PreprintPDF Available
Preprints and early-stage research may not have been peer reviewed yet.

Abstract and Figures

This study investigates the effect of call-to-action (CTA) direct mailings (DMs) on customers' purchase behavior and how this effect is moderated by incentive type as well as customer characteristics. The authors analyze individual purchase behavior of a panel of 179,525 customers across 40 Dutch optical retailers over 9 years. The empirical results show that CTA DMs that include an incentive have a higher positive impact on customer's purchase probability compared with those without an incentive. Furthermore, non-monetary incentives, especially utilitarian ones, have a higher positive impact than monetary incentives. Our results also show that customer heterogeneity plays an important role in the influence of CTA DMs on purchase incidence. More specifically, CTA DMs have a higher impact on purchase incidence for customers with higher past purchase frequency, lower purchase recency, longer relationship duration and for customers who received DM more recently and frequently. The results of the study provide valuable insights for managers allocating their direct marketing budget: it is better to use non-monetary (preferably utilitarian) incentives and to target customers based on relationship history as well as DM history.
Content may be subject to copyright.
Calling Customers to Take Action: The Impact of Incentive and
Customer Characteristics on Direct Mailing Effectiveness
Saeid Vafainia
a,
& Els Breugelmans
a
& Tammo Bijmolt
b
a
KU Leuven - Faculty of Economics and Business, Korte Nieuwstraat 33, B-2000 Antwerp, Belgium
b
University of Groningen - Faculty of Economics and Business, Nettelbosje 2, NL-9747AE Groningen, the Netherlands
Abstract
This study investigates the effect of call-to-action (CTA) direct mailings (DMs) on customers' purchase behavior and how this effect is
moderated by incentive type as well as customer characteristics. The authors analyze individual purchase behavior of a panel of 179,525 customers
across 40 Dutch optical retailers over 9 years. The empirical results show that CTA DMs that include an incentive have a higher positive impact on
customer's purchase probability compared with those without an incentive. Furthermore, non-monetary incentives, especially utilitarian ones, have
a higher positive impact than monetary incentives. Our results also show that customer heterogeneity plays an important role in the inuence of
CTA DMs on purchase incidence. More specically, CTA DMs have a higher impact on purchase incidence for customers with higher past
purchase frequency, lower purchase recency, longer relationship duration and for customers who received DM more recently and frequently. The
results of the study provide valuable insights for managers allocating their direct marketing budget: it is better to use non-monetary (preferably
utilitarian) incentives and to target customers based on relationship history as well as DM history.
© 2018
Keywords: Call-to-action direct mailings; Monetary incentives; Non-monetary incentives; Customer heterogeneity; Retailing; Purchase incidence
Introduction
For decades, direct mailing has been the workhorse of direct
marketing. Direct mailing is one of the most popular marketing
communication channels, accounting for 19% of all direct
marketing expenditures in the US, standing higher than any
other direct marketing channel (Direct Marketing Association
2015). A large portion of these mailings are call-to-action direct
mailings (hereafter abbreviated as CTA DMs), that aim to
directly influence customer buying behavior, e.g., visit a store
or make a purchase, oftentimes by providing some kind of
incentive (Prins and Verhoef 2007; Rust and Verhoef 2005;
Verhoef 2003). In today's marketplace, retailers frequently
communicate via such CTA DMs, while customer's response
rate is quite low, approximately 5% on average for such DMs
(Direct Marketing Association 2017). Hence, it is crucial for the
retailers to understand how to effectively influence a customer's
purchasing decision via DM communications.
CTA DMs can be divided based on the type of incentive.
DMs can use either monetary or non-monetary incentives.
While monetary incentives are typically price discounts, the
most frequently used type of non-monetary incentives are
premiums; a product or service for free or at a relatively low
price (D'Astous and Jacob 2002; Palazon and Delgado-
Ballester 2013). Among the non-monetary incentives, a
distinction can be made between DMs including a hedonic
versus a utilitarian premium. According to Palazon and
Delgado-Ballester (Palazon and Delgado-Ballester 2013),
hedonic incentives may provide more experiential consump-
tion, fun, enjoyment, pleasure and excitement, whereas
utilitarian premiums are primarily instrumental and functional.
Sometimes retailers invite customers via CTA DMs to visit the
store and trigger them to make a purchase, without explicitly
rewarding customers for taking this action. The sales promotion
literature has shown that different promotional incentives have
Corresponding author.
E-mail address: Saeid.Vafainia@KULeuven.be (S. Vafainia).
www.elsevier.com/locate/intmar
https://doi.org/10.1016/j.intmar.2018.11.003
1094-9968© 2018
Available online at www.sciencedirect.com
ScienceDirect
Journal of Interactive Marketing 45 (2019) 62 80
different mechanisms to influence customers' buying behavior
(Foubert et al. 2017; Palazon and Delgado-Ballester 2009).
While there is some research done in a sales promotion context,
at this moment it is largely unknown whether and how these
DM incentives have a differential effect on the probability of
customers responding to the CTA DM.
Previous studies have demonstrated that customer's response
to direct marketing is highly heterogonous and have called for
further research on the moderating effects of customer and
relationship characteristics (Gázquez-Abad et al. 2011; Rust and
Verhoef 2005). More specifically, retailers can differentiate their
DM communications based on customer differences in relation-
ship history (such as recency, frequency and monetary value
(RFM) measures), communication history (how recently and
frequently customers have received DMs) and socio-demo-
graphics (age and gender) (Neslin et al. 2013; Rust and Verhoef
2005). While past research has investigated some of these
customer characteristics (e.g., Gázquez-Abad et al. 2011; Rust
and Verhoef 2005), we are the first to investigate an extensive set
of customer characteristics on the impact of CTA DMs on
customer purchase behavior and thereby provide additional
insights on the moderating role of these characteristics.
In summary, we investigate (1) the differential impact of
various CTA DM incentives on purchase incidence; and (2)
whether and how the influence of a CTA DM on purchase
incidence is moderated by a customer's relationship history,
DM history and socio-demographic characteristics. Table 1
shows most relevant previous studies and highlights the
contribution of this paper relative to the existing literature.
Our findings shed more light on the conditions under which
CTA DM's effectiveness can be enhanced and, hence, will be
useful for managers in shaping their direct marketing strategy.
As the Table 1 illustrates, we have a unique stance in the
literature, as we are the first to investigate the impact of the
incentive type as well as a large set of consumer characteristics,
in the context of CTA DMs.
In pursuit of our research objectives, we shaped a conceptual
framework (Fig. 1) that includes: (i) CTA DM as the main
antecedent, (ii) type of incentive (no incentive, monetary, non-
monetary hedonic, non-monetary utilitarian), and (iii) relation-
ship history, DM history and socio-demographic characteristics
as potential moderating variables. We use a probit model in
which we control for the endogeneity of DMs, following (Van
Diepen, Donkers, and Frances 2009). We use a purchase and
DM database of a panel of 179,525 customers across 40 optical
retailers in the Netherlands from 2008 to 2016 in order to
answer the research questions.
We organize the remainder of this article as follows. We
begin with a discussion of our conceptual framework and
hypotheses. Subsequently, we discuss our data and model.
Then, we present the empirical results pertaining to the impact
of CTA DM on purchase incidence and several moderators. We
conclude with a discussion of the findings, implications, and
limitations.
Table 1
Literature overview.
Paper Context Focal construct/objective Outcome measure Endogeneity
correction
Moderators
Incentive
type
Relationship
history
DM
history
Socio-
demos
Thomas, Feng,
and Krishnan (2015)
Non-profit Impact of CTA DM Donation incidence Y N N N N
Feld et al. (2013) Financial service and
non-profit
Impact of CTA DM
design characteristics
Opening and keeping
rate
NNNNN
Gázquez-Abad
et al. (2011)
Retailing Impact of CTA
vs. relational DM
Purchase incidence,
amount, number of
visits and products
purchased
N N [Y]
a
[Y] N
Van Diepen, Donkers,
and Franses (2009)
Non-profit Impact of own and
competitive CTA DM
Donation incidence Y N [Y] N N
Donkers et al. (2006) Non-profit Optimization of CTA
DMs
Donation incidence
and amount
YNNNN
Gönül and Ter Hofstede
Hofstede (2006)
DM service company Impact of frequency
and timing of CTA
DM
Donation incidence
and amount
NNNNN
Rust and Verhoef (2005) Financial service Optimization of marketing
interventions including DMs
Profit N N Y N Y
Verhoef (2003) Financial service Impact of CTA DM vs. mass
communication
Change in customer
share
NNNNN
Bawa and Shoemaker
1987)
Retailing Impact of CTA DM Brand choice
probability
NNNNN
Our study Retailing Impact of CTA DM Purchase incidence Y Y Y Y Y
a
We indicate in brackets when a characteristic was included as a main effect in the paper, yet was not investigated as a moderator of DM effectiveness.
63S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Conceptual Framework and Hypotheses
Main Effect of CTA DMs
CTA DMs focus mainly on directly influencing customer's
buying behavior, for instance by providing an incentive (call)
that triggers customer responses (action)(Prins and Verhoef
2007; Rust and Verhoef 2005) or, in cases where such
incentives are not included, by providing customers a (textual)
trigger (call) to respond (act) now. The focal aim of such DMs
is to encourage customers to react within a short period of time,
by including a specific call to actionthat stimulates them, for
instance, to visit the store and make a purchase.
According to reciprocal action theory, direct marketing
communications positively influence repurchase, because firm
investments in customer relationships result in psychological
bonding, as customers perceive a greater resource investment
by the firm. Consequently, customers feel obliged to return
good for goodby making a purchase (Bagozzi 1995; De
Wulf and Odekerken-Schröder 2003; Godfrey, Seiders, and
Voss (2011)). Moreover, information directed toward cus-
tomers to reduce the effort required as well as the uncertainty
surrounding a decision increases the confidence associated with
the decision and is viewed positively by customers (Fitzsimons
and Lehmann 2004). Previous studies have confirmed this
positive CTA DM effect on a variety of behavioral measures
(e.g., purchase incidence, frequency and amount, (Gázquez-
Abad et al. 2011); customer retention, (Rust and Verhoef
2005); customer retention and lifetime value, Risselada,
Verhoef, and Bijmolt (2014).
On the other hand, CTA DM may also work negatively,
following reactance theory (Brehm 1966) that posits that
customers may resist marketing efforts perceived as an effort to
manipulate them, exercise control over their purchasing, or limit
their freedom of choice (Godfrey, Seiders, and Voss (2011)). A
CTA DM may be seen in such a way and may therefore diminish
or, at extremes, backlash on customer purchase behavior
(Fitzsimons and Lehmann 2004; Godfrey, Seiders, and Voss
(2011)). Several studies have reported a negative effect of
unsolicited marketing communications (Fitzsimons and
Lehmann 2004; Gázquez-Abad et al. 2011; Simonson, Carmon,
and Curry (1994)). Given the prevalence of arguments in both
directions, the following hypothesis is proposed:
H
1
. : A CTA DM has a (a) positive, or (b) negative effect on
purchase incidence.
Moderators
In the first subsection below, we focus on the difference in
effectiveness between a DM with vs. without an incentive.
Next, we differentiate between monetary and non-monetary
incentives, then split the latter in hedonic vs. utilitarian. In the
second subsection, we discuss the moderating impact of
customer characteristics (relationship history, DM history and
socio-demographics).
The Moderating Role of CTA DM Incentive
CTA DM with vs. without an incentive. Building on reactance
theory, unsolicited CTA DMs with incentive may be perceived
by customers as an attempt to manipulate their choices and
therefore, may lead to reactance, where customers try to
punishthe offending marketers Simonson, Carmon, and
Curry (1994). Conversely and relying on reciprocal action
theory, CTA DMs that use an incentive can also be conceived
by customers as an extra effort and resource investment. As a
result, customers may have a higher likelihood to reciprocate
the retailer's action by making a purchase. Empirical research
has found some support for the latter: (Bawa and Shoemaker
1987) found that marketing communications that provide
customer with an incentive lead to higher customer shares and
additional sales (Bawa and Shoemaker 1987). In a similar vein,
(Arora and Stoner 1992) also found that response rates are
Incentive Types
Incentive vs. No incentive
Monetary vs. Non-monetary
Hedonic vs. Utilitarian
Call-to-action DM Purchase Incidence
Customer Characteristics
Relationship history
DM history
Socio-demographics
Fig. 1. Conceptual Framework.
64 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
higher for DMs with an incentive compared to those without.
Finally, the customer may value the incentive itself, which also
suggests a larger effect of the CTA DM with an incentive. Still,
as we cannot rule out the reactance effect a priori, we formulate
opposing hypotheses:
H
2
. : A CTA DM with an incentive has a (a) lower, or (b)
higher impact on purchase incidence compared to ones without
an incentive.
CTA DM with monetary vs. non-monetary incentive. The
incentives of CTA DMs can be framed as monetary, where
customers receive a price reduction or percentage off, or non-
monetary, where customers receive a gift when making a
purchase of a specific product in the store. This difference
between incentive types has been shown to require different
levels of information processing (Nunes and Park 2003).
According to (Sinha and Smith 2000), customers perceive
monetary incentives as a reduced loss because they are in the
same metric as the product prices. Non-monetary incentives are
perceived as separate gains due to their incommensurate nature
(Nunes and Park 2003). According to Diamond and Sanyal
(Diamond and Sanyal 1990), a loss reduction tends to have a
relatively small impact, which causes customers to prefer
promotions that are framed as gains, leading to non-monetary
incentives outperforming monetary ones. On the contrary,
many studies stated that a non-monetary incentive might not be
as effective as a monetary incentive (Foubert et al. 2017). First,
a non-monetary incentive is more likely to be perceived as a
marketing gimmick to boost sales Simonson, Carmon, and
Curry (1994). Customers may attribute this to the brand's low
quality and the fear of disappointing sales (Blattberg and Neslin
1990). As a result, customers may feel manipulated and may
therefore refuse to act in accordance with the suggested
behavior (Brehm 1966). Second, Thaler (Thaler 1985) argues
that monetary incentives enable customers to hedge against
uncertainty about future needs and product performance.
Money saved from a discount is completely fungible and can
be allocated flexibly, while the benefits of non-monetary
rewards are fixed (Heilman, Nakamoto, and Rao (2002);
Leclerc 1997). Finally, non-monetary rewards can be seen as
a form of indulgence and the purchase may therefore be hard to
justify. When taking advantage of monetary rewards, customers
are perceived as smart shoppers Gedenk, Neslin, and Ailawadi
(2010); Simonson, Carmon, and Curry (1994). As we have
arguments for both sides, we pose the following hypotheses:
H
3
. : A CTA DM with a non-monetary incentive has a (a)
higher, or (b) lower impact on purchase incidence compared to
a monetary CTA DM.
CTA DM with a hedonic vs. utilitarian non-monetary
incentive. Research has shown there is a fundamental
difference between goals that hedonic and utilitarian non-
monetary incentives pursue (e.g., (Chitturi, Raghunathan, and
Mahajan (2007); Palazon and Delgado-Ballester 2013)). While
hedonic incentives tend to provide more experiential consump-
tion, fun, enjoyment, pleasure and excitement, they are also
more difficult to be justified because hedonic consumption can
arouse guilty feelings (Palazon and Delgado-Ballester 2013).
On the other hand, building on justification-based theory
developed by (Okada 2005), customers may prefer hedonic
premiums when they would see the CTA DM as a promotional
offer, because the justification of choice is easier in a
promotional context. In such a context, the sense of guilt is
mitigated as the benefit is offered for free (Palazon and
Delgado-Ballester 2013). Hence, we propose that:
H
4
. : A CTA DM with a hedonic non-monetary incentive has a
(a) lower, or (b) higher impact on purchase incidence compared
to a CTA DM with a utilitarian non-monetary incentive.
The Moderating Role of Relationship History, DM History and
Socio-Demographics
Following the direct marketing and CRM literature (e.g.,
(Reinartz and Kumar 2003; Rust and Verhoef 2005; Verhoef
2003)), we focus on three types of customer characteristics that
may moderate the impact of CTA DMs on purchase behavior:
(i) characteristics that relate to the relationship and prior
behavior of customers exposed in the company (relationship
duration, purchase frequency, and recency of purchase); (ii)
characteristics that relate to the retailer's prior DM communi-
cation with customers (DM recency and frequency) and (iii)
socio-demographics (age and gender). We will discuss each of
these variables below.
Impact of purchase frequency. Customer behavior models
rooted in learning theories suggest that customers, in general,
like to simplify their decisions by reducing the complexity of
the decision-making situation (Jayachandran et al. 2005; Sheth
and Parvatlyar 1995). Customers who purchase more fre-
quently, or so-called heavy users, have gained more experi-
ences that they can store, process, and retrieve to use in
subsequent purchase decision situations. They also tend to have
developed an ability to more easily distinguish among stimuli
they may receive (Sheth and Parvatlyar 1995). Thus, heavy
users have higher capabilities in terms of understanding and
engaging with firm contacts compared to less experienced
users. Relationship marketing literature has also demonstrated
that customers with higher purchase frequency are more likely
to respond in a positive way to a DM (Fader and Hardie 2007;
Liu 2007). Therefore, we propose:
H
5
. : The impact of a CTA DM on purchase incidence is higher
for customers with higher purchase frequency.
Impact of purchase recency. If a customer has not purchased
for a long time, then it is likely that s/he has stopped being a
customer. For these customers, a CTA DM may be perceived as
a persuasion attempt by the retailer to influence their purchasing
decisions and the reactance response for this group of customers
may grow stronger (Fitzsimons and Lehmann 2004). There is
some empirical evidence that indicates that customers with
higher recency (longer time since last purchase) are indeed less
likely to respond to marketing efforts (e.g., Bult et al. 1997;
65S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Neslin et al. 2013;Van Diepen, Donkers, and Franses (2009)).
Thus, we propose the following:
H
6
. : The impact of a CTA DM on purchase incidence is lower
for customers with a longer time since last purchase.
Impact of relationship duration. Customers in the early
phases of a relationship do not yet have much experience
with the company. As the duration of the relationship increases,
the level of intimacy also increases, leading to an increased
richness of the customer's impressions about the firm (Swann
and Gill 1997; Verhoef, Franses, and Hoekstra (2002)).
Therefore, a customer with higher relationship duration has a
better evaluation of a new received DM and might engage more
with it. Empirical findings in the relationship marketing
literature support this reasoning Verhoef, Franses, and
Hoekstra (2002). Hence, we hypothesize that:
H
7
. : The impact of a CTA DM on purchase incidence is higher
for customers with a longer relationship duration with the retailer.
Impact of DM frequency. Communication frequency has an
important impact on a customer's reaction to future contacts
(Drèze and Bonfrer 2008; Neslin et al. 2013). Receiving many
DM communications may trigger reactance against the repeated
message (Campbell and Keller 2003). In extremes cases,
customers with high levels of DM frequency in the past may
develop negative attitudes and defensive strategies with regard
to yet another DM from that company as a result of irritation
(Ansari, Mela, and Neslin 2008). Conversely, information in
memory decays and fades away more slowly, when customers
are confronted with a message again and again (Sheth and
Parvatlyar 1995). Consequently, for the customers who
frequently receive communications, information processing is
simplified through repetition of a (similar) message. Given the
prevalence of arguments in favor of both sides, we hypothesize:
H
8
. : The impact of a CTA DM on purchase incidence is (a)
lower, or (b) higher for customers with a higher DM frequency.
Impact of DM recency. Not only does prior communication
frequency impacts customers' reactions to a new DM, also the
recency of the previous communication has been shown to play
an important moderating role (Drèze and Bonfrer 2008; Neslin
et al. 2013). The moderating impact of DM recency on the
effect of CTA DMs on purchase incidence can also be predicted
by saturation and information overload mechanisms on the one
hand, and the role of memory and repetition of the message on
the other hand. According to Van Diepen, Donkers, and
Franses (2009), after receiving CTA DMs in short time
intervals, additional DMs send to customers will be ignored
because of DM clutter (Van Diepen, Donkers, and Frances
2009). However, relying on the advertising literature, receiving
messages with shorter intervals may enhance the impact of the
message on the memory (Sheth and Parvatlyar 1995), resulting,
in our case, in increased effects of a CTA DM for customers
who have received other DMs more recently. Hence we have a
two-sided hypothesis:
H
9
. : The impact of a CTA DM on purchase incidence is (a)
higher, or (b) lower for customers with a longer time since they
received the last CTA DM.
Impact of socio-demographics. Research comparing young
and older customers has concentrated on differences in the
information-processing abilities needed to evaluate a product
(Smith and Baltes 1990). Most previous studies conclude that
information processing declines with age (Gilly and Zeithaml
1985). Older people have lower information-processing capabil-
ities; and therefore, their reactions to marketing communications
might also decline. Yet, reactance theory indicates that young
people possess higher reactance than older people, mainly because
older people are more prepared to manage several dimensions of
reactance (Brehm 1966). Hence, younger customers may exhibit a
negative attitude toward a received CTA DM because they
perceive it as an attempt of retailers to limit their freedom of
choice. The research by (Simpson and Mortimore 2015)shows
that as age increases, favorable perceptions for DM increase and
(Barnes and Peters 1982) revealed that the elderly tend to have
more positive attitudes toward DMs. Hence:
H
10
. : The impact of a CTA DM on purchase incidence is (a)
higher, or (b) lower for younger customers.
Similar to age, gender often appears as a moderator variable
in marketing studies (Saad and Gill 2000). Drèze and Bonfrer
(Drèze and Bonfrer 2008) argue that a more positive attitude for
females may be due to higher female involvement in the
purchase process. Therefore, we predict:
H
11
. : The impact of a CTA DM on purchase incidence is
lower for males than for females.
Methodology
Empirical Setting and Data
To test our hypotheses, we examine customer purchase
behavior of a large group of 40 optical retailers in the Netherlands.
These optical retailers are independent and do not belong to large
chains. There is also no overlap in their trading area, so customers
in our dataset are uniquely linked to one retailer. A marketing
consultancy firm coordinates the customer databases for these
retailers and provided us the data for this project. Given the high
involvement and low purchase frequency of consumer durable
products such as optical product categories (Mittal 1989), the
main objective of the DMs is to help retailers to attract the
attention of existing customers with their offers and trigger them
to make a purchase. The usage of CTA DMs to influence purchase
behavior of existing customers is wide-spread, and extends well
beyond the setting of optical retailing (see Table 1 for an overview
of other retail sectors where CTA DMs are widely used). We
obtained a unique dataset with information on purchase as well as
DM history (for instance, timingand details of the CTA DMs sent
by the retailers) of all customers for these 40 optical retailers
across nine years (20082016).
For our analyses, we use all customers who (1) made at least
two purchases within the observation period (to rule out trial
66 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
customers that make just one single purchase in nine years), (2)
received at least one CTA DM (to make sure we focus on
customers that pass the selection criteria of receiving such a CTA
DM), and (3) are 18+ (to exclude children) and have valid
information on age, gender and registration date (to make sure
we can operationalize the independent variables). We believe
that our sample selection criteria are very lenient and reasonable,
because we use an observation window of 9 years, and the focus
of this paper is on the impact of CTA DMs and the moderation
effect of incentive types and customer characteristics. These
constraints leave us with 179,525 customers across the 40
retailers. The smallest retailer has 906 customers, while the
largest one has 12,915 customers. CTA DMs have been sent to
customers on a quarterly basis and each retailer sends CTA DMs
to the customers registered in their database (the marketing
consultancy firm helps them with doing this, so that we can be
confident that CTA DMs are designed and send in an identical
way). On average, customers receive 3.6 CTA DMs across these
nine years, while they make 4.08 purchases during this period.
Measures
Dependent Variable
The outcome variable in our research is customer purchase
incidence in each quarter during the estimation period (2009
2016), equal to one if the customer imakes a purchase in quarter
t. The reason for choosing quarter as our time unit is that
customers at optical retailers do not purchase glasses or lenses
every day or even every month; unlike other industries such as
fast-moving consumer goods, where customers make purchases
on a daily or weekly basis. Our data support this claim since the
median inter-purchase time is 192 days. Furthermore, retailers in
our dataset send DMs on a quarterly basis as a common practice.
The estimation window for a customer starts the first quarter
of 2009 if a customer was client before 2008 or one year after the
customer registers (typically purchases) for the first time if the
customer was a new customer/registered during the observation
period. The first year (Q1Q4 2008 for those that registered
before; and the first year for new customers) is used as
initialization period for the independent variables. The reason
that we start modeling a customer's behavior after his/her first
purchase is that retailers start sending DMs after a customer
enters their database. The estimation window for each customer
ends at the last quarter of our observation period (Q4 2016).
Independent Variables
The variable operationalization is described in detail in
Table 2. We use dummy variables to indicate whether a
customer has received a CTA DM in a quarter. We removed
109 (0.0005%) customers who received more than one CTA
DM in a quarter, because this pattern is not in line with the
company's policy of not sending customers multiple DMs per
quarter. If there was a CTA DM, we categorized which type of
incentive was included. These incentives were split into non-
monetary incentives, monetary incentives, and CTA DMs
without incentive based on their content. DMs that are inviting
customers to visit the store and/or make a purchase, without
indicating any reward after taking this (visit/purchase) action
are labeled as CTA DMs without incentive. A CTA DM that
provides a customer with a monetary reward, like a price
discount or a buy one get a second for free offer, was identified
as a monetary incentive. A CTA DM that rewards a customer's
purchase with a gift was classified as a non-monetary incentive.
A further distinction was made between a hedonic and
utilitarian non-monetary incentive, building on (Palazon and
Delgado-Ballester 2013). For instance, a gift such as a piece of
artwork was labeled as hedonic, while a sun tan crème that
offers a very functional benefit was classified as utilitarian.
The incentive characteristics for each of the CTA DMs were
coded independently by two coders based on the original visual
provided by the data provider. Cross-checks between the two
coders showed high inter-rater reliability for all of the data
accumulated (see for a similar approach, (Feld et al. 2013)).
Intensity of CTA DMs varies across the quarters in a sense that
retailers send more DMs in quarter 2 (AprilJune) and 4
(OctoberDecember) than in quarter 1 and 3. CTA DMs mainly
stimulate existing customers to purchase, for instance, new
glasses and therefore are focused on cross-selling; an objective
that does not vary across incentive types. Monetary incentives
appear either as a percentage discount on the purchase of a new
pair of glasses or as the offer to receive a second pair of glasses
for free. Non-monetary incentives, on the other hand, include a
reward such as a free collection item, an artwork or a free sun
cream. Across all CTA DMs sent to customers, the average
value of monetary incentives is 15 euro and of non-monetary
incentives is 34.6 euro. We control for these value differences
in the analysis (see Section Control variables). Of the total
CTA DM campaigns between 2008 to 2016 in our study, 60%
included an incentive, and about one-third of these ones are
with a monetary incentive and two-thirds are with a non-
monetary incentive. The latter is further divided in 33.4% of
CTA DMs with an utilitarian incentive and 66.6% with a
hedonic incentive. Taking into account the number of DMs sent
to customers between 2008 and 2016 per incentive type,
customers received in 31.52% of the cases a CTA DM with no
incentive, in 30.46% of the cases a CTA DM with a monetary
incentive and in 38.48% of the cases a CTA DM with a non-
monetary incentives, split between 46.2% and 53.8% for,
respectively, utilitarian and hedonic incentives.
Next, we operationalize relationship history measures. For
purchase frequency, we use the customer's last year (past four
quarters) to calculate this variable (similar in spirit to (Van
Diepen et al. 2009)). We discount the purchase incidence in each
quarter by weighting with a decay parameter (0.75) for each
quarter and then summing them up to capture fading. Purchase
recency is defined as the number of quarters without a purchase
immediately preceding the current period. Therefore, a higher
recency means longer duration since the previous purchase.
Customer relationship duration is measured as the number of
quarters since a customer has been registered in the retailer's
database. We use a similar operationalization to calculate past
DM communications: DM frequency is calculated by summing
the occurrence of CTA DMs in each of the past four quarters, and
discounting them with a decay parameter (0.75) for each quarter.
67S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Table 2
Variable Operationalization.
Variables Operationalization Equation
CTA_DM Dummy variable that indicates whether a customer ihas received a CTA DM in quarter t.
Dummy variable that indicates whether a customer ihas received a relational DM in quarter t.Relational_DM
Store_DM Dummy variable that indicates whether a customer ihas received a store-specific DM in quarter t.
Relationship history
Purchase_frequency Sum of purchase incidence of customer iin each quarter weighted
with a decay parameter (0.75) for each quarter going back up to
four quarters in the past (t-4, t-3, t-2, t-1). Pt¼tq1
t¼tq4δtqtIncidenceit , where t
q
is the calendar time of the
current quarter and δ= .75
Purchase_recency Number of quarters with no purchase of customer igoing back up
to the start of that customer's observation window (Q1 of 2008 for
those customers that were already in the dataset before the
observation window, or the first quarter that a new customer
registers (typically purchases) for the first time).
t
q
t
ip
, where t
ip
is the last calendar time where customer i made a
purchase (or in the case of no purchase, the start of the observation
window) and t
q
is the calendar time of the current quarter
Relationship_duration Number of quarters since the first purchase/registration of customer
iin the retailer's database.
t
q
t
ir
, where t
ir
is the calendar time where customer i registered for
the first time and t
q
is the calendar time of the current quarter
DM history
DM_frequency Sum of the CTA DM dummies of customer iin the past four
quarters (t-4, t-3, t-2, t-1), discounted with a decay parameter
(0.75). Pt¼tq1
t¼tq4δtqtCTADM it where t
q
is the calendar time of the current
quarter and δ= .75
DM_recency Number of quarters with no CTA DM for customer igoing back up
to the start of that customer's observation window.
t
q
t
i,cta_dm
, where t
i,cta_dm
is the last calendar time where customer
i received a CTA DM (or in the case no CTA DM was received, the
start of the observation window) and t
q
is the calendar time of the
current quarter
Socio-demographics
Age Age of customer imeasured in years.
Gender Indicator variable for customer i's gender (male = 1).
Control variables
Value_Incentive Market value of the incentive (in ) at the time that the incentive was send to customers.
Relational_DM_frequency Sum of relational DM dummies of customer iin the past four
quarters (t-4, t-3, t-2, t-1), discounted with a decay parameter
(0.75). Pt¼tq1
t¼tq4δtqtRelationalDMit where t
q
is the calendar time of the
current quarter and δ= .75
Store_DM_frequency Sum of store-specific DM dummies of customer iin the past four
quarters (t-4, t-3, t-2, t-1), discounted with a decay parameter
(0.75). Pt¼tq1
t¼tq4δtqtStoreDMit where t
q
is the calendar time current
quarter and δ= .75
Relational_DM_recencey Number of quarters with no relational DM for customer igoing
back up to the start of that customer's observation window.
t
q
t
i,rel_dm
, where t
i,rel_dm
is the last calendar time where customer
i received a relational DM (or in the case no relational DM was
received, the start of the observation window) and t
q
is the calendar
time of the current quarter
Store_DM_recency Number of quarters with no store-specific DM for customer igoing
back up to the start of that customer's observation window.
t
q
t
i,store_dm
, where t
i,store_dm
is the last calendar time where
customer i received a store-specific DM (or in the case no store-
specific DM was received, the start of the observation window) and
t
q
is the calendar time of the current quarter
Lag_CTA_DM The lagged value of the CTA_DM of the previous quarter (t-1).
Lag_Relational_DM The lagged value of the Relational_DM of the previous quarter
(t-1).
Lag_Store_DM The lagged value of the Store_DM of the previous quarter (t-1).
AVG CTA_DM Average number of CTA DMs that customer i has received up until
quarter t, going back up to the start of that customer's observation
window.
Pt¼tq
t¼tiCTADM it
ðtqtiþ1Þ, where t
q
is the calendar time of the current quarter, t
i
is the start of customer i's observation window
GDP_GrowthRate Percentage of GDP increase at quarterly level, compared to the same quarter in the previous year.
Instrumental variables
Attractiveness of DM
across all customers
Number of customers who received a CTA DM in quarter tacross
all 39 competitive retailers, divided by the total number of
customers in that quarter across these 39 competitive retailers.
(idem for relational DM)
Pj¼39
j¼1Pi¼njtq
i¼1CTADM ijtqPj¼39
j¼1njtq
, where i goes from 1 to njt
q
(the
number of active customers at retailer j in quarter q), and j captures
the 39 competitive retailers
Attractiveness of DM
across the customers of
a retailer
Number of customers other than customer ithat are receiving a
CTA DM at retailer j in quarter t, divided by the total number of
customers of retailer j in quarter tminus one.
(idem for relational DM)
Pnjtq1
i¼1CTADM ijtqnjtq1
, where i goes from 1 to njt
q
(the number of
active customers at retailer j in quarter q)
68 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
DM recency is the number of quarters with no DM received
(similar to (Van Diepen, Donkers, and Franses 2009).
Control Variables
We control for the effects of several other variables in our
study. First, we add the market value (monetary value, expressed
in ) of the incentives that were offered in the CTA DMs at the
time they were offered to a customer. We do this to make sure
that our results are not affected by differences in monetary value
of the incentives included in CTA DM. Second, we control for
other DMs sent to customers as well as their recency and
carryover effects. Retailers may send some other DM to
customers with a very different objective from CTA DMs that
are the focus of this study. One type is store-level DMs,
informing customers about shop-specific news (for instance,
construction works, remodeling at the store, etc.). Another type
includes DMs that are sent to customers after making a purchase.
These DMs are not CTA mailings and their explicit purpose is to
reinforce the relationship with the customers (Verhoef 2003).
Since the objective of these two types of DMs is merely fostering
the relationship with the customer or inform them about store-
specific news, without inviting any specific action such as making
a purchase, they are not the focus of this research. Third, we
include the lagged effect of all three DM types (CTA DM,
relational DM, and store-level DM), to capture the delay in
response of customers to a DM received in the last quarter.
Fourth, we include the average number of CTA DMs that each
customer has received in order to control for cross-sectional
correlation of the DMs with the random intercept, following
Datta, Foubert, and van Heerde (2015). Fifth, we include the
GDP growth rate as a proxy for the state of the economy in our
model, to control for any effect it might have on purchase
incidence. Finally, we control for (unobserved) differences
between quarters, years and retailers by including quarter, year
and store dummies.
Table 3 presents descriptive statistics for the variables. The
DM history measures for CTA DMs point to a low level of
carryover (average DM frequency of 0.23) and a long time of not
receiving CTA DMs (about 8.31 quarters). The average
purchase frequency (as operationalized in Table 2, that is, with
the summed and decay-weighted logarithm) equals 0.18, with a
substantial variation (SD = 0.35). Furthermore, customers do
not buy during 8.6 quarters (about 2 years), again with quite
some variation (SD = 7.05), and the average relationship length
is 14.51 quarters (about 3.5 years). 44% of our sample are males
and the average age of customers is 58 years. Average GDP
growth rate is 0.23%, with variations between 3.5% and 1.14%
and the monetary value of incentives as observed in the dataset
(including the zero observations) is 2.2. If we calculate the
average value of the incentive across the CTA DM campaigns
with an incentive only, it amounts to 27.6on average.
Model
We model individual i's purchase decision at quarter t,t=1,
,Τusing a probit regression. The corresponding equation can
be written as:
Pr PIit ¼1ðÞ¼β0þβ1CTADM it
þtype¼3
type¼1β2;typeCT ADM it Incentivetype

þchar ¼1char ¼7β3;charCT ADM it Consumerchar

þchar ¼1char ¼7β4;charConsumerchar

þcontrol¼12
control¼1β5;control Variablecontrol

þX
dummy¼50
dummy¼1
β6;dummyShop=Year=Seasondummy
!
þδiþit
ð1Þ
Table 3
Descriptive statistics of variables
a
.
Variable Mean SD Min. Max.
DM CTA DM 0.12 0.32 0 1
Relational DM 0.20 0.40 0 1
Store DM 0.07 0.26 0 1
Relationship history Purchase frequency 0.18 0.35 0 2.05
Purchase recency 8.60 7.35 1 36
Relationship duration 14.51 8.78 1 32
DM history DM frequency 0.23 0.331 0 1.73
DM recency 8.31 7.7 1 36
Socio-demographics Age 58 16 18 90
Gender (male = 1) 0.44 0.5 0 1
Control variables Value of incentive (in ) 2.22 12.22 0.057 100
Relational DM frequency 0.43 0.49 0 2.05
Store DM frequency 0.14 0.28 0 2.05
Relational DM recency 7.54 8.02 1 36
Store DM recency 11.37 8.04 1 36
Average CTA DM per customer 0.12 0.081 0 0.66
GDP growth rate per quarter (percentage) 0.23 0.67 3.5 1.14
a
Note that these descriptives are for the variables as they are operationalized in Table 2. That is, we use carryover measures (with decay parameters) for the
purchase frequency and the DM frequency measures, and include the zero observations when calculating the market value of the incentive. Recency and relationship
duration are expressed in quarters.
69S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
In line with the measures described above, we include the
incentive types (first brackets: monetary, non-monetary he-
donic, non-monetary utilitarian) and the interactions of the
consumer characteristics (second brackets: purchase frequency,
purchase recency, relationship duration, DM frequency, DM
recency, age and gender). Next to the main effects of these
consumer characteristics (third brackets), we include the
control variables (market value of the incentives, other DM
measures, average CTA DM, GDP growth rate; fourth
brackets) and the shop, year and season dummies (fifth
brackets). We use random effects in the probit model to capture
unobserved heterogeneity across customers (Van Diepen,
Donkers, and Franses 2009). All independent variables, with
the exception of the dummy variables, are mean-centered for
ease of interpretation.
With several independent variables, multi-collinearity might
be a concern. Table 4 provides the correlations between
variables. Checking the variation inflation factors (VIFs),
there are some variables with VIF values above the recom-
mended value of 10 ((Hair et al. 2010), p. 204), namely the two
most recent year dummies (2015 and 2016) and the main effect
of customer relationship duration. This is not unsurprising,
given the operationalization of relationship duration where
relationship duration increases as the year increases (see also
robustness checks section below).
A crucial issue that needs to be addressed in our modeling
approach is that retailers may use target selection techniques to
decide which customers receive a CTA DM. That is, retailers
do not send CTA DMs randomly to every customer and rather
take into consideration the likelihood of making a new purchase
and select customers with high purchase probabilities. Building
on Germann, Ebbes, and Grewal (2015), we introduce an
equation that explains the decision to send DMs by retailers in
order to control for this endogeneity (control function
approach). We add instrumental variables (IVs) that explain
the retailer's DM decisions, at the same time making sure they
are not correlated with the error term of the purchase incidence
equation. As instruments, we use two measures that capture the
(unobserved) attractiveness of CTA DMs among retailers. We
conjecture that CTA DMs are likely to be more attractive, when
more consumers (at other retailers) and more consumers (at the
focal retailer) receive the CTA DM in that quarter. Hence, we
use (i) the prevalence of a CTA DM across the other retailers in
the same quarter t(the % of customers at other retailers
receiving the CTA DM) and (ii) the prevalence of a CTA DM
across all customers of the same retailer in the same quarter t
(the % of customer of the retailer, except the focal customer,
receiving the CTA DM). See for more details on the
operationalization of these instruments in Table 2. We also
control for the endogeneity of relational DMs with the same
logic for selecting instruments.
Instruments are conceptually good, when they are relevant
and do not correlate with the error term of the purchase
incidence equation. The prevalence of the DM across all
customers reflects the decisions of a large number of retailers at
different locations. It therefore is indicative of the attractiveness
of a certain DM. It is unlikely that these retailers collectively
make decisions to send one particular DM to their customers,
given that they are located in different areas, and do not belong
to the same chain. In addition, the prevalence of DMs within
customers of a retailer can capture the idea that some DMs are
inherently more interesting for customers which can influence
the decision of a particular retailer to distribute these DMs
among its customers. Our instruments are uncorrelated with the
omitted variables that affect the customer's purchase incidence
decision, mainly because the purchase decision of an individual
customer cannot be influenced by the prevalence of a DM
across all customers as well as all customers within a specific
retailer, simply because that customer is not aware of the
purchase decisions of other customers.
We also tested for the strength and validity of our IVs, in
line with Germann, Ebbes, and Grewal (2015)s recommenda-
tions. Using the Wald test of the exogeneity ((Wooldridge
2010), pp. 472477) to test the endogeneity issue with multiple
potential endogenous regressors, we reject the null hypothesis
of no endogeneity (pb.01), which confirms that endogeneity
indeed exists. First, to check for the strength of the IVs to
explain each endogenous variable in the first stage, we follow
the approach suggested by (Stock and Yogo 2005). We reject
that our instruments are weak because all first stage F-statistics
for IVs are higher than Stock-Yogo weak ID F-test critical
values (at 5%). Hence, the IVs are sufficiently strong. Second,
testing for the validity of our instruments, the Hansen J test is
not significant for the model (p. 15), which suggests that the
null hypothesis (the IVs are uncorrelated with the error term)
cannot be rejected. Thus, the IVs are valid.
Estimation Results
We estimate four versions of the probit model predicting
customer purchase decisions, with different levels of aggrega-
tion for incentive types. This approach allows us to compare the
coefficients of different types of incentive, depending on what
is hypothesized in H1 to H4. In model 1, all CTA DMs, no
matter what incentive they include, are represented by a single
dummy variable to see the overall effect of CTA DMs on
purchase incidence, which allows us to test H1. In model 2, we
combine all CTA DMs with an incentive, either monetary or
non-monetary, into one variable, to test H2 on how a CTA DM
with an incentive performs compared to a CTA DM without
incentive. In a similar vein, in model 3, we separate monetary
and non-monetary (not distinguishing between hedonic and
utilitarian) incentives using a dummy variable, to check the
relative performance of monetary and non-monetary incentives
as stipulated in H3. Finally, in model 4 (presented in Eq. 1), we
have separate dummy variables for monetary, non-monetary
utilitarian and non-monetary hedonic incentives, to test H4 by
comparing the coefficients of the latter two and to test H5H
10
.
Table 5 gives an overview of our estimation results. Because of
space constraints, we do not provide the results of the first stage
(endogeneity-correction) models, nor for the retailer dummies,
but they can be provided upon request. We compared the
results of the purchase incidence model corrected for
endogeneity (see Table 5) with the model that is not corrected
70 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Table 4
Correlations between variables.
1 23456 789101112131415
CTA_DM 1
Hedonic 0.529 1
Utilitarian 0.399 0.031 1
Monetary incentive 0.427 0.033 0.025 1
*Purchase_frequency 0.426 0.243 0.211 0.197 1
*Purchase_recency 0.293 0.411 0.005 0.048 0.486 1
*Relationship _duration 0.235 0.600 0.242 0.134 0.035 0.530 1
*DM_frequency 0.202 0.355 0.019 0.065 0.172 0.347 0.244 1
*DM_recency 0.113 0.125 0.089 0.006 0.202 0.170 0.099 0.648 1
*Age 0.101 0.041 0.197 0.188 0.200 0.104 0.103 0.036 0.178 1
*Gender (male = 1) 0.638 0.333 0.258 0.272 0.293 0.196 0.141 0.129 0.075 0.048 1
Purchase_frequency 0.126 0.071 0.061 0.058 0.268 0.131 0.012 0.048 0.055 0.054 0.086 1
Purchase_recency 0.091 0.130 0.000 0.014 0.154 0.319 0.169 0.110 0.054 0.033 0.061 0.524 1
Relationship_duration 0.071 0.205 0.090 0.052 0.007 0.183 0.349 0.083 0.036 0.037 0.042 0.070 0.514 1
DM_frequency 0.075 0.126 0.009 0.021 0.062 0.122 0.086 0.348 0.225 0.013 0.048 0.136 0.200 0.158 1
DM_recency 0.056 0.054 0.038 0.004 0.080 0.066 0.033 0.240 0.367 0.067 0.037 0.152 0.096 0.120 0.644
Age 0.035 0.015 0.062 0.060 0.064 0.033 0.031 0.012 0.056 0.310 0.017 0.174 0.108 b0.001 0.088
Gender (Male = 1) b0.001 0.002 0.002 0.001 0.009 0.004 0.004 0.000 0.002 0.008 0.275 0.038 0.020 b0.001 0.001
Value_incentive 0.482 0.207 0.025 0.827 0.222 0.271 0.241 0.037 0.022 0.147 0.308 0.065 0.085 0.079 0.015
Relational_DM 0.059 0.015 0.052 0.041 0.036 0.054 0.049 0.002 0.022 0.025 0.040 0.005 0.145 0.088 0.058
Relational_DM_frequency 0.102 0.005 0.072 0.046 0.069 0.041 0.041 0.034 0.055 0.039 0.067 0.075 0.155 0.094 0.159
Relational_DM_recency 0.091 0.011 0.061 0.050 0.076 0.017 0.053 0.046 0.100 0.047 0.061 0.074 0.206 0.272 0.153
Store_DM 0.149 0.121 0.063 0.094 0.087 0.077 0.089 0.059 0.033 0.028 0.094 0.056 0.039 0.072 0.085
Store_DM_frequency 0.091 0.113 0.012 0.018 0.043 0.084 0.108 0.111 0.071 0.019 0.055 0.068 0.124 0.194 0.230
Store_DM_recency 0.063 0.030 0.030 0.027 0.040 b0.001 0.039 0.053 0.103 0.003 0.040 0.039 0.098 0.244 0.146
lag_CTA_DM 0.103 0.017 0.056 0.057 0.036 0.014 0.002 0.053 0.019 0.007 0.067 0.088 0.086 0.068 0.692
lag Relational_DM 0.050 0.008 0.043 0.017 0.033 0.038 0.029 0.007 0.022 0.022 0.032 0.035 0.113 0.066 0.081
lag Store_DM 0.001 0.034 0.032 0.043 0.001 0.014 0.046 0.051 0.022 0.020 0.002 0.046 0.051 0.082 0.137
AVG CTA_DM 0.252 0.171 0.087 0.096 0.174 0.088 0.022 0.223 0.225 0.048 0.161 0.208 0.093 0.000 0.443
GDP_growth 0.016 0.062 0.308 0.218 0.017 0.053 0.098 0.022 0.006 0.007 0.009 0.031 0.159 0.330 0.054
71S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
CTA_DM
Hedonic
Utilitarian
Monetary incentive
*Purchase_frequency
*Purchase_recency
*Relationship _duration
*DM_frequency
*DM_recency
*Age
*Gender (male = 1)
Purchase_frequency
Purchase_recency
Relationship_duration
DM_frequency
DM_recency 1
Age 0.177 1
Gender (Male = 1) 0.006 0.011 1
Value_incentive 0.001 0.047 b0.001 1
Relational_DM 0.081 0.044 0.004 0.018 1
Relational_DM_frequency 0.184 0.078 0.007 0.032 0.220 1
Relational_DM_recency 0.330 0.091 0.011 0.032 0.272 0.668 1
Store_DM 0.053 0.015 0.001 0.138 0.036 0.101 0.080 1
Store_DM_frequency 0.153 0.027 0.002 0.027 0.057 0.127 0.121 0.116 1
Store_DM_recency 0.334 0.038 0.003 0.016 0.098 0.180 0.346 0.096 0.573 1
lag_CTA_DM 0.382 0.033 . b0.001 0.054 0.040 0.102 0.088 0.022 0.138 0.082 1
lag Relational_DM 0.096 0.045 0.004 0.006 0.041 0.742 0.425 0.061 0.060 0.098 0.065 1
lag Store_DM 0.088 0.015 0.002 0.029 0.034 0.084 0.075 0.009 0.714 0.352 0.142 0.037 1
AVG CTA_DM 0.473 0.144 0.001 0.134 0.137 0.250 0.313 0.107 0.185 0.268 0.235 0.141 0.102 1
GDP_growth 0.021 0.018 b0.001 0.190 0.038 0.048 0.072 0.068 0.019 0.031 0.048 0.031 0.003 0.035 1
72 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Table 5
Model estimation results
a
.
Variables Model 1 Model 2 Model 3 Model 4
Coef. SE Coef. SE Coef. SE Coef. SE
Intercept 1.010** 0.019 1.011** 0.019 1.020** 0.019 1.057** 0.022
CTA_DM 0.151** 0.005 0.125** 0.007 0.112** 0.007 0.105** 0.008
Incentive 0.044** 0.008
Monetary incentive 0.001 0.009 0.001 0.009
Non-monetary incentive 0.108** 0.010
Hedonic 0.072** 0.016
Utilitarian 0.114** 0.010
CTA_DM * Relationship history
*Purchase_frequency 0.060** 0.011 0.060** 0.011 0.052** 0.011 0.051** 0.011
*Purchase_recency 0.011** 0.001 0.011** 0.001 0.011** 0.001 0.012** 0.001
*Relationship _duration 0.003** 0.001 0.003** b0.001 0.007** 0.001 0.007** 0.001
CTA_DM * DM History
*DM_frequency 0.063** 0.013 0.057** 0.013 0.064** 0.013 0.065** 0.013
*DM_recency 0.005** 0.001 0.005** 0.001 0.005** 0.001 0.005** 0.001
CTA_DM * socio-demographics
*Age 0.004* b0.001 0.004* 0.000 0.004* b0.001 0.004** b0.001
*Gender(male = 1) 0.014** 0.006 0.014** 0.006 0.015** 0.006 0.015** 0.006
Relationship history
Purchase_frequency 0.584** 0.004 0.584** 0.004 0.584** 0.004 0.584** 0.004
Purchase_recency 0.015** b0.001 0.015** b0.001 0.015** 0.000 0.015** b0.001
Relationship_duration 0.011** 0.001 0.011** 0.001 0.011** 0.001 0.011** 0.001
DM history
CTA_DM_frequency 0.165** 0.004 0.165** 0.004 0.164** 0.006 0.165** 0.004
CTA_DM_recency 0.003** b0.001 0.003** b0.001 0.003** b0.001 0.003** b0.001
Socio-demographics
Age 0.006** b0.001 0.006** b0.001 0.006** b0.001 0.006** b0.001
Gender (male = 1) 0.048** 0.003 0.048** 0.003 0.048** 0.003 0.048** 0.002
Control variables
Value_incentive 0.001** b0.001 0.001** b0.001 0.0001 b0.001 0.0001* b0.001
Relational_DM 0.209** 0.004 0.208** 0.004 0.207** 0.005 0.207** 0.004
Relational_DM_frequency 0.064* 0.003 0.064* 0.003 0.064** 0.004 0.064** 0.003
Relational_DM_recency 0.002** b0.001 0.002** b0.001 0.002** b0.001 0.002** b0.001
Store_DM 0.035** 0.004 0.036** 0.004 0.036** 0.004 0.037** 0.004
Store_DM_ frequency 0.010 0.004 0.010 0.004 0.009 0.004 0.009 0.004
Store_DM_ recency 0.0003 b0.001 0.0003** b0.001 0.0001 b0.001 0.0001 b0.001
AVG CTA_DM 2.726** 0.019 2.727** 0.019 2.700** 0.019 2.700** 0.019
Lag_CTA_DM 0.095** 0.004 0.094** 0.004 0.094** 0.004 0.094** 0.004
Lag_Relational_DM 0.039** 0.005 0.039** 0.005 0.040** 0.005 0.039** 0.005
Lag_Store_DM 0.039** 0.005 0.040** 0.005 0.040** 0.005 0.041** 0.005
GDP_growth 0.005 0.041 0.003 0.003 0.001 0.003 0.003 0.003
Quarter
Q2 0.046** 0.003 0.048** 0.003 0.050** 0.003 0.050** 0.003
Q3 0.055** 0.003 0.056** 0.003 0.056** 0.003 0.056** 0.003
Q4 0.027** 0.004 0.026** 0.004 0.025** 0.004 0.025** 0.004
Year
2010 0.034** 0.005 0.033** 0.005 0.033** 0.005 0.032** 0.005
2011 0.075** 0.008 0.074** 0.008 0.072** 0.005 0.073** 0.005
2012 0.174** 0.011 0.172** 0.011 0.168** 0.007 0.170** 0.007
2013 0.248** 0.013 0.245** 0.013 0.238** 0.009 0.240** 0.009
2014 0.320** 0.014 0.318** 0.014 0.311** 0.010 0.314** 0.010
2015 0.396** 0.015 0.396** 0.015 0.382** 0.012 0.384** 0.012
2016 0.502** 0.016 0.502** 0.016 0.492** 0.014 0.494** 0.014
*pb.05 (two-sided) **pb.01 (two-sided).
a
Because of space constraints, we do not provide the results of the rst stage (endogeneity-correction) models, nor for the retailer dummies, but they can be
provided upon request.
73S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
for endogeneity (see Appendix A). This comparison indicates
that correcting for endogeneity makes the effect of CTA DM as
well as the interaction with incentive types smaller in size, and
is therefore important to apply (Ebbes, Papies, and Van Heerde
2011).
Hypotheses Tests
Main Effect of DMs
To test H1, we look at model 1, where we have one dummy
variable that captures all forms of CTA DM. The significant
positive coefficient of the CTA DM (β= 0.151, pb.01)
indicates that such DMs have a positive impact on the
customer's purchase probability and therefore, H1(a) is
confirmed. Note that this coefficient should be interpreted as
the effect of the CTA DM for females at the mean value for all
continuous moderators (because they were mean-centered), due
to the interaction terms in the model.
The Moderating Role of DM Incentive
To test H2, we focus on model 2, where we can interpret the
coefficient of the dummy variable that captures whether a CTA
DM had an incentive. This coefficient indicates the additional
effect of an incentive beyond the main effect of sending a CTA
DM (which is captured by the baseline coefficient). Given the
positive significant coefficient of the incentive variable
(β= 0.044, pb.01), we can conclude that a CTA DM with
an incentive has a higher impact on purchase incidence than
a CTA DM without an incentive. Thus we find support for
H2(a).
In model 3, we estimate the coefficients of monetary and
non-monetary incentives and compare these to test H3.To
assess whether the coefficients of monetary versus non-
monetary incentives are significantly different from each
other, we calculate the difference between these coefficients
and divide it by the standard error derived via the Delta method
(Greene 2003). This comparison is significant (pb.01). The
coefficient of the non-monetary incentive (β= 0.108, pb.01) is
significantly higher than that of the monetary incentive (β=
0.001, p= .13), with the latter not being significantly different
from the no incentive CTA DM. Thus, we conclude that non-
monetary incentives have a higher (more positive) impact on
purchase incidence compared with monetary incentives,
supporting H3(a).
We further test whether the coefficients of non-monetary
hedonic and utilitarian incentives in model 4 are significantly
different from each other. Parameter tests of the difference
between these coefficients show that the difference is
significant (pb.01). Hence, H4a is supported. Although both
a CTA DM with a non-monetary hedonic (β= 0.072, pb.01)
as well as one with a utilitarian incentive (β= 0.114, pb.01)
positively influence a customer's purchase incidence compared
to a CTA DM without an incentive, our results also indicate
that a utilitarian incentive has a higher impact on customer
purchase incidence than a hedonic one.
The Moderating Role of Customer Characteristics
For the hypotheses tests of the moderating role of customer
characteristics, we focus on model 1, but the results are robust
across the four versions of the model (see Table 5). To enhance
the interpretation of the interaction effects, we estimate and test
the simple effect of the CTA DM at different levels of the
moderators, which is called a spotlight test (Spiller et al. 2013).
In particular, we calculate the simple effect of a CTA DM (and
its significance level) for low (one standard deviation below the
mean value) and high (one standard deviation above the mean
value) levels of the moderators. We also calculate threshold
values (see (Melis et al. 2015) for a similar approach) to
identify the range of the moderator values where the simple
effect of the CTA DM is significant. The results are presented
in Table 6.
Relationship history. The positive significant coefficient of
the interaction between a CTA DM and purchase frequency (β
= 0.060, pb.01) shows that the impact of a CTA DM on
purchase incidence is more pronounced for customers with
higher purchase frequency (support for H5). The impact of
CTA DMs on purchase incidence is positive and significant for
all values of purchase frequency as observed in our data and
increases as purchase frequency increases.
The coefficient of the interaction term between the CTA DM
and purchase recency is negative and significant (β=0.011,
pb.01) (support for H6). The spotlight analyses show that the
effect of CTA DM on purchase incidence is highly positive at
the low purchase recency value (mean value minus 1 SD), and
lower but still positive at the high level (mean value plus 1 SD).
The effect of CTA DM is positive and significant at a wide
range of recency levels, namely for 1 to 21.6 quarters.
However, it is decreasing and becomes negative and significant
at the very high levels of time since purchase (25.7 quarters and
higher).
The significant and positive coefficient of the interaction
between relationship duration and the CTA DM (β= 0.003, p
b.01) indicates that effect of the CTA DM on purchase
incidence increases as the duration of the relationship between
the customer and retailer increases (support for H7). The effect
Table 6
Overall effect of CTA DM at different levels of continuous moderators.
Total effect
CTA_DM
Low level
of
moderator
High
level of
moderator
Range of
values where
the effect is
significant and
negative
Range of
values where
the effect is
significant
and positive
Purchase_frequency 0.128** 0.149** [0, 1.86]
Purchase_recency 0.226** 0.074** [25.7, 32] [1, 21.6]
Relationship_duration 0.104** 0.195** [1, 32]
DM_frequency 0.139** 0.261** [0, 2.05]
DM_recency 0.186** 0.025** [1, 32]
Age 0.002** 0.278** [18, 70]
We use non-mean-centered values and define low = mean1 SD (or 0, if
mean1SDb0), high = mean + 1 SD. The hypotheses are two-sided with p-
values: *pb.05, **pb.01. The minimum and maximum values in the reported
ranges are the ones observed in the dataset.
74 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
is positive and significant across all values of relationship
duration as observed in our data and increases as the
relationship duration goes up.
DM history. The interaction between the CTA DM and DM
frequency is positive and significant (β= 0.063, pb.01),
suggesting that message repetition works (support of H8(b)).
The impact of CTA DM frequency on purchase incidence
increases as the value of CTA DM frequency increases and is
significant at the whole range of values of CTA DM frequency.
The coefficient of the interaction between CTA DM and DM
recency is negative and significant (β=0.005, pb.01), which
indicates that the impact of CTA DM on purchase incidence
decreases if the time since the last CTA DM mailed to this
customer increases (support of H9(b)). The impact of the CTA
DM on purchase incidence is positive and significant in the
whole range of recency values as observed in our data.
Socio-demographics. The results show that a CTA DM has a
higher influence on stimulating older customers to make a
purchase (β= 0.004, pb.01), supporting H10(b). The simple
effect of a CTA DM on purchase incidence is small but
significant at low age values and increases at higher age levels.
Lastly, the interaction term between CTA DM and gender is
positive and significant (β= 0.014, pb.01). The effect of CTA
DMs on purchase likelihood is higher for males than females.
This is in contrast to our expectations where males were
expected to be less likely to react to a CTA DM than females
(no support for H11).
The Impact of Control Variables
Table 5 shows that quarter and year dummies are all
significant, indicating that not controlling for these variables
would potentially cause an omitted variable bias. Also, the
retailer dummies (omitted because of space constraints) are all
significant. The impact of the market value of the incentive
used in the CTA DM on purchase incidence is positive and
significant in all models (except model 3), in line with the
expectation that the higher the market value of the incentive
involved, the more likely consumers are to make a purchase.
Next, controlling for the effect of other DMs that a customer
receives is important, given the significant effects of current
(captured by the dummy variables) as well as past DMs
(captured by recency and frequency measures of these DMs).
Negative effects of both currentrelational DMs and store-
specific DMs on purchase incidence may seem unexpected, but
store-specific DM communication is typically send when there
is a disruption at the store or its environment; and relational
DM is often send to build a relationship instead of generating
purchases in the same period. The lagged effects of all three
DMs are significant. It is negative for the CTA DM which
suggests that consumers that received a CTA DM in the
previous quarter are less likely to make a purchase in the
following quarter, while the opposite holds for relational and
store DMs where consumers are more likely to make a purchase
in the quarter following the one where they received such a
DM. The average number of CTA DMs that customers have
received is significant and negative, which may capture the
effect that customers who receive more direct marketing are
less likely to purchase in the first place. The coefficient of GDP
growth rate is not significant. Finally, the main effects of the
moderator variables on purchase incidence are in line with the
expectations, corroborating the validity of our model: cus-
tomers are more likely to make a purchase at the retailer when
they have made more purchases in the past, have bought more
recently, are with the retailer during a longer time, received
CTA DM more frequently and more recently, are younger and
are female.
Simulation
To assess the effect size of changes in customer character-
istics as well as incentive types on purchase probability, we
simulate a meaningful change in values of each explanatory
variable (/+ 1 SD from the mean for continuous variables and
from 0 to 1 for binary variables), while fixing other continuous
variables at their mean and dummy variables at a certain value,
then compute the predicted change in purchase probability. By
doing this, we gain insights in the effect size for a change in the
value of a variable of interest on purchase probability. We use
model 1 to calculate the predicted purchase probabilities for
changes in the customer characteristics, and model 4 for
changes in the incentive types. Result of these simulations are
provided in Table 7.
Table 7
Simulation results
a
.
Variable Value range Purchase
probability
Model 1 Model 4
No CTA DM 15.6% 15.6%
Overall effect of CTA DM (irrespective of incentive) 19.5%
Effect of CTA DM without incentive 18.4%
Effect of CTA DM with monetary incentive 18.5%
Effect of CTA DM with non-monetary hedonic incentive 20.5%
Effect of CTA DM with non-monetary utilitarian incentive 21.7%
Effect of CTA DM for changes in:
Relationship history Purchase frequency Low (0) 16.5%
High (0.5) 26.3%
Purchase recency Low (1) 24.8%
High (16) 15.7%
Relationship duration Low (6) 16.3%
High (22) 23.0%
DM history DM frequency Low (0) 18.1%
High (0.5) 21.6%
DM recency Low (1) 19.8%
High (16) 19.2%
Socio-demographics Age Low (42) 20.2%
High (73) 18.8%
Gender Female 19.5%
Male 18.6%
When doing the simulations, we fixed the non-focal continuous variables at
their mean and the non-focal dummy variables at a fixed given number (quarter
= 4, year = 2013, male = 0 and shopdummy25 = 1).
a
Low = mean of continuous variables - 1 SD; High = mean of continuous
variables +1 SD.
75S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
The results show that targeting a customer with a CTA DM
(irrespective of the incentive) increases a customer's purchase
probability from 15.6% to 19.5% in comparison with not
sending a CTA DM. A CTA DM with an utilitarian incentive
increases a customer's purchase probability the most (21.7%),
closely followed by a hedonic incentive (20.5%), while the
monetary incentive only slightly increases the purchase
probability (18.5%) compared with a CTA DM without an
incentive (18.4%). Table 7 further shows that the CTA DM
effect on customer's purchase probability is affected more by
changes in values of relationship history variables compared to
changes in values of DM history and socio-demographics. For
instance, the predicted purchase probability varies between
16.5% and 26.3% when a CTA DM is mailed to a customer
with low versus high value of purchase frequency. DM
frequency also plays a role because the predicted purchase
probability increases from 18.1% to 21.6% if the value of DM
frequency changes from low to high. The moderating roles of
DM recency, age and gender are relatively small in terms of
effect size, indicated by the low differences in predicted
purchase probabilities.
To illustrate implications for retailer targeting decisions
based on customer characteristics, we computed how much
customer's purchase probability changes if a retailer would
target two very different customers (while the customer profiles
are hypothetical, they are viable in the dataset). Customer A is a
42 year old male who does not purchase very frequently
(purchase frequency in last four quarters = 0, time since last
purchase = 5 quarters), has a low relationship duration (6
quarters) and receives CTA DMs not very frequently (low CTA
DM frequency = 0, long time since last time received a CTA
DM = 5 quarters). Customer B is a 73-year-old female who
purchases frequently (high purchase frequency = 0.5, short time
since last purchase = 1 quarter), has a high relationship duration
(22 quarters) and receives CTA DMs from retailers often (high
CTA DM frequency = 0.5, short time since last time received a
CTA DM = 1 quarter). The purchase probability of customer A
changes from 10.9% to 13.2% by receiving a CTA DM, while
purchase probability of customer B changes from 24.8% to
37.1%. This indicates that the predicted purchase probability
increases much more if the retailer targets Customer B instead
of Customer A.
As managers need to decide on the monetary value of the
incentive as well as on the incentive type, it is interesting to
assess the relative importance of these CTA DM decisions. To
do so, and following the procedure described above, we
calculate the effect size of the monetary value of incentives on
purchase probability. More specifically, we show that the
purchase probability increases only by 0.3% (from 19.5% to
19.8%, respectively) if the monetary value of the CTA DM
incentive increases by one standard deviation from its mean.
This change is very small compared to when the CTA DM type
changes: 3.2% (if a non-monetary utilitarian CTA DM is sent
instead of a monetary one) or 2.0% (if a non-monetary hedonic
CTA DM is sent instead of a monetary one). We therefore
conclude that the value of the incentive matters much less than
the type of incentive (monetary vs. non-monetary).
Robustness Checks
To check whether the model results are robust, several
alternative models have been tested. First, we changed the level
of aggregation of our analysis from quarter to year. This change
did not result in substantively different findings. Hence we
decided to stick to quarterly level data analysis. Second, we
tested the sensitivity of our result to the selection of IVs.
Instead of using the two IVs that we discussed above, we used
another IV, namely the prevalence of a DM across other
retailers in quarter t, to replace one of these IVs (we did this
robustness check twice). The parameters and corresponding
significance levels of other variables were hardly affected.
Third, given the high VIF value for the main effect of the
relationship duration variable, we excluded that variable and its
interaction with the CTA DM. The findings for the rest of
variables stayed the same, mitigating the concerns regarding
multi-collinearity. Fourth, there is some evidence in the
literature Van Diepen, Donkers, and Franses (2009) that
suggests that the interaction of a CTA DM with purchase
recency might have an inverted U-shaped form, that is, an
intermediate level of purchase recency resulting in a more
significant impact than a very high or a very low level of
purchase recency. We added the quadratic term of the
interaction to our model, however, it was not significant.
Fifth, we did sensitivity analyses on the decay parameters used
in the operationalization of DM frequency and purchase
frequency. Results for parameters between 0.7 and 0.9 suggest
that our results are not sensitive to the decay parameter value.
Therefore, we use 0.75, because this number is most commonly
used in the direct marketing literature (Gázquez-Abad et al.
2011). Sixth, we split the DM frequency and recency variables
per incentive level, leading to different DM frequency and
recency variables for CTA DMs with a monetary and non-
monetary incentive. The operationalization of these recency/
frequency variables at the incentive level was similar in spirit to
the one we used at the aggregate level. By including these
variables in interaction with the CTA DM dummy, we test
whether there exist a wear-our effect (Ansari et al. 2008), that
predicts that the frequent and/or recent use of monetary (non-
monetary) incentives decreases the impact of the CTA DM on
purchase incidence. This split between monetary and non-
monetary frequency plus recency did not result in a model fit
improvement, nor did it change the substantive findings.
Discussion
Given the very large expenditures retailers dedicate to direct
marketing communications in consumer durables (Gázquez-
Abad et al. 2011), such as optical products in our study,
understanding under which conditions a DM is effective in
triggering a purchase is a relevant issue for both academics and
retailers. In a context like ours that is characterized by low
purchase frequencies and high involvement (Grewal, Mehta,
and Kardes 2004). it becomes critical for retailers to target
customers with the right offers in order to keep close contact
with existing customers. Our primary goal in this research was
76 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
to develop insights on the impact of CTA DMs with different
incentive types. In addition, we were also interested in the
moderation effects of customer characteristics. This study adds
to the existing knowledge on the impact of CTA DMs on
customer's purchase incidence in the retail context, by
investigating the impact of a variety of incentives types, a
large set of consumer characteristics and this in a setting of
consumer durables (notably optical retailing).
The Moderating Role of CTA DM Incentive
The unique dataset analyzed in this study reveals that a CTA
DM with an incentive has a higher impact on customer's
purchase behavior, compared with a CTA DM without an
incentive. Customers perceive a CTA DM with an incentive, as
an extra effort and resource investment from retailer to foster
the relationship. Consequently, customers reciprocate a
retailer's action by making a new purchase, in line with what
(Arora and Stoner 1992) found. Next, we found that the impact
of a CTA DM on purchase incidence depends on the type of
incentive. Our results show that non-monetary incentives are
more influential compared to monetary incentives in stimulat-
ing customers to make a purchase. As demonstrated by (Nunes
and Park 2003), non-monetary incentives are perceived as
separate gains due to their incommensurate nature, which
makes it difficult to convert them into a common unit of
measurement. Hence, we find that in our research, customers
attach a higher perceived value to these non-monetary
incentives than to monetary ones. Another interesting finding
is that among the two possible non-monetary incentive, a CTA
DM with a utilitarian incentive has a slightly higher impact on
customer's purchase probabilty, compared to a CTA DM with a
hedonic incentive. This is in line with the stream of research
that argues that justification for the incentive is easier for
utilitation rewards that are primarily instrumental and func-
tional, leading to less guilty feelings for such incentives than for
hedonic incentives (Palazon and Delgado-Ballester 2013).
The Moderating Role of Customer Characteristics
Our findings substantiate the idea that customer heteroge-
neity in response to CTA DMs exists. More specifically, our
results show that customers who purchase more frequently have
an even higher purchase incidence when receiving a CTA DM.
Customers with higher purchase frequency are heavy users and
according to the relationship marketing literature, heavy users
engage more with a firm because they developed higher
capabilities in terms of understanding and engaging with firm
contacts compared to less experienced users. Our results
support the idea that this makes them more likely to react
positively to firm offers (Fader and Hardie 2007; Neslin et al.
2013).
Next, a CTA DM has a lower effect on purchase incidence
for customers with high purchase recency values (and thus,
longer time since past purchase), compared to customers who
have purchased more recently. According to reactance theory,
when (almost) defected customers receive DMs, they find these
DMs more manipulative and irritating than other (more
active) customers (Fitzsimons and Lehmann 2004). More-
over, the longer the time since the last purchase, the more likely
a lapsed customer is to have shifted to a new retailer or simply
to have developed new behaviors (Thomas, Blattberg, and Fox
2004). Our results thus support the idea that customers with
higher elapsed purchase time may ignore the newly received
CTA DM.
The CTA DM effect increases also as the duration of
customer relationship increases. This suggests that as the
duration of the relationship increases, the level of intimacy also
increases, leading to an increased richness of the customer's
impressions about the firm (Swann and Gill 1997). The
relationship marketing literature has shown that strong
customer relationships are those in which customers have
maintained a longer relationship, as the parties acquire
experience and get to know what they can expect from one
another (Bolton 1998). Our results support this idea, by
showing that customers with a longer relationship are more
likely to react to a CTA DM. Taken together, our results are in
line with well-established findings in customer relationship
management literature that suggests to target customers based
on RFM variables as well as relationship age with the firm
(Rust and Verhoef 2005).
Not only does the customer relationship history play a
critical role, the impact of a CTA DM on purchase incidence
also depends on DM history. Our results reveal that how recent
and frequent a customer received DM in the past plays a
significant role: customers who received DM with higher
frequency and lower recency have a higher probability to make
a purchase when targeted with a new CTA DM. This
corroborates with the idea suggested in advertising literature
that repeated exposure to advertising can lead to familiarity, a
positive attitude toward the company, and less likelihood of the
customer forgettingthe company over time (Naik and
Piersma 2002).
Finally, we found that socio-demographics like age and
gender play significant albeit small roles: older customers and
males are more likely to make a purchase when targeted with a
CTA DM. Contrary to our prediction, CTA DMs have more
impact on the purchase decision of males. One potential
explanation can be found in the product category under study,
which seems to make especially male consumers likely to
respond to a call-to-action for an extra pair of glasses, while
female consumers are more likely to make a purchase (without
a CTA DM) as is evidenced by the significant negative main
effect of gender.
Managerial Implications
The findings from our study provide valuable insights for
retailers. First, our results suggests that it pays off to provide
customers with a non-monetary incentive. Our results challenge
the practice of some retailers who spend substantial amounts of
their budget on sending DMs with monetary incentives (e.g.,
price discounts) in pursuit of stimulating customers to buy. We
demonstrate that retailers can enhances customer's purchase
77S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
probability by more than 14.6% if they switch from sending a
CTA DM with a monetary incentive to a non-monetary
incentive, especially with an utilitarian one.
Second, our findings show that targeting customers based on
their relationship history and DM history significantly affects
the customer's purchase probability. For instance, targeting
customer with a high rather than low purchase frequency with a
CTA DM can increase a customer's purchase probability by
59%. For the relationship duration, the purchase probability
increases by 41% if retailers target a customer with relatively
high relationship duration (22 quarters) instead of a customer
with low relationship duration (6 quarters). Targeting cus-
tomers with a low (1 quarter) vs. a high purchase recency (16
quarters) increases purchase probability by 57%. Targeting
customers based on RFM measures thus pays off, in line with
customer relationship management practices.
Third, we encourage retailers to get to know their customers
better, by keeping track of some essential information like
purchases made, timing of DMs sent and to a lesser extent
socio-demographics, because this enables retailers to manage
DM strategies more effectively. Our research substantiates that
a retailer can leverage customer and transaction data to increase
purchases via a more refined direct marketing strategy. After
acquiring such information, it also makes sense for retailers to
invest in internal data analysis skills necessary to take
advantages of such data.
Limitations and Suggestions for Future Research
In this study, we examined a large group of optical retailers
across the Netherlands. Comparable to other empirical studies,
the first question is whether our findings are generalizable to
other contexts as well. In this research we focused on optical
retailers that belong to consumer durables and that are
characterized by low purchase frequency and high involvement
level. There is a need to extend this study to other consumer
durable categories, but certainly also to high purchase
frequency and low involvement categories. A second limitation
of this study is that we could not observe the full behavioral
response of a customer to a DM and focused on purchase
incidence only. Next to investigating other aspects of consumer
purchase behavior, future research could also refine this
measure by, for instance, investigating the intermediate stages
of the direct mail response, such as the opening and keeping
rates of DMs (Feld et al. 2013). Third, our data captures the
actual behavior of customers (purchase incidence) and retailers
(timing and frequency of sending DM). It was not collected
with the purpose of theory testing. Our results are indicative of
the relevance of the various drivers, but field experiments
should be carried out to disentangle the causal influences and
corroborate the validity of our findings. However, such field
experiments would be quite costly and hard to implement,
given the number of moderators and control variables in our
model. A next limitation is that our data do not cover the
competition. This concern is, however, alleviated because the
retailers that we study are spread over the Netherlands and they
do not compete for the same customers. Finally, we focus on
CTA DMs targeted to existing customers (i.e., customers that
already made a purchase at the company in the past). Studies
that focus on how CTA DMs may help with acquiring new
customers would complement our study.
Acknowledgment
We thank Kay Peters of the Department of Marketing of the
University of Hamburg (Germany) and Anne ter Braak of the
Department of Marketing of KU Leuven (Belgium) for their
feedback. We also thank the participants and faculty of the
EMAC doctoral consortium 2017 in Groningen (the
Netherlands).
Appendix A
Result of model without controlling for endogeneity.
Variables Coef. SE
Intercept 1.117** 0.024
CTA_DM 0.065** 0.006
Non-monetary incentive
Hedonic 0.183** 0.016
Utilitarian 0.195** 0.012
Monetary incentive 0.018 0.009
CTA_DM * relationship history
*Purchase_frequency 0.092** 0.020
*Purchase_recency 0.035** 0.001
*Relationship_duration 0.013** 0.001
CTA_DM * DM history
*DM_frequency 0.103** 0.027
*DM_recency 0.009** 0.001
CTA_DM * Socio-demographics
*Age 0.004** b0.001
*Gender (male = 1) 0.011 0.006
Relationship history
Purchase_frequecny 1.009** 0.008
Purchase_recency 0.023** 0.001
Relationship_duration 0.006** 0.001
DM history
DM_frequency 0.256** 0.008
DM_recency 0.006** b0.001
Socio-demographics
Age 0.072** 0.007
Gender (male = 1) 0.100** 0.005
Control variables
Value_incentive 0.0003* 0.000
Relational_DM 0.143** 0.005
Relational_DM_frequency 0.009* 0.005
Relational_DM_recency 0.009** b0.001
Store_DM 0.084** 0.007
Store_DM_frequency 0.073** 0.008
Store_DM_recency b0.001 b0.001
AVG CTA_DM 3.421 0.027
Lag_CTA_DM 0.054** 0.004
Lag_Relational_DM 0.022** 0.005
Lag_Store_DM 0.023** 0.006
GDP_growth 0.002 0.003
78 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
(continued)
Variables Coef. SE
Quarter
2 0.052** 0.003
3 0.051** 0.003
4 0.033** 0.004
Year
2010 0.045** 0.005
2011 0.080** 0.007
2012 0.177** 0.009
2013 0.255** 0.010
2014 0.323** 0.013
2015 0.392** 0.015
2016 0.507** 0.017
*pb.05 (two-sided) **pb.01 (two-sided).
Because of space constraints, the results of the coefficients of the retailer
dummies are omitted.
References
Ansari, Asim, Carl F. Mela, and Scott A. Neslin (2008), Customer Channel
Migration,Journal of Marketing Research, 45, 1, 6076.
Arora, Raj and Charles Stoner (1992), Variables Influencing Direct Mail
Response Rate: A Meta-Analysis,Journal of Marketing Management,2,1,
715.
Bagozzi, Richard P. (1995), Reflections on Relationship Marketing in
Consumer Markets,Journal of the Academy of Marketing Science, 23, 4,
2737.
Barnes, Nora Ganim and Michael P. Peters (1982), Modes of Retail
Distribution-Views of the Elderly,Akron Business and Economic Review,
13, 3, 2631.
Bawa, Kapil and Robert W. Shoemaker (1987), The Effects of a Direct Mail
Coupon on Brand Choice Behavior,Journal of Marketing Research, 24, 4,
3706.
Blattberg, Robert C. and Scott Neslin (1990), Sales Promotion: Concepts,
Methods, and Strategies. Prentice Hall.
Brehm, Jack W. (1966), A Theory of Psychological Reactance. New York (N.
Y.): Academic Press.
Bolton, Ruth N. (1998), A Dynamic Model of the Duration of the Customer's
Relationship with a Continuous Service Provider: The Role of Satisfaction,
Marketing science, 17, 1, 4565.
Bult, Jan Roelf, Hiek van der Scheer, and Tom Wansbeek (1997), Interaction
Between Target and Mailing Characteristics in Direct Marketing, with an
Application to Health Care Fund Raising,International Journal of
Research in Marketing, 14, 4, 3018.
Campbell, Margaret C. and Kevin Lane Keller (2003), Brand Familiarity and
Advertising Repetition Effects,Journal of Consumer Research, 30, 2,
292304.
Chitturi, Ravindra, Rajagopal Raghunathan, and Vijay Mahajan (2007), Form
Versus Function: How the Intensities of Specific Emotions Evoked in
Functional Versus Hedonic Trade-Offs Mediate Product Preferences,
Journal of Marketing Research, 44, 4, 70214.
D'Astous, Alain and Isabelle Jacob (2002), Understanding Consumer
Reactions to Premium-based Promotional Offers,European Journal of
Marketing, 36, 11/12, 127086.
Datta, Hannes, Bram Foubert, and Harald J. Van Heerde (2015), The
Challenge of Retaining Customers Acquired with Free Trials,Journal of
Marketing Research, 52, 2, 21734.
De Wulf, Kristof and Gaby Odekerken-Schröder (2003), Assessing the Impact
of a Retailer's Relationship Efforts on Consumer's Attitudes and Behavior,
Journal of Retailing and Consumer Services, 10, 2, 95108.
Diamond, W.D. and A. Sanyal (1990), The Effect of Framing on The Choice
of Supermarket Coupons,Personality and Social Psychology Bulletin, 16,
3, 496510.
Direct Marketing Association (2015), Statistical Fact Book The Definitive
Source for Marketing's Most Important Benchmarks. New York, USA.
——— (2017), Statistical Fact Book The Definitive Source for Marketing's
Most Important Benchmarks. New York, USA. Statistical Fact Book The
Ultimate Source for Data-Driven Marketing Insight.
Donkers, Bas, Philip Hans Franses, and Peter C. Verhoef (2006), Selective
Sampling for Binary Choice Models,Journal of Marketing Research, 40,
4, 4927.
Drèze, Xavier and André Bonfrer (2008), An Empirical Investigation of the
Impact of Communication Timing on Customer Equity,Journal of
Interactive Marketing, 22, 1, 3650.
Ebbes, P., D. Papies, and H.J. Van Heerde (2011), The Sense and Non-sense of
Holdout Sample Validation in the Presence of Endogeneity,Marketing
Science, 30, 6, 111522.
Fader, Peter S. and Bruce G.S. Hardie (2007), How to Project Customer
Retention,Journal of Interactive Marketing, 21, 1, 7690.
Feld, Sebastian, Heiko Frenzen, Manfred Krafft, Kay Peters, and Peter C.
Verhoef (2013), The Effects of Mailing Design Characteristics on Direct
Mail Campaign Performance,International Journal of Research in
Marketing, 30, 2, 14359.
Fitzsimons, Gavan J. and Donald R. Lehmann (2004), Reactance to
Recommendations: When Unsolicited Advice Yields Contrary Responses,
Marketing Science, 23, 1, 8294.
Foubert, Bram, Els Breugelmans, Karen Gedenk, and Charlotte Rolef (2017),
Something Free or Something Off? A Comparative Study of the Purchase
Effects of Premiums and Price Cuts,Journal of Retailing, 94, 1, 520.
Gázquez-Abad, Juan Carlos, Marie Hélène De Canniére, and Francisco J.
Martínez-López (2011), Dynamics of Customer Response to Promotional
and Relational Direct Mailings from an Apparel Retailer: the Moderating
Role of Relationship Strength,Journal of Retailing, 87, 2, 16681.
Gedenk, Karen, Scott A. Neslin, and Kusum L. Ailawadi (2010), Sales
Promotion,Retailing in the 21st Century, Berlin, Heidelberg: Springer.
Germann, Frank, Peter Ebbes, and Rajdeep Grewal (2015), The Chief
Marketing Officer Matters!,Journal of Marketing, 79, 3, 122.
Gilly, Mary C. and Valarie A. Zeithaml (1985), The Elderly Consumer and
Adoption of Technologies,Journal of Consumer Research, 12, 3,
3537.
Godfrey, Andrea, Kathleen Seiders, and Glenn B. Voss (2011), Enough Is
Enough! The Fine Line in Executing Multichannel Relational Communi-
cation,Journal of Marketing, 75, 4, 94109.
Gönül, Füsun F. and Frenkel Ter Hofstede (2006), How to Compute Optimal
Catalog Mailing Decisions,Marketing Science, 25, 1, 6574.
Greene, W.H. (2003), Econometric Analysis. Pearson Education India.
Grewal, Rajdeep, Raj Mehta, and Frank R. Kardes (2004), The Timing of
Repeat Purchases of Consumer Durable Goods: The Role of Functional
Bases of Consumer Attitudes,Journal of Marketing Research, 41, 1,
10115.
Hair, J.F., W.C. Black, B.J. Babin, R.E. Anderson, and R.L. Tatham (2010),
Multivariate Data Analysis. Pearson.
Heilman, Carrie M., Kent Nakamoto, and Ambar G. Rao (2002), Pleasant
Surprises: Consumer Response to Unexpected In-Store Coupons,Journal
of Marketing Research, 39, 2, 24252.
Jayachandran, Satish, Subhash Sharma, Peter Kaufman, and Pushkala Raman
(2005), The Role of Relational Information Processes and Technology Use
in Customer Relationship Management,Journal of Marketing, 69, 4,
17792.
Leclerc, Fratice (1997), Monetary Promotions vs. Non-Monetary Promotions,
Which Is More Effective?Advances in Consumer Research, 24, 2201.
Liu, Yuping (2007), The Long-Term Impact of Loyalty Programs on
Consumer Purchase Behavior and Loyalty,Journal of Marketing, 71, 4,
1935.
Melis, Kristina, Katia Campo, Els Breugelmans, and Lien Lamey (2015), The
Impact of the Multi-Channel Retail Mix on Online Store Choice: Does
Online Experience Matter?Journal of Retailing, 91, 2, 27288.
Mittal, Banwari (1989), Measuring Purchase-Decision Involvement,Psy-
chology & Marketing, 6, 2, 14762.
79S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
Naik, Prasad A. and Nanda Piersma (2002), Understanding the Role of
Marketing Communication in Direct Marketing,Econometric Instutite
Report, EI 2002-13, Rotterdam: Erasmus Univeristy.
Neslin, Scott A., Gail Ayala Taylor, Kimberly D. Grantham, and Kimberly R.
McNeil (2013), Overcoming The recency Trapin Customer Relationship
Management,Journal of the Academy of Marketing Science, 41, 3,
32037.
Nunes, Joseph C. and C. Whan Park (2003), Incommensurate Resources: Not
Just More of the Same,Journal of Marketing Research, 40, 1, 2638.
Okada, Erica Mina (2005), Justification Effects on Consumer Choice of
Hedonic and Utilitarian Goods,Journal of Marketing Research, 42, 1,
4353.
Palazon, Mariola and Elena Delgado-Ballester (2009), Effectiveness of Price
Discounts and Premium Promotions,Psychology & Marketing, 26, 12,
110829.
Delgado-Ballester, Elena (2013), Hedonic or Utilitarian Premiums: Does It
Matter?European Journal of Marketing, 47, 8, 125675.
Prins, Remco and Peter C. Verhoef (2007), Marketing Communication Drivers
of Adoption Timing of a New E-Service,Journal of Marketing, 71, 2,
16983.
Reinartz, Werner J. and V. Kumar (2003), The Impact of Customer
Relationship Characteristics on Profitable Lifetime Duration,Journal of
Marketing, 67, 1, 7799.
Rust, Roland T. and Peter C. Verhoef (2005), Optimizing the Marketing
Interventions Mix in Intermediate-Term CRM,Marketing Science, 24, 3,
47789.
Risselada, Hans, Peter C. Verhoef, and Tammo H.A. Bijmolt (2014), Dynamic
Effects of Social Influence and Direct Marketing on the Adoption of High-
Technology Products,Journal of Marketing, 78, 2, 5268.
Saad, Gad and Tripat Gill (2000), Applications of Evolutionary Psychology in
Marketing,Psychology and Marketing, 17, 12, 100534.
Sheth, Jagdish N. and Atul Parvatlyar (1995), Relationship Marketing in
Consumer Markets: Antecedents and Consequences,Journal of the
Academy of Marketing Science, 23, 4, 25571.
Simonson, Itamar, Z.I.V. Carmon, and Suzanne O. Curry (1994), Experimen-
tal Evidence on the Negative Effect of Product Features and Sales
Promotions on Brand Choice,Marketing Science, 13, 1, 2340.
Simpson, James and Helen Mortimore (2015), The Influence of Direct Mail
Marketing on Buyer Purchasing Decisions: A Aualitative Analysis of
Perceptions by Age Froup,Journal of Research Studies in Business and
Management, 1, 1, 11942.
Sinha, I. and M.F. Smith (2000), Consumers' Perceptions of Promotional
Framing of Price,Psychology & Marketing, 3, 25775.
Smith, Jacqui and Paul B. Baltes (1990), Wisdom-Related Knowledge: Age/
cohort Differences in Response to Life-Planning Problems,Developmental
Psychology, 26, 3, 494505.
Spiller, Stephen A., Gavan J. Fitzsimons, John G. Lynch, and Gary H.
McClelland (2013), Spotlights, Floodlights, and the Magic Number Zero:
Simple Effects Tests in Moderated Regression,Journal of Marketing
Research, 50, 2, 27788.
Stock, J. and M. Yogo (2005), Asymptotic Distributions of Instrumental
Variables Statistics with Many Instruments, Vol. 6.
Swann, William B. and Michael J. Gill (1997), Confidence and Accuracy in
Person Perception: Do We Know What We Think We Know About Our
Relationship Partners?Journal of Personality and Social Psychology, 73,
4, 74757.
Thaler, Richard (1985), Mental Accounting and Consumer Choice,
Marketing Science,4,3,199214.
Thomas, Jacquelyn S., Robert C. Blattberg, and Edward J. Fox (2004),
Recapturing Lost Customers,Journal of Marketing Research, 41, 1, 3145.
Thomas, Suman Ann, Shanfei Feng, and Trichy V. Krishnan (2015), To
Retain? To Upgrade? The Effects of Direct Mail on Regular Donation
Behavior,International Journal of Research in Marketing, 32, 1, 4863.
Van Diepen, Merel, Bas Donkers, and Philip Hans Franses (2009), Dynamic
and Competitive Effects of Direct Mailings: A Charitable Giving
Application,Journal of Marketing Research, 46, 1, 12033.
Verhoef, Peter C., Philip Franses, and Janny C. Hoekstra (2002), The Effect of
Relational Constructs on Customer Referrals and Number of Services
Purchased from a Multiservice Provider: Does Age of Relationship
Matter?Journal of the Academy of Marketing Science, 30, 3, 20216.
——— (2003), Understanding the Effect of Customer Relationship Manage-
ment Efforts on Customer Retention and Customer Share Development,
Journal of Marketing, 67, 4, 3045.
Wooldridge, J.M. (2010), Econometric Analysis of Cross Section and Panel
Data. MIT press.
80 S. Vafainia et al. / Journal of Interactive Marketing 45 (2019) 6280
... Moreover, companies can contacts their customer base through multiple DM channels, such as direct mail, direct response advertising, telemarketing, email, and sales personnel, to inform customers about the multitude of options available for purchasing products (Kumar & Venkatesan, 2005;Rettie, Grandcolas, & Deakins, 2005), to reduce the uncertainty and the effort surrounding a decision (Vafainia, Breugelmans, & Bijmolt, 2019). ...
... For Gázquez-Abad et al. (2011), the objectives of DM are categorized as cognitive (transferring information, brand awareness), affective (image building) and behavioral (accomplishing sales or information inquiries). In addition, according to Vafainia, Breugelmans, & Bijmolt (2019), DM communications positively influence consumer behavior because retailers investments in customer relationships result in psychological bonding are perceived by the customers. The authors conclude that, as customers increase their confidence associated with the decision after the DM campaigns, they feel obliged to return "good for good" by making a purchase. ...
... One of the uses of DM is to encourage consumers to react immediately by including a specific 'call to action' (CTA) with short-term rewards that can create interest in a product/service and thereby lead to a final purchase (Gázquez-Abad et al., 2011). CTA DMs that include an incentive have a higher positive impact on customer's purchase compared with those without an incentive (Vafainia, Breugelmans, & Bijmolt, 2019). ...
Thesis
Mobile adoption has seen an exponential increase worldwide in recent years, which has also been reflected in the number of mobile applications, or apps. Many companies have developed their own apps to display their brand entities and have started using them as a direct marketing channel with push notifications, messages displayed on their mobile device screen. This research seeks to understand the extent of the impact on purchasing behavior of push notifications sent by branded applications from Portuguese supermarket chains. It also aims to segment the results to different buyer groups based on psychographic variables. This research was conducted by adopting a quantitative approach and non-probabilistic sampling. With an online questionnaire, 265 responses were collected and treated with statistical procedures such as descriptive, multivariate and cluster analysis. The findings confirm that customers who received push notifications had a higher average monthly spend in supermarkets than customers with branded apps who received no push notifications. However, these results were observed only for buyers deeply involved in all attributes of the purchase and for buyers with a greater concern for the products. Academically, this dissertation contributes to the discussion of push notifications as a direct marketing channel and the effects of customer characteristics on the effect of direct marketing. This research also provides insight to the marketer, stating that push notifications can be used to drive users' offline purchases.
... Therefore, marketers usually offer monetary incentives to brand fans to increase their enthusiasm and nurture the brand community [14,63]. At the same time, firms providing monetary incentive to customers could make customers perceive their extra effort and resource investment, which results in consumers' psychological bonding [64]. Accordingly, a brand can enable its fans to fulfill their monetary needs as it keeps its promise by offering incentives to them; and then, the monetary needs fulfillment may result in brand fans being more satisfied with the host brand. ...
Article
Full-text available
This study adopts a brand relationship quality (BRQ) perspective to reveal the reason firms’ investments in social media-based brand communities should increase their social relationship marketing performances. An empirical analysis with 234 Facebook users who joined brand communities was conducted to examine the proposed hypotheses, revealing that fan needs fulfillments—information, entertainment, social interaction, and monetary ones—had positive effects on BRQ. Further, BRQ was found to have positive effects on fans’ engagement behavioral intentions toward brands, including willingness to buy, member continuance intention, and electronic word of mouth intention. This study contributes to existing research that indicates a new mechanism of BRQ improvement via the social media-based brand community. Implications corresponding to the research findings as well as study limitations and future directions are also addressed.
... In our evaluation, we consider multiple scenarios that were executed in four weeks with relevant variations in a number of exogenous factors. Following previous findings in the literature, in our analysis, we explicitly consider if the products recommended in the email are on promotion (Reibstein, 2002;Vafainia, Breugelmans, & Bijmolt, 2019) and whether there was a banner for the corresponding category on the homepage of the retailer (Goic, Álvarez, & Montoya, 2018). Additionally, we consider dummy variables that indicate if the delivery occurred in the morning and if the national soccer team was playing in the World Cup, which might reduce the likelihood of shopping on that day. ...
Article
Triggered emails are personalized messages that are automatically sent as a response to specific actions or states of customers. Typical examples of this type of campaign include cross-selling recommendations, cart abandonment reminders, and re-engagement emails. Despite the widespread growth of these strategies, there has been no formal evaluation of their effectiveness. This paper investigates the impact of one type of triggered email campaign by using an experimental approach. We identify customers who had recently browsed the website of a multichannel retailer but had abandoned the process before making a purchase. Approximately half of the sample was randomly selected to receive automated emails with different configurations, while the other half receive no message at all. Comparison of the sales of these two groups indicates that browse abandonment emails have increase revenues in the online channel and in the triggered category. In terms of the design of the campaign, we found that the implementation of triggered emails plays an important role in their effectiveness. In this regard our result indicates that retargeting based on longer navigation histories is associated with larger conversions and that recommendations of wider assortments are associated with larger revenues.
... In addition, these mechanisms could also enhance store switching, stimulating consumers to visit the store that runs the promotion instead of other stores (Chandon et al., 2000). Several empirical studies have reported the positive impact of promotions on store visits and the number of buyers (e.g., Guyt & Gijsbrechts, 2020;Vafainia et al., 2019). Therefore, a VAT-free promotion is expected to lead to a positive transaction effect. ...
Article
Full-text available
This study examines the impact of value-added tax (VAT)–free promotions on sales performance. VAT-free promotions are a recently adopted form of price promotions where consumers are exempted from paying the VAT amount across almost all products in the assortment during a limited number of days. They are typically organized once per year and surrounded by a large amount of media attention from the involved retailer. To test the effects of this promotion on store and category sales, and investigate the differences between loyalty program (LP) members and non-LP members, the authors use scanner data from a Dutch durable goods retailer across a range of categories. The results show that VAT-free promotions positively impact store performance. Moreover, the findings indicate that more non-LP members are attracted to the store and that they increase the amount they spend in the store. While LP members also spend more in the store, this increase in shopping basket size does not compensate for the significant drop in the number of LP members that visit the store, leading to an overall decrease in sales coming from LP members during VAT-free days. We furthermore find that the positive effect of VAT-free promotions for non-LP members (rather than LP members) generalizes across all investigated categories. Our results provide key insights for retailers and direct marketers with regard to the effectiveness of VAT-free promotions in order to strategically segment the customer base.
Article
Social media, and in particular Instagram, can provide a suitable communication channel to achieve customer inspiration. However, it is essential to implement the right strategy. This research aims to advance in the concept of social media customer inspiration by analysing three dimensions: inspiration by, inspiration to buy, and inspiration to brand community. To this end, the effect of different executions that combine the call to action with the visual content of the brand (brand name presence) and the fragmentation of the content (publication in carousel) is explored. Three experimental studies are conducted in studies in three sectors (travel, food and beverages), with different targets (brand followers and non-followers) and in two Instagram communication formats (feed posts and stories). Empirical results show interesting implications that contribute significantly to advancing both theory and practice. Supplemental data for this article is available online at https://doi.org/10.1080/02650487.2021.2014702 .
Article
Purpose This quantitative research paper examined factors influencing re-engagement intentions and re-engagement behavior of lapsed health club members and identified which sales promotion incentives are most effective in re-engaging this lapsed member market. While previous studies examined re-engagement intentions and re-engagement behavior in isolation, no research could be located which examined both simultaneously and the impact of promotion incentives in the health and fitness industry. Design/methodology/approach Study A (mail survey) examined re-engagement intentions of one hundred lapsed members of a medium size suburban health club. Study B (an experimental field study) measured actual true re-engagement behavior using price and gift token as incentives to re-engage 300 lapsed members of the same health club. Ten hypotheses were tested using chi-square, logistic regression and correlation analysis. Findings Results indicate that price is the most influential incentive in promoting re-engagement intentions, while the experimental field study tested this survey finding and reported that although incentives were more effective than nonincentives in terms of re-engagement behavior, price played a far greater significant role than gift token in actual re-engagement behavior of lapsed members. Highly satisfied customers are likely to have stronger re-engagement intentions with frequent attendees more influenced by price in terms of re-engagement intentions, while those who lapsed most recently demonstrated stronger re-engagement behavior. Research limitations/implications Study A: While acknowledging that the sample size was limited, a respectable response rate of 49% was recorded for the mail survey in Study A but a greater response rate and larger sample size would have allowed for a more comprehensive analysis. However, the study is primarily exploratory in nature and serves to achieve the objectives of the research. The study was based on a medium size suburban health club, therefore no comparisons can be drawn between the results of this study and that involving larger health clubs, chains, urban clubs or those outside of Ireland. Although the inclusion of moderating effects was considered, it was not feasible to develop these interactions due to the limited sample size. Study B: Generalization of the findings in this study with other international markets is not possible due to differences in demographic factors, promotions and differences in the health and fitness industry. No comparison can be drawn either between health clubs which operate in a different environment, such as public or nonmembership-based clubs, urban clubs or those constituting a chain. While the offer incentives were of equal value, it is recognized that the gift token may have appealed more to females than to males. Practical implications Health clubs should consider changing strategic focus, from being overly concentrated on new customer acquisition to actively re-engaging the lapsed member market by finding out why they left and offer an incentive such as price to re-engage. Although loyalty, duration, age, gender and income were found not to be significant in this study, satisfaction, frequency and recency of lapse were significant. Therefore, health clubs should strive to keep members satisfied, monitor and increase frequency of attendance with creative programming and commence the re-engagement process prior to the membership expiry so as to maximize re-engagement and customer retention using price discount as an incentive. Originality/value The originality of this study is that is tests statistically consumer re-engagement intentions and actual re-engagement behavior simultaneously in a health club setting using a mail survey in Study A and Experimental Design in Study B.
Article
Full-text available
Many service firms acquire customers by offering free-trial promotions. However, a crucial challenge is to retain the customers acquired with these free trials. To address this challenge, firms need to understand how free-trial customers differ from regular customers in terms of their decisions to retain the service. This article conceptualizes how marketing communication and usage behavior drive customers' retention decisions and develops hypotheses about the impact of free-trial acquisition on this process. To test the hypotheses, the authors model a customer's retention and usage decisions, distinguishing usage of a flat-rate service and usage of a pay-per-use service. The model allows for unobserved heterogeneity and corrects for selection effects and endogeneity. Using household panel data from a digital television service, the authors find systematic behavioral differences that cause the average customer lifetime value of free-trial customers to be 59% lower than that of regular customers. However, free-trial customers are more responsive to marketing communication and usage rates, which offers opportunities to target marketing efforts and enhance retention rates, customer lifetime value, and customer equity.
Article
Full-text available
The authors examine the effects of a manufacturer coupon on brand choice behavior. The level of coupon redemption and changes in brand choice behavior after redemption are examined as a function of the household's prior probability of purchasing the promoted brand, likelihood of buying a favorite competitive brand, and coupon face value. A model of the coupon redemption decision is developed to predict response to the coupon promotion by different consumer segments. Predictions from the model are tested by using scanner panel data from a field experiment on coupon face values. Coupon redemption rates are found to be much higher among households that have purchased the brand on a regular basis in the past. The results also suggest that most consumers revert to their precoupon choice behavior immediately after their redemption purchase. These and other findings have important implications for the profitability of coupon promotions.
Article
The authors examine the effects of a manufacturer coupon on brand choice behavior. The level of coupon redemption and changes in brand choice behavior after redemption are examined as a function of the household's prior probability of purchasing the promoted brand, likelihood of buying a favorite competitive brand, and coupon face value. A model of the coupon redemption decision is developed to predict response to the coupon promotion by different consumer segments. Predictions from the model are tested by using scanner panel data from a field experiment on coupon face values. Coupon redemption rates are found to be much higher among households that have purchased the brand on a regular basis in the past. The results also suggest that most consumers revert to their precoupon choice behavior immediately after their redemption purchase. These and other findings have important implications for the profitability of coupon promotions.
Article
As an alternative to promotional price cuts, retailers and manufacturers often rely on non-price promotion techniques, such as premium promotions, where consumers receive a free gift with the purchase of a product. We compare the effectiveness of premiums to that of price cuts, and study moderators of this comparative premium effectiveness. We use data from a large online shopping simulation study with more than 2,000 participants to model consumers' purchase decisions in response to premiums and price cuts. Results indicate that the impact of premiums on purchase behavior is systematically lower than that of equivalent price cuts. However, a premium's smaller sales impact may be offset by a cost advantage. This is especially true for private label brands where the premium's purchase effects do not differ too much from those of a price cut. We calculate how large the cost advantage has to be for a premium to be more profitable than a price cut, and show that premiums entail risks as well as opportunities, for both manufacturers and retailers.
Article
Market response models based on field-generated data need to address potential endogeneity in the regressors to obtain consistent parameter estimates. Another requirement is that market response models predict well in a holdout sample. With both requirements combined, it may seem reasonable to subject an endogeneity- corrected model to a holdout prediction task, and this is quite common in the academic marketing literature. One may be inclined to expect that the consistent parameter estimates obtained via instrumental variables (IV) estimation predict better than the biased ordinary least squares (OLS) estimates. This paper shows that this expectation is incorrect. That is, if the holdout sample is similar to the estimation sample so that the regressors are endogenous in both samples, holdout sample validation favors regression estimates that are not corrected for endogeneity (i.e., OLS) over estimates that are corrected for endogeneity (i.e., IV estimation). We also discuss ways in which holdout samples may be used sensibly in the presence of endogeneity. A key takeaway is that if consistent parameter estimates are the primary model objective, the model should be validated with an exogenous (rather than endogenous) holdout sample.
Article
Marketing academics and practitioners alike remain unconvinced about the chief marketing officer's (CMO's) performance implications. Whereas some studies propose that firms benefit financially from having a CMO in the C-suite, other studies conclude that the CMO has little or no effect on firm performance. Accordingly, there have been strong calls for additional academic research regarding the CMO's performance implications. In response to these calls, the authors employ model specifications with varying identifying assumptions (i.e., rich data models, unobserved effects models, instrumental variable models, and panel internal instruments models) and use data from up to 155 publicly traded firms over a 12-year period (2000-2011) to find that firms can indeed expect to benefit financially from having a CMO at the strategy table. Specifically, their findings suggest that the performance (measured in terms of Tobin's q) of the sample firms that employ a CMO is, on average, approximately 15% greater than that of the sample firms that do not employ a CMO. This result is robust to the type of model specification used. Marketing academics and practitioners should find the results intriguing given the existing uncertainty surrounding the CMO's performance implications. The study also contributes to the methodology literature by collating diverse empirical model specifications that can be used to model causal effects with observational data into a coherent and comprehensive framework.
Article
As an alternative to promotional price cuts, retailers and manufacturers often rely on non price promotion techniques, such as premium promotions, where consumers receive a free gift with the purchase of a product. We compare the effectiveness of premiums to that of price cuts, and study moderators of this comparative premium effectiveness. We use data from a large online shopping simulation study with more than 2,000 participants to model consumers’ purchase decisions in response to premiums and price cuts. Results indicate that the impact of premiums on purchase behavior is systematically lower than that of equivalent price cuts. However, a premium’s smaller sales impact may be offset by a cost advantage. This is especially true for private label brands where the premium’s purchase effects do not differ too much from those of a price cut. We calculate how large the cost advantage has to be for a premium to be more profitable than a price cut, and show that premiums entail risks as well as opportunities, for both manufacturers and retailers.