV for
wireless text messages.
Shephard et al.
(2016)
408 US college student participants
(232 male and 176 female), 97.8%
between 18 and 29 years old
Mass media: television, billboard, store display, worn by
persons on television programs, in music videos.
Personalized media: catalogue, magazine, recommended
by a sales associate
Mass media, personalized media, fashion consciousness,
fashion leadership, traditional store patronage, non-
traditional store patronage
Mass media had a positive effect on fashion
consciousness regardless of gender.
The effect of personalized media on fashion
leadership was only positively significant for males.
Fashion leadership had a positive effect on non-
traditional over traditional retail channels for both
male and female consumers.
Present study Two distinct (gender-specific) samples:
170 male and 162 female German
college students, average age 21.9
years
Package inserts (offline), email, banner ads Part-worth utilities/importance via CBC/HB for different
levels of
For both samples, the advertising channel is the most
important attribute for the respondents’ intention to
adopt product recommendations.• Advertising channels
• Recommender algorithm of product
recommendations
• Explanation for the recommended products
• Number of recommendations
• Provider of advertisement,
Banner advertisements provide a comparably low
utility to female as well as male students.
Email advertising provides the greatest utility to
females and ads in package inserts are preferred most
by males.
While female respondents prefer a set of twelve
product recommendations, male participants favour
considerably smaller recommendation sets.
Reasons for rejecting product recommendations
Females do more often reject product
recommendations due to privacy concerns and a
minor recommendation quality.
T. Schreiner, et al. Journal of Retailing and Consumer Services 48 (2019) 87–9989

2009). For instance, in a comparative study of the customers’ percep-
tions towards personalized advertising in offline mail, email and tele-
phone advertising, Yu and Cude (2009) found that personalization is
generally perceived negatively, with personalized letters still being
perceived most positively. Their study revealed that the “respondents
showed comparably more favourable responses towards delivery via
offline mail than the other two types of media. They were less likely to
reject the mail immediately, more likely to take it seriously, less
threatened by the personalized advertisement as a violation of their
privacy, and somewhat more likely to view it as personal attention” (Yu
and Cude, 2009, p. 511). Furthermore, Baek and Morimoto (2012)
show that the relationship between perceived privacy concerns and
advertising avoidance is significantly weaker for direct mail than for
email advertising. Other studies into the advertising effectiveness of
different media unrelated to personalization confirm the supremacy of
advertising in traditional (print) media over digital communication
channels (Danaher and Dagger, 2013; Zantedeschi et al., 2016). While
Shephard et al. (2016) report positive influences of mass and perso-
nalized media on fashion consciousness and fashion leadership, both
factors were only significant for males.
At this point, our study is added to the list as it provides, on the one
hand, valuable insights into the customers’ preferences regarding per-
sonalized advertising across three different types of media, and on the
other hand also yields gender-specific insights.
With reference to research about the effects of personalization
across different media channels, customer preferences for product re-
commendations in three different advertising channels are investigated
in the study at hand: namely package inserts for traditional print media
and banner as well as email advertisements for digital media.
Based on prior findings we hypothesize:
H1a. Consumers prefer advertising in print media (package inserts) to
advertising in digital communication channels (email and banner
advertising).
With regard to the two digital communication channels, research
has shown that overall email advertising is more effective at influencing
sales compared to banner advertising with a weaker immediate, but a
much stronger long-term effect (Breuer et al., 2011). Investigating a
four-week advertising campaign across several media channels of an
Australian retailer and relying on members of the loyalty program for
respondents, Danaher and Dagger (2013) found that banner advertising
had no effect on sales in contrast to email advertising. According to the
results of this study “only 7% recalled having seen one or more of [the
retailer’s online display ads]” (Danaher and Dagger, 2013, p. 528). Li
and Kannan (2014) confirmed the enduring impact of emails compared
to banner ads. Based on these findings we hypothesize:
H1b. Consumers prefer email advertising to banner ads.
2.3. Identification of relevant design characteristics of product
recommendations for personalized advertising
Several empirical and literature-based, conceptual studies identify
various success factors of recommender systems (Jugovac and Jannach,
2017; Knijnenburg et al., 2012; Schafer et al., 2001; Xiao and Benbasat,
2007). Based on those studies, potential success factors of recommender
systems can be divided into different categories such as system-related
aspects or personal characteristics. We focus on system-related aspects
of product recommendations as these factors can be easily controlled
and modified by companies.
Product recommendations as a specific personalization tool can be
based on different recommendation sources ranging from user-gener-
ated content such as customer reviews to automatic recommender
systems (Lin, 2014; Senecal and Nantel, 2004). Since personalization
refers to firm-initiated practices, the focus is exclusively on re-
commender systems as a recommendation source.
The term ‘recommender system’ refers to any system that proposes a
personalized subset of interesting or useful objects from a large number
of options to a user (Burke, 2002). While recommender systems can
significantly improve the decision-making quality of consumers in
ecommerce and reduce information overload as well as search costs
(Xiao and Benbasat, 2007), the primary goal of recommender systems
from a firm’s point of view is an increase in product sales or conversion
rates (Aggarwal, 2016). Based on different underlying data sources such
as product ratings of other customers or a specific customer’s purchase
history, different recommendation techniques are distinguished (Burke,
2002; for an overview: Table 2).
Collaborative Filtering (CF) is by far the most widespread and popular
recommendation technique and Amazon’s item-to-item CF approach
(Linden et al., 2003) is the best-known and most influential example of CF
in ecommerce. In contrast to the user-based CF approach described in
Table 2, the item-based CF approach by Amazon generates its re-
commendations based on similar items, i.e. products that are often pur-
chased together (Linden et al., 2003). While content-based filtering ap-
proaches are also used quite frequently, hybrid recommender systems
nowadays represent the state of the art. For instance, Netflix uses a
combination of various algorithms to generate video recommendations.
Their recommender systems include e.g. not personalized recommenda-
tions from the most popular videos, recommendations based on similar
videos, as well as personalized recommendations based on the movie
genre (Gomez-Uribe and Hunt, 2016). By applying several recommenda-
tion techniques at a time, 80% of hours streamed at Netflix are triggered
by its own recommender systems (Gomez-Uribe and Hunt, 2016).
Previous research has shown that the success of recommender sys-
tems highly depends on the underlying recommender algorithm. In
particular, several studies confirmed for various domains such as video
clips and movies (Knijnenburg et al., 2010, 2012; Said et al., 2013) or
cultural events (Dooms et al., 2011) that personalized algorithms out-
perform random recommendations or recommendations of the gen-
erally most popular items from a customer’s point of view. Therefore,
we propose:
H2. Consumers prefer product recommendations generated by a CF
algorithm to product recommendations of bestselling products.
Beyond algorithms, but with reference to system-related aspects,
many recent studies identify several other aspects, e.g. the number of
product recommendations (Beierle et al., 2017; Bollen et al., 2010;
Ozok et al., 2010; Tam and Ho, 2005; Willemsen et al., 2016) or the
Table 2
Overview of recommendation techniques (Adomavicius and Tuzhilin, 2005; Burke, 2002; Ricci et al., 2015).
Technique Description/approach
Collaborative-filtering (CF) Recommendations are based on product ratings or purchases of other users with similar profiles.
Content-based filtering Recommendations are based on specific product features (e.g. colour, form) that were included in previously preferred items.
Demographic filtering Recommendations are based on the sociodemographic user profiles.
Knowledge-based/utility-based filtering Recommendations are based on specific domain knowledge about how a particular item meets a specific user need.
Community-based/social Recommendations are based on the preferences of a user’s friends.
Hybrid methods Combination of two or more of the previous methods in order to compensate for the disadvantages of a single technique by making use
of the advantages of another technique.
T. Schreiner, et al. Journal of Retailing and Consumer Services 48 (2019) 87–9990