Content uploaded by Nikolaus Franke
Author content
All content in this area was uploaded by Nikolaus Franke on Feb 13, 2014
Content may be subject to copyright.
103
Journal of Marketing
Vol. 73 (September 2009), 103–121
© 2009, American Marketing Association
ISSN: 0022-2429 (print), 1547-7185 (electronic)
Nikolaus Franke, Peter Keinz, & Christoph J. Steger
Testing the Value of Customization:
When Do Customers Really Prefer
Products Tailored to Their
Preferences?
Recently, researchers have paid increasing attention to the marketing strategy of customization. A key assumption
is that customized products create higher benefits for customers than standard products because they deliver a
closer preference fit. The prerequisite for this effect is the ability to obtain precise information on what customers
actually want. But are customers able to specify their preferences that precisely? Several theoretical arguments
raise doubts about this, implicitly challenging the value of customization. The authors conduct two studies in which
they find that products customized on the basis of expressed preferences bring about significantly higher benefits
for customers in terms of willingness to pay, purchase intention, and attitude toward the product than standard
products.The benefit gain is higher if customers have (1) better insight into their own preferences, (2) a better ability
to express their preferences, and (3) greater product involvement. This suggests that customization has the
potential to be a powerful marketing strategy if these conditions are met. In the opposite case, firms willing to serve
heterogeneous customer preferences need to adapt their customization systems in such a way that they explicitly
address the customers’ inability to provide valid preference information.
Keywords
: mass customization, self-design, toolkits for user innovation and design, heterogeneous preferences,
segmentation
Nikolaus Franke is a professor and Director of the Institute for Entrepre-
neurship and Innovation (e-mail: nikolaus.franke@wu.ac.at), Peter Keinz
is an assistant professor (e-mail: peter.keinz@wu.ac.at), and Christoph J.
Steger is a lecturer (e-mail: christoph.steger@wu.ac.at), Institute for
Entrepreneurship and Innovation, Vienna User Innovation Research Initia-
tive, Vienna University of Economics and Business (WU Wien). The
authors thank Thomas Funke, Andreas Pinterits, Martin Kober, and the
students in the Entrepreneurship and Innovation research course at WU
Wien (summer 2006), especially Paul Barborka, for their valuable contri-
butions, as well as MindTake New Media Consulting and its chief execu-
tive officer, Klaus Oberecker, for providing the online survey. Finally, the
authors thank the two anonymous
JM
reviewers for their valuable com-
ments, and the Jubilaeumsfonds der Oesterreichischen Nationalbank for
funding this research project.
In this article, we measure the benefits customers derive
from customized products and the factors that affect the
magnitude of the benefits achieved. In recent years,
researchers and practitioners have paid increasing attention
to the marketing strategy of customization (e.g., Dellaert
and Stremersch 2005; Gilmore and Pine 2000; Kotha 1995;
Syam, Ruan, and Hess 2005; Varki and Rust 1998). These
efforts have been driven by both the supply side and the
demand side. There is a constantly increasing supply of
technology that facilitates customization. The production
costs for individualized offerings are declining (Duray et al.
2000; Kahn 1998; Peppers and Rogers 1997), and the Inter-
net has led to a decline in the costs of communication with
customers (Ansari and Mela 2003; Sheth, Sisodia, and
Sharma 2000). Research has also found ways to reduce the
customer effort required (Alba et al. 1997; Diehl, Kornish,
and Lynch 2003; Randall, Terwiesch, and Ulrich 2005,
2007). At the same time, the customers’ demand for cus-
tomized products has increased, and customer preferences
have become increasingly heterogeneous in many markets
(Gilmore and Pine 1997; Smith 1956). Scholars and practi-
tioners alike have high expectations regarding the promise
of customization (Ansari and Mela 2003; Sheth and Sisodia
1999). As Simonson (2005, p. 42) notes, “It has been
assumed in recent years that the age-old practice of target-
ing market segments is dominated and will be displaced by
individual marketing.”
However, some scholars have questioned the merits of
customization because it requires extensive customer par-
ticipation (e.g., Fang 2008; Huffman and Kahn 1998;
Simonson 2005; Zipkin 2001). Spectacular failures in cus-
tomization, such as Levi Strauss’s “Original Spin” jeans
and Mattel’s “My Design Barbie” (see Franke and Piller
2004), appear to support these doubts. Moreover, customers
sometimes prefer the default configurations provided by the
producer and fail to recognize the opportunities offered
(Hill 2003).
Delivering positive value to the customer is a prerequi-
site for the long-term success of any customization strategy.
Therefore, several scholars have begun to analyze the bene-
fits customization strategies create for customers. However,
empirical findings have yielded mixed results. Franke and
Piller (2004), Schreier (2006), and Franke and Schreier
104 / Journal of Marketing, September 2009
(2008, 2009) compare students’ willingness to pay (WTP)
for both standard and customized products in different
low-price consumer goods categories and find a higher WTP
for the customized products. Schoder and colleagues (2006)
measure consumer acceptance of traditional and customized
newspapers using conjoint analysis and conclude that
people prefer customized newspapers but are not willing to
pay more for them than for traditional ones, thus question-
ing the benefits of customization. The stimuli provided in
that study were relatively abstract (operationalized as verbal
stimuli: “regular newspaper” versus “personalized news-
paper”). Bardakci and Whitelock (2004) investigate con-
sumer agreement or disagreement with statements related to
the benefits and disadvantages of customized cars. The
results show that though people seem to be interested in
customized products, only 58% of the participants claimed
to be willing to pay a slight premium for a customized car.
Again, “customization” was only given as an abstract repre-
sentation, meaning that the participants did not actually
experience customized products, which casts doubt on the
validity of the findings.
These conflicting findings call for further systematic
analysis of the benefits created by customization compared
with other strategies and of the conditions under which
those benefits take effect. When possible, it would appear
useful to employ controlled experiments and actual prod-
ucts as experimental stimuli to overcome the limitations of
abstract stimuli. To permit general conclusions, a broad and
representative sample is required because special-interest or
convenience samples might differ from the mass consumer
market in important aspects.
In this research project, we attempt to take this next step
by empirically analyzing the benefits of customization
(compared with segmentation and mass marketing) and the
variables that moderate these benefits. For this purpose, we
conducted two studies. Study 1 is an experiment with a rep-
resentative sample of 1589 participants in which we simu-
lated the three different strategies of customization, seg-
mentation, and mass marketing and then measured the
customers’ resulting subjective benefit attributions. We
selected the product category of newspapers because the
strategy of customization is currently being discussed in
this area (Schoder et al. 2006). We find that newspapers
individualized on the basis of stated customer preferences
generate significantly higher benefits for customers in terms
of WTP, purchase intention, and attitude toward the prod-
uct. We also find that the magnitude of such benefits
increases in line with the person’s insight into his or her
own preferences and ability to express them.
Study 2 aims to complement this research by enhancing
the validity and generalizability of Study 1. For this pur-
pose, we used a different methodology, comprising a sur-
vey, latent construct measurement, and structural equation
modeling. Again, we used a representative sample (1039
participants). In addition to replicating the analysis in the
newspaper market, we tested our hypotheses in four diverse
product categories: fountain pens, kitchens, skis, and cere-
als. The findings in the newspaper market are similar to
Study 1, thus validating our findings. In the other product
fields, we also find that customization results in increased
benefits for customers. Again, the magnitude of the benefits
created is affected by the person’s insight into his or her
preferences and ability to express them. Moreover, we find
that product involvement affects the benefits obtained from
customization in these five product categories.
Research Hypotheses
The Benefits of Customized Products
It is a straightforward economic argument that a closer fit
between preferences and product attributes brings about
increased benefits for the customer (see Simonson 2005).
The crux of the matter is that true customer preferences
may differ from expressed preferences on the basis of
which the product is customized. As with any measurement
task, the companies’ measurement of customer preferences
is likely to contain a true component and an error term or
random component (Morrison and Silva-Risso 1995). To
ensure valid measurement, a customer must be aware of his
or her preferences and must be able to express them
(Simonson 2005). However, consumer preferences are often
poorly defined. Research on consumer decision making (see
Bettman, Luce, and Payne 1998; Fischhoff 1991; Slovic
1995; Yoon and Simonson 2008) suggests that preferences
are often constructed on the basis of how the options are
framed when the consumer needs to make a decision. Given
the nearly infinite solution spaces of even simple mass cus-
tomization configurators, Huffman and Kahn (1998) warn
that the preferences consumers construct when faced with
the challenge of defining a product are almost certainly a
local optimum, if not an almost random choice; the authors
use the term “mass confusion” for such futile mass cus-
tomization endeavors. Thus, it is not easy to transfer
“sticky” customer preference information from a customer
to a producing company (Morrison, Roberts, and Von Hip-
pel 2000; Von Hippel 1994). On the whole, the relevant lit-
erature indicates that the measurement of an individual cus-
tomer’s preferences might contain a large error term.
In such a case, it is questionable whether customized
products are more beneficial to customers than standard
products (Kramer 2007). It could even be argued that in
extreme cases, an adept segmentation strategy might yield
higher benefits for customers than customization. Segmen-
tation involves grouping customers with similar preferences
and then targeting them in the same manner. If these seg-
ments capture the true (common) component of the prefer-
ences within the segment and eliminate random error
through aggregation, the true component of the preferences
is revealed. In extreme cases, such a “wisdom-of-the-
crowd” effect (Surowiecki 2004) could yield a segment-
specific product that generates even higher benefits for cus-
tomers within the segment than products customized on the
basis of erroneous individual preference statements.
Kramer, Spolter-Weisfeld, and Thakkar (2007) provide evi-
dence that there are indeed consumers who prefer products
based on the aggregated preferences of other consumers
over products based on their own preferences.
In summary, there are arguments against the core
assumption of mass customization and individualization lit-
Testing the Value of Customization / 105
erature—namely, that products customized on the basis of
the preferences expressed yield benefits for customers.
Therefore, we formulate the following testable hypotheses:
H1: Products customized on the basis of measured preferences
generate higher benefits for customers than (a) segment-
specific products and (b) mass-marketing products.
H2: In a more general form, the closer the fit between mea-
sured preferences and product attributes, the higher is the
resulting benefit for the customer.
When Customized Products Create Benefits
We cannot assume that there is a universal pattern in cus-
tomer benefits from customization. Instead, the effect
depends on specific conditional variables. Simonson (2005,
p. 42) postulates that the “investigation [of responses to cus-
tomized offers] must consider the moderating variables …
instead of just searching for main effects.” This is also sup-
ported by Von Hippel (2001, p. 255), who, in the context of
the design interface of customization systems, states that
“the design freedom provided by user toolkits for innova-
tion may not be of interest to all or even to many users of a
given type of product or service.”
The extant literature suggests preference insight as an
important moderating variable. As we argued, the mea-
surement of preferences (which is the basis for customiza-
tion) can be effective only if consumers actually have
well-defined preferences and are sufficiently aware of
them (Kramer 2007; Simonson 2005). Kramer (2007, p.
231) argues that “the value of personalization may be lim-
ited when customers do not have stable preferences from
which to derive customized offers.” If consumers do not
really know what they want, they are more inclined to con-
struct preferences based on situational cues when asked to
specify product requirements, which brings about a high
error term in their preference measurement. A product con-
structed on the basis of this measurement is of less value to
the customer than when a customer has clear insight into his
or her preferences. Furthermore, a customer with low pref-
erence insight might also be less able to evaluate whether an
offering truly fits his or her preferences (Kramer 2007).
Prior research has suggested that preference insight varies
greatly among people (Bettman, Luce, and Payne 1998;
Chernev, Mick, and Johnson 2003), and therefore we argue
that this variable moderates the benefits obtained from
customization.
H3: A customer’s preference insight has a positive effect on
the benefit derived from customization.
A valid measurement of preferences requires not only
that preference information actually exists on the customer
side but also that the customer has the ability to reveal this
information to the manufacturer. Polanyi’s (1983, p. 136)
famous dictum, “we know more than we can tell,” points to
the problem that this is not always an easy task. A wide
body of literature has confirmed empirically that preference
information on the customer side is often sticky and that
people may have different abilities to express their prefer-
ences to an external institution (e.g., Aggarwal and
Vaidyanathan 2003; Huber, Ariely, and Fischer 2002; Mor-
rison, Roberts, and Von Hippel 2000; Von Hippel 1994; Von
Hippel and Katz 2002; West, Brown, and Hoch 1996). A
person who has difficulty expressing his or her preferences
creates a potentially high measurement error. Preference
insight and the ability to express preferences might be posi-
tively correlated because both refer to the clarity of the pref-
erence system. However, these dimensions can also differ in
many cases—for example, when a person’s verbal skills or
skills in handling communication technology are particu-
larly high or low (Aggarwal and Vaidyanathan 2003).
Therefore, we postulate the following hypothesis:
H4: A customer’s ability to express his or her preferences
has a positive effect on the benefit derived from
customization.
Even if two people have an identical level of insight into
their preferences and an identical ability to express them to
the manufacturer, we argue that the benefit they derive from
customization might still be different. People with high
product involvement might put far more effort (in terms of
time, ambition, and cognitive effort) into the product defini-
tion task than those with low involvement. Thus, the prefer-
ences they express might contain a smaller error term,
which in turn results in higher benefits from customization.
People with high product involvement might also respond
more negatively (i.e., with higher disutility) if the product
does not fit their preferences. In general, product involve-
ment refers to the relevance of a specific product (category)
as perceived by a customer based on his or her individual
needs, preferences, and interests (Zaichkowsky 1985). Con-
sumers with high product involvement tend to be portrayed
as less willing to satisfice than consumers with low product
involvement (e.g., Baker and Lutz 2000). As Von Hippel
(2005, p. 32) notes, “A serious mountain biker may be will-
ing to simply buy any couch,… even if he or she is not fully
happy with it. On the other hand, that same biker may be
totally unwilling to compromise about getting mountain
bike equipment that is precisely right for his or her specific
needs.” Accordingly, research has found that consumers
with high involvement in a product category carry out more
comparisons between different offers to ensure the best
preference fit (Gordon, McKeage, and Fox 1998). Kaplan,
Schoder, and Haenlein (2007) show that the higher the con-
sumption frequency of a certain product category (which
should correlate with the consumer’s involvement level),
the higher is the intention to adopt a customized product.
Gordon, McKeage, and Fox (1998) also demonstrate that
the relative effectiveness of customization is positively
associated with product involvement. Thus, we formulate a
testable hypothesis to describe this commonly held belief:
H5: Product involvement has a positive effect on the benefit
derived from customization.
Study 1: Simulation of Customized
Versus Standard Products in the
Newspaper Market
In Study 1, we tested our five hypotheses in the newspaper
market. We chose this product category because it enabled
us to employ concrete and relatively realistic experimental
106 / Journal of Marketing, September 2009
stimuli—namely, newspapers that were actually tailored
to the participants’ preferences (with varying proximity).
In addition, newspapers constitute a market in which
customization is currently being discussed (Schoder et
al. 2006) because of highly heterogeneous customer
preferences.
The rationale underlying this study is simple: We simu-
late the three strategies of customization, segmentation, and
mass marketing and measure the resulting benefits for cus-
tomers. As a first step, we captured the participants’ prefer-
ences with regard to a collection of 90 newspaper headlines.
In the second step, we confronted participants with the
experimental stimuli—namely, simulated “newspapers” that
were each composed of a small selection of 10 headlines
(taken from the 90 headlines) in random order. We then
measured the benefits these newspapers generated for the
participants (Figure 1).
Experimental Groups
We structured the sample into ten subgroups to enhance the
robustness of our findings. We first distinguish two inde-
pendent samples (1 and 2). The difference between them is
that the time between the first step (measurement of prefer-
ences) and the second step (measurement of reactions to
stimuli) was ten days in Sample 1, whereas the second step
immediately followed the first (using the same question-
naire) in Sample 2.
This process enabled us to analyze whether our findings
are affected by participants’ efforts to appear consistent in
their response behavior (a form of “on-stage effect; see
Podsakoff et al. 2003). Typically, participants tend to
remember their preference statements for a limited time. In
a pilot study with 25 participants, we had conducted a sur-
prise recall test in which we asked participants to rate the 90
headlines according to their preferences, and ten days later,
we confronted them with their responses, with changes in a
random selection of 20 headline ratings. On average, par-
ticipants were able to identify and correct only 24% of the
modified ratings. Moreover, participants erroneously identi-
fied changes in 20% of the ratings that had not been modi-
fied. As we expected, this suggests that ten days can indeed
provoke a substantial forgetting effect. If we find similar
support for our hypotheses in Samples 1 and 2, we can sur-
mise that the effects were not only due to an on-stage effect.
We subdivided both samples into five subsamples, each
of which received different experimental stimuli. The par-
ticipants in the two main samples were each randomly
assigned to one of the following subsamples:
•Participants in Samples 1a and 2a each received an identical
“mass-marketing newspaper” constructed on the basis of
Sample 1’s preferences (for a list of the stimuli, see the next
section).
•Participants in Samples 1b and 2b each received the one
“segment-specific newspaper” that came closest to their indi-
vidual preferences and was taken from a five-segment
solution.
•Participants in Samples 1c and 2c each received the one
“segment-specific newspaper” that came closest to their indi-
vidual preferences and was taken from a ten-segment
solution.
•Participants in Samples 1d and 2d each received a “cus-
tomized newspaper” that matched their individual prefer-
ences precisely.
•Participants in Samples 1e and 2e received all four of these
stimuli in randomized order.
In all cases, the stimulus newspapers were introduced
uniformly with the following sentences: “Here are news-
papers consisting of 10 headlines each. If you compare this
newspaper [stimulus newspaper] with the newspaper on the
right side [the anchor stimulus—i.e., the expert newspaper;
see next subsection], how do you like it?” We used
between-subjects analyses in Samples 1a–1d and 2a–2d and
within-subject analyses in Samples 1e–2e. Between-
subjects and within-subject analyses each have their spe-
cific advantages and disadvantages (Maxwell and Delaney
2004); thus, we employed both to challenge the robustness
of our findings. Overall, we employ a 2 ×2 matrix of four
independent tests (within-subject versus between-subjects
design and short-time versus long-time delta between pref-
erence measurement and measurement of reaction to
stimuli).
Experimental Stimuli
We constructed the stimuli on the basis of the preferences
expressed by participants in Sample 1. As a first step, we
asked them to rate 90 different headlines on a five-point rat-
ing scale ranging from 1 (“I would really like to read this
article”) to 5 (“I would definitely not read this article”). We
selected these anchors in accordance with the recommenda-
tions of Kramer (2007), who notes that consumers under-
stand their own preferences better if they are provided with
a transparent elicitation task. In general, the headlines in
our study included two introductory lines to give partici-
pants an idea of each article’s subject.
The decision to use 90 headlines resulted from another
pilot study we had conducted. We asked 45 randomly
selected participants to read and rate a selection of head-
lines. One-third received 60 headlines, one-third 90 head-
lines, and one-third 120 headlines. Afterward, we con-
ducted in-depth interviews to determine whether the
respective tasks overstrained them and whether the elicita-
tion task was clear and transparent. These interviews
revealed that the maximum number participants could read
completely and evaluate thoroughly was 90 headlines. In
addition, it became clear that a multi-item measurement
approach would require a much smaller number of head-
lines. As a result, we restricted the number of items to one
per headline, which should yield information of the same
validity as multiple-item measurement in this particular
case of a clear question (Bergkvist and Rossiter 2007) and
is less time consuming and strenuous for participants to
complete (Gardner et al. 1998). The 90 headlines were ran-
domly selected from the 4964 real articles released by the
Austrian Press Agency in the period between March 24 and
March 31, 2006. The articles were from the categories of
foreign affairs, domestic affairs, current events, culture,
economics, science, education, media, and sports. We used
the preference information from Sample 1 to construct the
different stimuli for Samples 1 and 2 (second step) as
follows.
Testing the Value of Customization / 107
Sample 1
•Preferences
•Reference price
a
•Product involvement
•Demographics
Segmentation of
preference data
Sample 1a
•Mass-marketing newspaper
•DV = WTP/PI/ATT
Sample 1c
•Best cluster (10 clusters, LCA)
•DV = WTP/PI/ATT
Sample 1d
•Customized newspaper
•DV = WTP/PI/ATT
Sample 1b
•Best cluster (5 clusters, LCA)
•DV = WTP/PI/ATT
Sample 1e
•Mass-marketing newspaper
•Best cluster (5 clusters, LCA)
•Best cluster (10 clusters, LCA)
•DV = WTP/PI/ATT
Assignment to best cluster via minimum
squared Euclidean distance
Time
Sample 2
•Preferences
•Reference price
a
•Product involvement
•Demographics
Sample 2a
•Mass-marketing newspaper
•DV = WTP/PI/ATT
Sample 2b
•Best cluster (5 clusters)
•DV = WTP/PI/ATT
Sample 2c
•Best cluster (10 clusters)
Sample 2d
•DV = WTP/PI/ATT
Sample 2e
•Mass-marketing newspaper
•Best cluster (5 clusters)
•Best cluster (10 clusters)
•DV = WTP/PI/ATT
Segmentation of
preference data
•Customized newspaper
•DV = WTP/PI/ATT
•Customized newspaper
•Customized newspaper
Time delta = 10 days
Time delta = 0 days
Between-
Group
Analysis 2
Within-
Group
Analysis 2
Within-
Group
Analysis 1
Between-
Group
Analysis 1
FIGURE 1
Data Collection Process (Study 1)
aWTP for a newspaper assembled by experts.
Notes: DV = dependent variables, PI = purchase intention, ATT = attitude toward the product, and LCA = latent class cluster analysis.
108 / Journal of Marketing, September 2009
Mass-marketing newspaper. We calculated the mean
preference rating for each headline in Sample 1 and ranked
the headlines on that basis. The ten best-rated headlines
constitute our mass-marketing newspaper, the one standard
product that comes closest to the participants’ preferences.
Segment-specific newspapers. Using the preference rat-
ings as a basis, we performed market segmentation using
latent class analysis (Green, Carmone, and Wachpress
1976), which offers fundamental advantages over tradi-
tional types of cluster analysis (Vermunt and Magidson
2002; Wedel and Kamakura 2000). To determine the num-
ber of segments, we used the Bayesian information criterion
to compare the models until model fit stopped improving
(Schwarz 1978). Because 90 variables for segmentation
would lead to negative degrees of freedom from 11 classes
upward, we decided to conduct latent class analysis with a
varied number of randomly chosen variables (20, 25, 30,
35, and 40). To account for the possibility of local optima,
we replicated the estimation 50 times, each with a different
set of randomly selected variables (similar to Kamakura and
Wedel 1995). We found that the ten-segment model outper-
formed the other models; therefore, we used ten segments
in our study (details on the Monte Carlo simulation are
available on request). In addition, we calculated a five-
segment solution because a meta-analysis of published mar-
ket segmentation studies found that this is the mean number
of segments used (Franke, Reisinger, and Hoppe 2009). The
ten best-rated headlines in each segment defined the respec-
tive segment-specific newspapers—that is, the product that
is best adapted to the average preferences of each segment.
Participants in the subsamples who were confronted with a
segment-specific newspaper (Samples 1b, 1c, 1e and 2b, 2c,
2e) received only the one segment-specific newspaper that
came closest to their individual preferences. This assign-
ment of participants was based on the classification results
of the latent class analysis in Sample 1, while in Sample 2,
we used the squared Euclidean distance between the partici-
pants’ (individual) preferences and each segment-specific
newspaper (calculated from Sample 1).
Customized newspapers. For each individual partici-
pant, we ranked the 90 headlines according to his or her
preferences and took the ten highest-rated headlines. When
equal scores precluded an exact solution (e.g., when 12
headlines were assigned a rating of 1), we randomly
selected the headlines from those that were tied.
Expert newspaper. To ensure an identical point of refer-
ence for our between-subjects comparisons, we defined a
standardized anchor. For this purpose, we provided all the
subgroups with an identical newspaper that was to be rated
by each participant before the stimulus newspaper (in terms
of WTP). This newspaper again consisted of ten headlines
and was developed on the basis of an expert evaluation. In a
pilot study conducted before the experiment, we asked 23
experts (journalists from different Austrian newspapers and
the Austrian Press Agency) to assess the attractiveness of
the 90 headlines to newspaper readers (which also consti-
tuted the basis for the preference ratings of the participants
in the experiment). They gave their assessment on a five-
point rating scale ranging from 1 (“A typical Austrian news-
paper reader would really like to read this article”) to 5 (“A
typical Austrian newspaper reader would not be interested
in this article at all”). We selected the ten headlines that
received the best mean ratings to constitute the “expert
newspaper.”
Measurement
In the first step, we measured the individual participants’
WTP for the expert newspaper in addition to measuring
their preferences. We also obtained data on the respondents’
product involvement and demographics (age, sex, income,
household size, and education) in this step. In the second
step, we measured the individual participants’ evaluative
reaction to each particular stimulus (Samples 1a, 1b, 1c, 1d
and 2a, 2b, 2c, 2d) and to all stimuli (Samples 1e and 2e).
For this purpose, we measured the participants’ WTP. The
WTP or reservation price denotes the maximum amount of
money a participant is willing to pay for a given product
and thus constitutes a hard proxy to measure the benefit
consumers (expect to) derive from that product (e.g., Hom-
burg, Koschate, and Hoyer 2005; Voelckner 2006). To
obtain additional indicators of consumer benefit, we also
measured the participants’ attitude toward the stimulus
product and their purchase intentions. The expert news-
paper was evaluated only in terms of the participants’WTP.
WTP. The literature proposes various methods of mea-
suring WTP (Wertenbroch and Skiera 2002). In the case of
the anchoring stimulus (the expert newspaper rated first by
any participant), we employed a bracketing technique
(Casey and Delquié 1995) to enhance the validity of the
ensuing stimulus ratings. We carried out a small pilot study
(n = 30) to check whether participants would accept the
procedure, and the results were affirmative. Thus, we first
asked participants to state their WTP for the expert news-
paper. Depending on their responses, participants were
given a higher follow-up price (their stated price plus €.10)
and asked whether they would also pay that sum. This step
was repeated with €.50 and then with €1, and ultimately the
amount was doubled until the participants gave a negative
answer. Finally, we provided participants with a scroll bar
that enabled them to adjust their final WTP in the range
between the last positive response and the ensuing negative
response. We measured WTP for the stimulus newspapers
(the mass-marketing newspaper, segment-specific newspa-
per, and customized newspaper) by providing the partici-
pants with a scroll bar ranging from €.00 to €10.00, which
constitutes a form of a contingent valuation method (Voel-
ckner 2006). The (changeable) default value was the indi-
vidual participant’s WTP for the expert newspaper (the
anchor). We carried out this elaborate procedure to obtain
measures of WTP with maximum validity and on the basis
of an identical reference point, but with a minimum of
effort on the participants’ part.
Attitude toward the product. We measured participants’
attitude toward the product using three five-point semantic
differential scales (the anchors were “like” versus “dislike,”
“good” versus “bad,” and “appealing” versus “not appeal-
ing”) adapted from the scale used by Schlosser and Shavitt
Testing the Value of Customization / 109
(2002). Reliability checks show a Cronbach’s alpha of .94
across the two samples.
Purchase intention. We measured purchase intention
using a Juster (1966) scale, an 11-point scale that measures
individual purchase probability and has been shown to pos-
sess superior validity (Uncles and Lee 2006). To verify
comprehensiveness and applicability for the purposes of
this study, we carried out a small pilot experiment (n = 25)
with affirmative results. The three benefit measures are sig-
nificantly correlated (WTP–purchase intention: r = .447, p<
.01; WTP–attitude toward the product: r = .384, p< .01;
purchase intention–attitude toward the product: r = .724,
p< .01).
Preference insight. In this first study, we measured
preference insight not by means of self-reports but on the
basis of the participants’ actual behavior. This delivered the
clear advantage of not demanding extra effort from the
participants because the questionnaire was relatively long.
The large body of literature on measurement errors sug-
gests that the less certain people feel when evaluating
objects, the greater is their tendency to answer uniformly.
Depending on the specific situation, this desire for unifor-
mity can result in a systematic tendency to confirm (yea-
saying), to deny (nay-saying), or to avoid extreme answers
(central tendency) (see Pedhazur and Schmelkin 1991).
Thus, we measured the respondents’ insight using the
squared variance of their preference statements regarding
the 90 headlines. A pilot study (n = 29) revealed that this
measure and preference insight modeled as a latent con-
struct based on the three items (used in Study 2) showed a
high correlation (r = .623, p< .01), in support of the validity
of this measurement.
Ability to express preferences. We also operationalized
this latent construct in a relatively straightforward way to
keep the questionnaire short. We built on the well-
established finding in social sciences that there is a clear
correlation between a person’s level of education and ability
to express him- or herself (see Pascarella and Terenzini
1991). Because the level of education is an ordinal measure,
we used a dichotomous scale (1 = no college degree, and
2 = college degree or higher).
Product involvement. We confronted participants with
the statement “For me, a newspaper (is) …” and asked them
to complete it using a reduced version of the personal
involvement inventory scale (see Zaichkowsky 1985),
which consists of six five-point semantic differential scales
measuring the subjective importance of newspapers,
anchored with “matters” versus “doesn’t matter,” “impor-
tant” versus “unimportant,” “useless” versus “useful,” “bor-
ing” versus “interesting,” “not needed” versus “needed,”
and “essential” versus “nonessential.” The reliability of this
construct was high, with a Cronbach’s alpha of .93 across
the two samples.
Questionnaire and Samples
We obtained the data from self-administered online ques-
tionnaires. Two random samples were drawn from Austria’s
leading online panel (provided by MindTake New Media
1Patterns in subsamples are weaker because of the lower vari-
ance in the independent variable of preference fit. However, they
are still visible.
Consulting), which is representative of Austrian residents
with an e-mail account. For Sample 1, 7500 panel partici-
pants were contacted (with one reminder) and asked to fill
out the first questionnaire (Step 1). Of these, 1213 partici-
pants answered this first questionnaire completely, for a
response rate of 16.2%. Comparisons of early and late
respondents showed no significant differences, which indi-
cates the absence of response bias (Armstrong and Overton
1977). Ten days after the first questionnaire (and after we
had carried out the segmentation analysis), the 1213 partici-
pants were contacted again and asked to fill out a second
questionnaire (Step 2) with random assignments to experi-
mental stimuli, for which the participants were divided into
Subsamples 1a, 1b, 1c, 1d, and 1e. A total of 854 respon-
dents from Step 1 agreed to participate, for a response rate
of 70.4% in Step 2. Again, no response bias was visible.
Along with Step 2 in Sample 1, another questionnaire was
sent to the second (independent) sample of 4116 partici-
pants. The questionnaire for Sample 2 merged the two ques-
tionnaires from Sample 1 into a single questionnaire. Imme-
diately after the participants stated their preferences
regarding the 90 newspaper headlines, the respective stimu-
lus was calculated. The assignment of participants to the
various subsamples (Subsamples 2a, 2b, 2c, 2d, and 2e) was
randomized by means of a database algorithm. The
response rate was 17.9% (735 complete questionnaires with
one reminder sent out). No response bias could be detected,
and tests show no significant differences between the sub-
samples in Sample 1 and Sample 2.
Findings
Difference test. As a first step, we compared the benefit
generated by customized products with segment-specific
and mass-marketing products (descriptive findings, such as
which headlines attracted the most interest and which ones
completely failed to do so, are available on request).
Table 1 shows the mean WTP, purchase intention, and
attitude toward the product for the mass-marketing newspa-
per, the two segment-specific newspapers, and the cus-
tomized newspaper for the between-subjects/within-subject
analyses in Samples 1 and 2 and for the overall sample. We
also provide the mean WTP for the expert newspaper. Tests
for differences in means show a consistent and robust pat-
tern in all four analyses.
We find that H1gains clear support in any case. Regard-
less of whether it is compared with the two segment-
specific newspapers or with the mass-marketing newspaper,
the customized newspaper received significantly more
favorable ratings with regard to WTP, purchase intention,
and attitude toward the product. (As a notable side finding,
we also discovered that the market experts were able to
empathize with customers to some degree.)
Dependency analysis. We used pooled data for the
dependency analysis.1Unlike Subsamples 1a–1d and 2a–
2d, the data in Subsamples 1e and 2e are nested (for each
110 / Journal of Marketing, September 2009
Aggregated Analysis Between-Subjects Analysis Within-Subject Analysis
Total Sample Sample 1 Sample 2 Sample 1 Sample 2
na
WTP:
M
(SD)
ATTb:
M
(SD)
PIc:
M
(SD) n
WTP:
M
(SD)
ATTb:
M
(SD)
PIc:
M
(SD) n
WTP:
M
(SD)
ATTb:
M
(SD)
PIc:
M
(SD) n
WTP:
M
(SD)
ATTb:
M
(SD)
PIc:
M
(SD) n
WTP:
M
(SD)
ATTb:
M
(SD)
PIc:
M
(SD)
Expert product
(1)
2522 .84
(.56)
— — 692 .84
(.57)
— — 586 .90
(.56)
— — 162 .82
(.58)
—- —- 149 .81
(.53)
——
Mass-marketing product
(2)
633 .88
(.53)
2.73
(.97)
3.09
(.94)
170 .92
(.55)
2.48
(.91)
2.92
(.92)
152 .92
(.51)
2.79
(1.07)
3.14
(.98)
162 .84
(.53)
2.76
(.88)
3.14
(.92)
149 .83
(.52)
2.93
(.96)
3.18
(.92)
Segment-specific
product assuming five
segments (3)
628 .91
(.53)
2.61
(.91)
2.97
(.94)
171 .92
(.55)
2.55
(.96)
2.90
(.96)
146 .96
(.52)
2.69
(.88)
2.90
(.91)
162 .89
(.53)
2.52
(.90)
2.98
(.94)
149 .87
(.52)
2.70
(.93)
3.10
(.93)
Segment-specific
product assuming ten
segments (4)
623 .91
(.56)
2.66
(.92)
2.99
(.94)
171 .92
(.55)
2.56
(.83)
2.92
(.93)
141 .96
(.55)
2.69
(.99)
2.89
(.95)
162 .89
(.55)
2.60
(.90)
3.06
(.90)
149 .87
(.52)
2.82
(.95)
3.09
(.96)
Customized product (5) 638 1.00
(.57)
2.30
(.82)
2.66
(.85)
180 1.03
(.60)
2.19
(.69)
2.49
(.74)
147 1.07
(.62)
2.39
(.83)
2.68
(.86)
162 .95
(.54)
2.26
(.88)
2.75
(.91)
149 .96
(.53)
2.38
(.90)
2.77
(.89)
Δ(5) – (3) (H1a)
p
<
.01d
p
<
.001d
p
<
.001d
p
<
.05d
p
<
.001e
p
<
.001d
p
<
.05d
p
<
.01e
p
<
.05d
p
<
.001d
p
<
.01d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
Δ(5) – (4) (H1a)
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.05d
p
<
.001e
p
<
.001d
p
<
.05d
p
<
.05e
p
<
.05d
p
<
.01d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
Δ(5) – (2) (H1b)
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.05d
p
<
.01e
p
<
.001d
p
<
.01d
p
<
.01e
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
p
<
.001d
Total
p
<
.001f
(F =
6.11)
p
<
.001f
(F =
28.19)
p
<
.001f
(F =
25.71)
n.s.f
p
<
.001f
(F =
7.61)
p
<
.001f
(F =
10.41)
n.s.f
p
< .01f
(F =
4.94)
p
<
.001f
(F =
6.27)
p
<
.001g
p
<
.001h
p
<
.001i
p
<
.001j
p
<
.001k
p
<
.001l
TABLE 1
Mean WTP, Attitude Toward the Product (ATT), and Purchase Intention (PI) (Study 1)
aTotal sample consists of 1278 participants from the between-subjects samples and 1244 participants from the within-subject samples (each participant in the within-subject sample counts for 4
data points).
b5-point rating scale (three items; 1 = “very high,” and 5 = “very low”).
c11-point probability scale transformed to a 5-point scale, in line with the work of Juster (1966) (1 = “very high,” and 5 = “very low”).
dLeast significant difference test (one-sided).
eDunnett T3 test (one-sided).
fAnalysis of variance (ANOVA).
gRepeated measures ANOVA [F(2.57, 413.43) = 12.37].
hRepeated measures ANOVA [F(2.85, 458.17) = 16.79].
iRepeated measures ANOVA [F(3, 483) = 14.96].
jRepeated measures ANOVA [F(2.72, 402.58) = 22.48].
kRepeated measures ANOVA [F(3, 444) = 22.15].
lRepeated measures ANOVA [F(3, 444) = 18.86].
Testing the Value of Customization / 111
participant, we had four observations of preference fit), and
therefore we randomly drew one observation for each par-
ticipant. The ordinary least squares regression shows that
the closer the fit between the preferences expressed and the
product obtained (i.e., the higher the “extent of customiza-
tion”), the higher is the benefits for customers, in support of
H2(see Table 2).
To test H3–H5, we used moderated regression analysis
(Cohen et al. 2003). We averaged involvement items to
form an index (similar to Homburg and Fürst 2005) and
standardized the predictor and moderator variables by mean
centering (Frazier, Tix, and Barron 2004). We then created
the product terms by multiplying the centered predictor and
moderator variables (Aiken and West 1991).
In line with H3and H4, we find that preference insight
and the ability to express preferences moderate the benefits
participants derive from customization (i.e., a closer fit
between preferences expressed and product characteristics
obtained), regardless of whether the benefits are measured
as WTP, purchase intention, or attitude toward the product
(Table 2). Perhaps because preference insight and the abil-
ity to express preferences are significantly correlated (r =
.07, p< .001), the effects diminish somewhat in the inclu-
sive model. However, they are still visible. Only product
involvement failed to show a significant moderator effect,
which refutes H5.
Discussion of Study 1
We found that customization newspapers deliver clear bene-
fits for a representative sample of customers because of
improved preference fit. We also found that the benefits
from customization are affected by the person’s insight into
his or her own preferences and ability to express them. We
failed to confirm the impact of product involvement. Thus,
the question arises whether this is a general finding that
requires an in-depth discussion of potential reasons or
whether it results from the peculiarities of the method used
or the market chosen. We offer a method-related explana-
tion: Newspapers are a media product, and they essentially
consist of information. When measuring product involve-
ment, we asked participants about the relevance of the
physical product—printed papers people peruse in the
morning. In the other part of the experiment, we heavily
emphasized the informational aspect of newspapers, opera-
tionalizing newspapers as a bundle of ten pieces of informa-
tion (the headlines). Given the many ways information can
be obtained—for example, on the Internet—it may well be
that a person is highly interested in new information but at
the same time reports a low level of involvement in (physi-
cal) newspapers. Therefore, we hypothesize that the lack of
support for H5in the newspaper market may be due to our
operationalization. Our second study is intended to shed
more light on this particular issue.
In general, the findings are based on four independent
data sets (within-subject versus between-subjects design
and short-time versus long-time delta between preference
measurement and the measurement of reaction to stimuli).
This constitutes a relatively rigorous analysis scheme, and
we obtained almost identical patterns in the four settings,
thus providing robust support for our hypotheses on the
benefits of customization.
The setting of our study required some simplifications.
Our stimuli (the simulated newspapers) consisted of only 10
headlines taken from a set of 90. This appeared necessary to
ensure that the task remained manageable for participants.
In reality, most newspapers consist of approximately 100–
300 articles, and the pool of possible news items is also
considerably larger (e.g., the Austrian Press Agency
releases approximately 600 articles, the German Press
Agency releases approximately 800, the Associated Press
releases approximately 20 million words of news per day).
In addition, our stimuli did not include other content, such
as advertisements, weather forecasts, or movie schedules,
which might also be of value to consumers (Sonnac 2000).
However, these simplifications lend our findings a con-
servative nature because there is little room for customiza-
tion to exhibit its specific strengths within the heavily
restricted setting of our study. Given heterogeneous prefer-
ences, participants can individualize only when there is a
sufficiently large solution space. We conjecture that the
benefits would be even more substantial if the experiment’s
solution space came closer to real life.
Another aspect worth discussing is that we provided
each participant in the subsamples with the optimum seg-
ment—that is, the segment that we knew came closest to his
or her preferences (assuming these preferences did not
change). Again, this is unrealistic because some assign-
ments of customers to segments would be erroneous in real
life. This can be regarded as a “subsidy” to the segmenta-
tion strategy and lends even more credibility to our finding
that customized products generate higher benefits for cus-
tomers than segment-specific products.
A research design such as ours involves the concrete
risk of on-stage effects (Podsakoff et al. 2003). Participants
were first asked about their level of interest in the 90 head-
lines; then, they were asked to evaluate simulated news-
papers comprising selections of the headlines they had pre-
viously rated. It is possible that at least some participants
were motivated by the desire to appear consistent and thus
tried to adjust their evaluations to match their previous
answers. We reduced this problem by making it impossible
to return to the headline evaluations after they had been
entered. Theoretically, participants could have kept some of
their answers in mind. Therefore, we drew two samples
with different periods between the measurement of prefer-
ences and the measurement of reactions to stimuli. If the
desire to appear consistent played a role at all, the effects in
Sample 2 should be clearer than in Sample 1. However, the
patterns in the findings are nearly identical in both samples,
suggesting that on-stage effects are unlikely to have
affected our findings.
A more critical aspect of this study is the operationali-
zation of “preference insight” and “ability to express prefer-
ences.” Although both variables can be considered latent
constructs, we employed a relatively simple measurement.
The length of the questionnaire precluded a more detailed
operationalization. This raises questions regarding the
extent to which the findings with regard to moderating
effects hold when a different measurement is used.
112 / Journal of Marketing, September 2009
TABLE 2
Moderated Multiple Regression Results (Study 1)
Model 1: Main Effects Only
Model 2: Main Effects
and Interaction with
Preference Insight
Model 3: Main Effects
and Interaction with Ability
to Express Preferences
Model 4: Main Effects
and Interaction with
Product Involvement
Model 5: Main Effects
and All Interactions
Dependent Variables Dependent Variables Dependent Variables Dependent Variables Dependent Variables
WTP ATT PI WTP ATT PI WTP ATT PI WTP ATT PI WTP ATT PI
Independent Variables
Preference fit .131*** .202*** .207*** .109*** .171*** .187*** .128*** .200*** .202*** .129*** .203*** .206*** .109*** .171*** .187***
Preference insight .011 .087** .037* .026 .108*** .051* .007 .084** .028 .012 .087** .038†.026 .108*** .051*
Ability to express
preferences –.026 .033†–.024 –.026 .034†–.023 –.026 .033†–.023 –.025 .032 –.023 –.025 .007 –.023
Product involvement .244*** .257*** .335*** .245*** .258*** .336*** .246*** .258*** .339*** .248*** .254*** .337*** .251*** .255*** .339***
Control Variables
Age –.076** –.102*** –.088*** –.074** –.099*** –.086*** –.077** –.103*** –.089*** –.077** –.101*** –.089*** –.075** –.102*** –.086***
Sex .063* –.013 .038†.064** –.011 .039†.061* –.014 .035†.064** –.014 .038†.065** –.013 .040*
Household size .049* .024 .023 .048* .024 .023 .048* .024 .023 .051* .023 .025 .050* .024 .024
Income –.011 .025 .081** –.014 .021 .078** –.008 .027 .086*** –.010 .025 .081** –.013 .025 .079**
Interactions
Preference fit ×
preference insight — — — .065* .091** .051* ——————.059* .095*** .057*
Preference fit ×ability to
express preferences ——————.049* .034†.097*** — — — .042†.007 .035†
Preference fit ×product
involvement —————————.032–.024 .017 .023 –.033 .008
R2.082*** .108*** .160*** .085*** .114*** .163*** .082*** .109*** .169*** .083*** .109*** .161*** .088*** .115 .164***
ΔR2— — — .003* .006** .003* .002* .001†.009*** .001 .001 .001 .006* .007** .004*
†
p
< .1 (one-sided).
*
p
< .05 (one-sided).
**
p
< .01 (one-sided).
***
p
< .001 (one-sided).
Note: Standardized coefficients are shown; n = 1589; ATT = attitude toward the product, and PI = purchase intention.
Testing the Value of Customization / 113
2We are indebted to one of the anonymous reviewers for point-
ing out this important issue.
Moreover, the generalizability of our findings is a weak
point in this study. Newspapers are frequently purchased,
low-cost products with a large number of potential variants.
They are hedonic (rather than utilitarian) products that are
consumed in private (Knox and Walker 2001). The question
arises whether our findings also hold for products that sys-
tematically differ from newspapers. Therefore, we con-
ducted a second study to complement the results of Study 1.
Study 2: Generalization Across
Other Markets
The rationale underlying Study 2 is different from that of
Study 1. We confronted each participant with two stimuli: a
standard product in the relevant product category and a
(simulated) customization configurator that would enable
the participant to tailor the respective product to his or her
specific preferences. Then, we measured the delta of benefit
each participant associated with the two products and his or
her resulting WTP as dependent variables. This means that
a demand artifact (Sawyer 1975) might be present in Study
2; that is, participants might in part react to their interpreta-
tion of the experiment’s purpose.2Unlike in Study 1, par-
ticipants could be distracted from the purpose of the study
only to a certain degree, and the design of the study does
not allow for a time delta, which would provoke forgetting
effects. We discuss the implications of the potential demand
artifact in the “Methodological Limitations” section.
The independent variables measured subsequently were
“preference insight,” “ability to express preferences,” and
“product involvement,” each of which we measured as a
latent construct. The study was conducted independently in
the product categories of fountain pens, kitchens, skis, and
breakfast cereals. These products differ in terms of price
level, hedonic value, and privacy of consumption. We also
repeated the analysis in the newspaper category to deter-
mine the extent to which the different methods used in
Studies 1 and 2 result in similar findings and in how far a
different measurement of product involvement alters our
findings with regard to H5.
Experimental Stimuli
Standard product. In the case of newspapers, we defined
the standard product in the same way as in Study 1; that is,
we asked experts to determine headlines that were most
appealing to newspaper readers from a new set of 90 ran-
domly selected headlines. In the other product categories,
we selected one standard market product from a set of five
popular offerings that we collected with the help of market
experts. The basis of the selection was a pilot study in
which we asked 31 participants which of the five products
they liked best. In general, the standard product was intro-
duced with the lines “Here you see a [fountain pen] which
fits the preferences and requirements of most consumers
quite well” and described along the most relevant dimen-
sions (Figure 2).
Customized product. For all the product categories, we
simulated a mass customization toolkit that enabled partici-
pants to customize the product. We designed the toolkits so
that participants could choose the most preferred parameter
value for the product dimensions that were used to describe
the standard product (Figure 2). The toolkit designs were
based on a pilot study in which we thoroughly analyzed 53
existing mass customization toolkits on the Web and inter-
viewed three market experts to make the design as realistic
as possible. Participants were informed that the technical
quality of the standard product and the customized product
was identical.
Measurement
Preference insight. The participants’ degree of aware-
ness of their own preferences regarding the specific product
category was measured on the basis of extant literature
(Kramer 2007; Simonson 2005). The items read “Regarding
[fountain pens], I know exactly what I want”; “When I pur-
chase a [fountain pen], I usually know quite soon what I
prefer”; and “When I purchase a [fountain pen], I find it
easy to choose among different alternatives” (1 = “high
agreement,” and 7 = “low agreement”; Cronbach’s α= .83).
Ability to express preferences. We measured this latent
construct on the basis of relevant literature (Kramer 2007;
Simonson 2005) and used the following items: “It would be
easy for me to describe what an ideal [fountain pen] should
look like”; “It would be no problem for me to name those
attributes of a [fountain pen] which are most important to
me”; “I could easily explain to someone else what kind of
[fountain pen] I like best”; and “If I had three minutes’ time
to explain to someone else what I like and what I dislike,
this person could theoretically choose a [fountain pen] for
me that would meet my requirements” (1 = “high agree-
ment,” and 7 = “low agreement”; Cronbach’s α= .89).
Product involvement. We measured product involve-
ment, with a new focus on involvement in news (instead of
newspapers as a physical product), using a reduced version
of Zaichkowsky’s (1985) personal involvement inventory
scale. To save space in the questionnaire, we selected three
five-point semantic differential scales. The participants
were asked to complete the short statement “For me, always
having the latest news is….” The scales were anchored at
“important” versus “unimportant,” “useless” versus “use-
ful,” and “essential” versus “nonessential.” A Cronbach’s
alpha of .87 across all samples (product categories) indi-
cates a sufficient level of reliability in this construct. In
addition, we measured participants’ involvement in news-
papers in the same way as in Study 1 (“For me, a newspaper
is …”), using the same three five-point semantic differential
scales. Cronbach’s alpha is .84, which points to high relia-
bility in the measurement. The questions regarding involve-
ment in news and newspapers were placed consecutively to
make it more clear that the objects were different (see
Oppenheim 2000). The two constructs are correlated (.52;
p< .001), which does not come as a surprise because news-
papers are indeed bundles of news.
Testing the Value of Customization / 115
ΔBenefit. We measured the additional benefit associated
with a customized product (compared with the standard
product) as a latent construct. We adapted the five items
from Schreier (2006): “Compared to the standard [fountain
pen], the customized [fountain pen] would …” (1) “better
satisfy my requirements,” (2) “better meet my personal
preferences,” (3) “more likely be the best solution for me,”
(4) “more likely be what I really want,” and (5) “more likely
fit my image of a perfect [fountain pen].” We measured
them on a seven-point scale (1 = “high agreement,” and 7 =
“low agreement”; Cronbach’s α= .96).
To ascertain the validity of the four latent constructs, we
employed exploratory factor analysis and confirmatory fac-
tor analysis. The former led to four extracted factors that
together explained 79% of the overall variance, which
points to the unidimensionality of the constructs. The latter
delivered satisfactory overall fit statistics (goodness-of-fit
index = .964, adjusted goodness-of-fit index = .949, com-
parative fit index = .982, incremental fit index = .982, and
root mean square error of approximation = .048). All factor
loadings were positive and significant, indicating a sound
degree of convergent validity (detailed results of the confir-
matory factor analysis are available from the authors on
request).
ΔWTP. The second dependent variable was the intra-
individual difference between WTP for the customized
product and WTP for the standard product. In both cases,
we measured WTP using the open-ended contingent valua-
tion approach (“How much would you be willing to pay for
[your self-designed] [fountain pen]?”) (Jones 1975). We
then calculated the delta as WTP (customized product) less
WTP (standard product).
Questionnaire and Samples
As in Study 1, we obtained the data from self-administered
online questionnaires. A random representative sample
comprising 6775 participants was drawn from Austria’s
leading online panel (provided by MindTake New Media
Consulting), and they were contacted through e-mail, with
an initial reminder after three days and a second reminder
after seven days. Each member of the sample received a
questionnaire that was adapted to one of the five product
categories and had been randomly assigned. A total of 1039
participants answered the questionnaire completely, for a
response rate of 15.3%. Again, comparisons of early and
late respondents showed no significant differences, indicat-
ing the absence of response bias (Armstrong and Overton
1977).
Findings
Difference test. We first measured the extent to which
the customized products provided higher benefits than stan-
dard products, as we conjectured in H1. We reproduced the
findings of Study 1 for the newspaper market (ΔWTP =
36%, p< .001) and found a stable pattern in the product
categories of kitchens (ΔWTP = 37%, p< .001), fountain
pens (ΔWTP = 40%, p< .001), skis (ΔWTP = 34%, p<
.001), and breakfast cereals (ΔWTP = 50%, p< .001). Thus,
H1was confirmed in any case (see Table 3). Regardless of
the product category, customization appears to increase the
benefit a customer derives from a product, though the
degree to which this increased differs between categories.
Dependency analysis. We used structural equation mod-
eling to test H2–H5(Figure 3) and tested the extent to which
the constructs of “preference insight,” “ability to express
preferences,” and “product involvement” affect the addi-
tional benefit participants derive from customization. We
set the latent construct of ΔBenefit as the first-order depen-
dent variable and ΔWTP as the second-order dependent
variable. We then estimated six structural equation models
(one for each product category and one overall model).
When first examining the results in the newspaper cate-
gory, we again confirm the results of Study 1. In addition,
when measured as latent constructs, preference insight and
the ability to express preferences affect the benefits partici-
pants derive from customization. Product involvement has a
significant impact, which, unlike Study 1, confirms H5. In
line with our conjecture, we find that if we take product
involvement as measured in Study 1 (i.e., involvement in
newspapers as a physical product), the path between prod-
uct involvement and ΔBenefit becomes insignificant (.037,
p= .329), while the other paths remain stable (preference
Newspaper Fountain Pen Kitchen Skis Cereal
M (SD) M (SD) M (SD) M (SD) M (SD)
Mass-marketing
product (1)
.75
(.52)
42.21
(27.54)
2481.80
(954.22)
211.82
(107.51)
2.11
(.84)
Customized
product (2)
1.02
(.66)
59.12
(39.57)
3406.76
(1369.04)
282.85
(147.67)
3.16
(1.08)
% benefit gain 36%,
p
< .001a
(t = –7.23)
40%,
p
< .001a
(t = –10.25)
37%,
p
< .001a
(t = –15.04)
34%,
p
< .001a
(t = –11.72)
50%,
p
< .001a
(t = –10.42)
p
< .05b
(F = 2.812)
N 231 200 198 201 209
TABLE 3
Mean WTP for Standard and Customized Products (Study 2)
at-test comparing WTP for the standard and customized product within each product category.
bANOVA comparing the relative deltas in WTP across the product categories.
Notes: WTP for each product is reported in euros.
116 / Journal of Marketing, September 2009
FIGURE 3
Structural Equation Model (Study 2)
Preference
insight
Ability to
express
preferences
Product
involvement
Delta
benefit Delta WTP
insight →ΔBenefit = .188, p< .05; ability to express pref-
erences →ΔBenefit = .217, p< .01; ΔBenefit →ΔWTP =
.361, p< .01). This suggests that the findings of Study 1 are
indeed replicated and that our improved measurement of
product involvement yields additional insights.
In the other product categories, the results of the struc-
tural equation models provide strong empirical support for
our hypotheses. We find that the constructs of preference
insight, ability to express preferences, and product involve-
ment significantly affect the benefits from customization
compared with standard products, thus confirming H3–H5.
We also find that these higher benefits result in higher WTP
(see Table 4). In all product categories, alternative models
performed worse. The model fit decreased significantly
when paths were eliminated, suggesting that the pattern
found is indeed robust across different product categories.
However, although the identified patterns are similar in all
product categories, we find substantial differences in effect
sizes. Multigroup analyses show that the product category
moderates the effects of preference insight, the ability to
express preferences, and product involvement on the benefit
a customer derives from an customized product. If the
regression weights are held constant within our model, its
overall fit decreases significantly.
Discussion of Study 2
In Study 2, we successfully replicated the findings from
Study 1 using a different method. This suggests that the pat-
terns found are robust and valid. We also discovered that the
findings are generalizable to other product categories,
which differ greatly from newspapers. Compared with
newspapers, fountain pens can be considered more hedonic,
higher-priced products that are consumed more publicly
and are not bought frequently. Skis also differ from news-
papers in terms of their price, frequency of purchase, public
consumption, and hedonic nature. Kitchens are high-priced
products that are purchased seldom and involve an exten-
sive decision-making process. Breakfast cereals may be
regarded as different from newspapers because of their
more utilitarian character. However, although we found
similar patterns in the other product categories, both analy-
sis of variance in the case of WTP increases and multigroup
analyses in the case of structural equation models show that
the coefficients differ significantly. Thus, we cannot claim
to have found a “natural constant.” In the context of our
study, it would appear premature to speculate about why,
for example, the WTP increase in the case of cereals is
approximately 50% while the respective value fluctuates
around 35% in the other product categories, because the dif-
ferences can result from the product category and its vari-
ous characteristics as well as the way we operationalized
the toolkit. However, investigating reasons for the different
levels might serve as a starting point for further research.
General Discussion
Theoretical Contribution
In our two studies, we found that products customized on
the basis of measured customer preferences deliver clear
benefits to the customer. This finding is highly relevant to
the literature because it provides evidence of a critical rela-
tionship that, until now, has only been assumed and can be
regarded as the foundation of management concepts, such
as mass customization (e.g., Pine 1993; Wind and Mahajan
2001), one-to-one-marketing (e.g., Peppers and Rogers
1993, 1997), customer relationship management (e.g.,
Lemon, White, and Winer 2002), and personalization and
smart agents (e.g., Alba et al. 1997). In our studies, we did
not use a convenience sample of students, as in Franke and
Piller (2004), Schreier (2006), or Franke and Schreier
(2008, 2009), but rather a truly representative sample drawn
from Austrian residents with an e-mail account (60% of
Austrian residents have Internet access, and there is still a
moderate bias toward younger, better-educated customers).
The finding that the high benefit of customization is not
merely a characteristic of one special-interest subgroup is a
crucial aspect of our study.
The relatively large increase in derived benefit (despite
identical technical quality) suggests that there is a great deal
of “money on the table,” which underscores the high rele-
vance of scholarly research on ways to reduce the costs of
customization and indicates that such efforts are indeed
highly promising. Cost reductions can come in various
forms and comprise further improvements in flexible pro-
duction technologies (e.g., Chua, Leong, and Lim 2003)
and lower process costs for customers through design tool-
kits that are easier to use (Randall, Terwiesch, and Ulrich
2007) or through more effective recommender systems
(Holzwarth, Janiszewski, and Neumann 2006). Given fur-
ther progress, this suggests that individual marketing will
indeed gain more importance relative to the traditional prac-
tices of segmentation and mass marketing, as several schol-
ars have predicted.
However, we also show empirically that the benefits of
customization are contingent on characteristics of the cus-
tomer—namely, his or her level of insight into own prefer-
ences, ability to express those preferences, and product
involvement (e.g., Kramer 2007; Simonson 2005). This
challenges the tendency in the popular press to advocate
customization as the best possible strategy for any con-
Testing the Value of Customization / 117
Newspaper Fountain Pen Kitchen Skis Cereal Aggregated
Dependent Variable Dependent Variable Dependent Variable Dependent Variable Dependent Variable Dependent Variable
ΔBenefit ΔWTP ΔBenefit ΔWTP ΔBenefit ΔWTP ΔBenefit ΔWTP ΔBenefit ΔWTP ΔBenefit ΔWTP
Preference insighta.175** — .331*** — .193** — .213** — .178** — .123*** —
Ability to express
preferencesb
.192** — .194** — .184** — .294*** — .258*** — .289*** —
Product involvementc.117**— .107* — .246*** — .168** — .184*** — .159*** —
ΔBenefitd— .360*** — .332*** — .254*** — .393*** — .283*** — .162***
Model Fit
χ2145,321 161,061 134,399 170,966 190,995 321,525
d.f. 989898989898
χ2/d.f. 1483 1660 1371 1708 1949 3281
GFI/AGFI .929/.901 .908/.872 .924/.894 .904/.867 .901/.863 .962/.947
IFI/CFI .973/.973 .975/.974 .979/.979 .972/.972 .957/.957 .980/.980
RMSEA .046 .057 .043 .061 .068 .047
N 231 200 198 201 209 1039
Rival Models
1. With fixed regression weights
Δd.f. —————+060.000***
Δχ2—————+144.531***
2. Without preference insight →ΔBenefit path
Δd.f. +1.00** +1.00*** +1.00*** +1.00*** +1.00*** —
Δχ2+3.51**+7.59*** +3.80*** +3.37*** +3.31*** —
3. Without ability to express preferences →ΔBenefit path
Δd.f. +1.00** +1.00*** +1.00*** +1.00*** +1.00*** —
Δχ2+4.48** +3.20*** +2.99*** +9.06*** +6.95*** —
4. Without product involvement →ΔBenefit path
Δd.f. +1.00** +1.00*** +1.00*** +1.00*** +1.00*** —
Δχ2+1.99** +2.16*** +7.93*** +2.93*** +5.76***—
*
p
< .1 (one-sided).
**
p
< .05 (one-sided).
***
p
< .01 (one-sided).
aSeven-point rating scale (three items: 1 = “agree completely,” and 7 = “disagree completely”).
bSeven-point rating scale (four items: 1 = “agree completely,” and 7 = “disagree completely”).
cSeven-point semantic differential scale following Zaichkowsky (1985).
dSeven-point rating scale (five items: 1 = “agree completely,” and 7 = “disagree completely”).
Notes: Standardized coefficients are shown. GFI = goodness-of-fit index, AGFI = adjusted goodness-of-fit index, IFI = incremental fit index, CFI = comparative fit index, and RMSEA = root mean
square error of approximation.
TABLE 4
Results from the Structural Equation Models (Study 2)
118 / Journal of Marketing, September 2009
sumer in any situation. If customers have difficulties con-
veying preference information to the company (either
because they are unaware of what they want or because they
are not able to express their preferences properly) or if they
have a low level of involvement, the benefits of customiza-
tion will be considerably lower. This is in line with findings
on user innovation activities (see Von Hippel 2005). The
consistent finding is that a small subgroup of users (termed
“lead users”) responds to the lack of products that fit their
preferences by creating such a customized product them-
selves. These lead users are characterized by a clear and
high demand for a solution to a problem they face (Von
Hippel 1988).
Our finding that customization is particularly beneficial
to customers with clear preference systems does not neces-
sarily mean that customization strategies do not make sense
in the opposite case. However, it indicates that customiza-
tion processes should be designed differently. In our stud-
ies, we focused on a customer-active means of preference
transmission in which customers must actively specify what
they want. However, there are alternatives that require less
skill and effort from the customer than laborious self-design
processes, such as smart agents or recommender systems
(which require little or no customer effort). The extent to
which these systems provide benefits in such situations
remains a question for further empirical research. In addi-
tion, it is important to bear in mind that interaction with a
customization toolkit might actually help the consumer
understand and articulate his or her preferences better
because it involves trial-and-error learning with simulated
feedback on the outcome (Von Hippel and Katz 2002).
Research from Yoon and Simonson (2008), who find that
the design of the configuration set affects the customers’
preference confidence and stability, can be viewed as a first
step toward a better understanding of how customization
might work when customers have no clear preferences.
Again, additional research is necessary.
Methodological Limitations and Avenues for
Further Research
Our studies only involved simulated products, which might
lead to a “hypothetical bias,” particularly when participants
reveal their WTP by means of the contingent valuation
method (Wertenbroch and Skiera 2002). The relatively high
sums participants claimed to be willing to pay for the prod-
ucts suggest that this effect also arose in our experiment. As
a result, conclusions regarding the absolute amounts of
money that can be skimmed through customization should
be drawn with due caution. Additional research on this issue
in a more realistic experimental setting is required. Fortu-
nately, our study focuses on the relative level and not the
absolute level of WTP. Several other studies have shown
that inflation factors are stable across experimental groups
(e.g., Franke and Piller 2004).
If we compare the WTP gains for customized news-
papers in Studies 1 (14%) and 2 (35%), we find substantial
differences, though both are based on a similar object
(newspapers consisting of 10 articles chosen from a set of
90 headlines) and the populations were similar. The differ-
ence may be attributed to the presence of a demand artifact
in Study 2 (Sawyer 1975). This means that participants
might have in part reacted to their interpretation of the
experiment’s purpose by making inflated statements of
WTP. We do not know the extent of this effect; however, we
found identical patterns in Studies 1 and 2 with regard to
newspapers, both in the main effects and in the moderator
variables, which provides evidence that the demand artifact
does not explain the entire WTP gain observed in Study 2.
Does the inflation of WTP by the demand artifact in
Study 2 affect the generalization that customization is also
beneficial in other markets? We do not believe that this is
the case. Assuming that (1) there is hardly any demand
effect in Study 1 (here, the objectives of the experiments
were unclear to participants—the study was entitled “What
are newspaper readers interested in?”—and at least Sample
1 involved a ten-day time delta) and (2) the demand artifact
in Study 2 had a similar inflating effect on all samples
(which is likely because the samples were randomized and
the setting was identical), we can “deflate” the WTP gain
from customization in the categories of fountain pens,
kitchens, and cereals by the same factor as in the newspaper
category. Regardless of the concrete deflation algorithm
(e.g., multiplicative or subtractive), the findings should still
indicate a positive “true” gain in the other product
categories.
We focused on the benefit component arising from
increased preference fit. Although preference fit is consid-
ered the most important positive component in the cus-
tomers’ “utility balance,” the process of self-designing a
product might also carry value for customers (Franke and
Schreier 2009). There may also be (positive) interaction
effects between these components, which would warrant
studies measuring both effects simultaneously. Further
research also should integrate cost components (with varied
methods of obtaining customized products, such as self-
design toolkits or recommender systems) because they
might also interact with the other components and yield
more realistic insights into the net value of customization.
All these components and their interplay might vary with
product categories and cultural factors, which would
explain why customization is common in some industries
and nonexistent in others or why the commercial use of
configuration toolkits differs in various countries (e.g., car
configurators are more popular in Europe than in the United
States).
In addition, the values attributed by consumers are not
the only variables that determine whether customization is a
viable business strategy in markets characterized by hetero-
geneous customer preferences. Other factors include net-
work externalities of the products (which instead under-
score the importance of standardization), experience effects,
economies of scale, production and distribution costs, orga-
nizational issues (e.g., the “not-invented-here” syndrome),
and the costs of organizational coordination. Further
research from different perspectives is necessary to enhance
the understanding of when customization constitutes a
promising marketing strategy.
Testing the Value of Customization / 119
REFERENCES
Aggarwal Praveen and Rajiv Vaidyanathan (2003), “Eliciting
Online Customers’ Preferences: Conjoint vs. Self-Explicated
Attribute-Level Measurements,” Journal of Marketing Man-
agement, 19 (February), 157–77.
Aiken, Leona S. and Stephen G. West (1991), Multiple Regres-
sion: Testing and Interpreting Interactions. Newbury Park, CA:
Sage Publications.
Alba, Joseph, John Lynch, Barton Weitz, Chris Janiszewski,
Richard Lutz, Alan Sawyer, and Stacy Wood (1997), “Inter-
active Home Shopping: Consumer, Retailer, and Manufacturer
Incentives to Participate in Electronic Marketplaces,” Journal
of Marketing, 61 (July), 38–53.
Ansari, Asim and Carl F. Mela (2003), “E-Customization,” Jour-
nal of Marketing Research, 40 (May), 131–45.
Armstrong, J. Scott and Terry S. Overton (1977), “Estimating
Nonresponse Bias in Mail Surveys,” Journal of Marketing
Research, 14 (August), 396–402.
Baker, William E. and Richard J. Lutz (2000), “An Empirical Test
of an Updated Relevance-Accessibility Model of Advertising
Effectiveness,” Journal of Advertising, 29 (Spring), 1–14.
Bardakci, Ahmet and Jeryl Whitelock (2004), “How ‘Ready’
Are Customers for Mass Customization? An Exploratory
Investigation,” European Journal of Marketing, 38 (Novem-
ber), 1396–1416.
Bergkvist, Lars and John R. Rossiter (2007), “The Predictive
Validity of Multiple-Item Versus Single-Item Measures of the
Same Constructs,” Journal of Marketing Research, 44 (May),
175–84.
Bettman, James R., Mary Frances Luce, and John W. Payne
(1998), “Constructive Consumer Choice Processes,” Journal of
Consumer Research, 25 (December), 187–217.
Casey, Jeff T. and Philippe Delquié (1995), “Stated vs. Implicit
Willingness to Pay Under Risk,” Organizational Behavior and
Human Decision Processes, 61 (February), 123–37.
Chernev, Alexander, David Glen Mick, and Michael D. Johnson
(2003), “When More Is Less and Less Is More: The Role of
Ideal Point Availability and Assortment in Consumer Choice,”
Journal of Consumer Research, 30 (September), 170–83.
Chua, Chee K., Kah F. Leong, and Chu S. Lim (2003), Rapid Pro-
totyping: Principles and Applications. Singapore: World
Scientific.
Cohen, Jacob, Patricia Cohen, Stephen G. West, and Leona S.
Aiken (2003), Applied Multiple Regression/Correlation Analy-
sis for the Behavioral Sciences, 3d ed. Mahwah, NJ: Lawrence
Erlbaum Associates.
Dellaert, Benedict G.C. and Stefan Stremersch (2005), “Marketing
Mass-Customized Products: Striking a Balance Between Util-
ity and Complexity,” Journal of Marketing Research, 42
(May), 219–27.
Diehl, Kristin, Laura J. Kornish, and John G. Lynch (2003),
“Smart Agents: When Lower Search Costs for Quality Infor-
mation Increase Price Sensitivity,” Journal of Consumer
Research, 30 (June), 56–71.
Duray, Rebecca, Peter T. Ward, Glenn W. Milligan, and William L.
Berry (2000), “Approaches to Mass Customization: Configura-
tions and Empirical Validation,” Journal of Operations Man-
agement, 18 (November), 605–625.
Fang, Eric (2008), “Customer Participation and the Trade-Off
Between New Product Innovativeness and Speed to Market,”
Journal of Marketing, 72 (July), 90–104.
Fischhoff, Baruch (1991), “Value Elicitation: Is There Anything
There?” American Psychologist, 46 (August), 835–47.
Franke, Nikolaus and Frank T. Piller (2004), “Value Creation by
Toolkits for User Innovation and Design: The Case of the
Watch Market,” Journal of Product Innovation Management,
21 (November), 401–415.
———, Heribert Reisinger, and Daniel Hoppe (2009), “Remain-
ing Within-Cluster Heterogeneity: A Meta-Analysis of the
‘Dark Side’ of Clustering Methods,” Journal of Marketing
Management, 25 (3–4), 273–93.
——— and Martin Schreier (2008), “Product Uniqueness as a
Driver of Customer Utility in Mass Customization,” Marketing
Letters, 19 (2), 93–107.
——— and ——— (2009), “Why Customers Value Mass-
Customized Products: The Importance of Process Effort and
Enjoyment,” Journal of Product Innovation Management,
forthcoming.
Frazier, Patricia A., Andrew P. Tix, and Kenneth E. Barron (2004),
“Testing Moderator and Mediator Effects in Counseling Psy-
chology Research,” Journal of Counseling Psychology, 51
(January), 115–34.
Gardner, Donald G., Larry L. Cummings, Randall B. Dunham,
and Jon L. Pierce (1998), “Single-Item Versus Multiple-Item
Measurement Scales: An Empirical Comparison,” Educational
and Psychological Measurement, 58 (December), 898–915.
Gilmore, James H. and B. Joseph Pine (1997), “The Four Faces
of Mass Customization,” Harvard Business Review, 75
(January–February), 91–101.
——— and ——— (2000), Markets of One: Creating Customer-
Unique Value Through Mass Customization. Boston: Harvard
Business School Press.
Gordon, Mary E., Kim McKeage, and Mark A. Fox (1998), “Rela-
tionship Marketing Effectiveness: The Role of Involvement,”
Psychology and Marketing, 15 (August), 443–59.
Green, Paul E., Frank J. Carmone, and David P. Wachspress
(1976), “Consumer Segmentation via Latent Class Analysis,”
Journal of Consumer Research, 3 (December), 170–74.
Hill, Kimberly (2003), “Customers Love/Hate Customization,”
CRM-Daily.com, (April 10), (accessed December 4, 2007),
[available at http://www.crm-daily.com].
Holzwarth, Martin, Chris Janiszewski, and Markus M. Neumann
(2006), “The Influence of Avatars on Online Consumer Shop-
ping Behavior,” Journal of Marketing, 70 (October), 19–36.
Homburg, Christian and Andreas Fürst (2005), “How Organiza-
tional Complaint Handling Drives Customer Loyalty: An
Analysis of the Mechanistic and the Organic Approach,” Jour-
nal of Marketing, 69 (July), 95–114.
———, Nicole Koschate, and Wayne D. Hoyer (2005), “Do Satis-
fied Customers Really Pay More? A Study of the Relationship
Between Customer Satisfaction and Willingness to Pay,” Jour-
nal of Marketing, 69 (July), 84–96.
Huber, Joel, Dan Ariely, and Gregory Fischer (2002), “Expressing
Preferences in a Principal-Agent Task: A Comparison of
Choice, Rating, and Matching,” Organizational Behavior and
Human Decision Processes, 87 (January), 66–90.
Huffman, Cynthia and Barbara E. Kahn (1998), “Variety for Sale:
Mass Customization or Mass Confusion?” Journal of Retail-
ing, 74 (Winter), 491–513.
Jones, D. Frank (1975), “A Survey Technique to Measure Demand
Under Various Pricing Strategies,” Journal of Marketing, 39
(July), 75–77.
Juster, F. Thomas (1966), “Consumer Buying Intentions and Pur-
chase Probability: An Experiment in Survey Design,” Journal
of the American Statistical Association, 61 (September),
658–96.
Kahn, Barbara E. (1998), “Dynamic Relationships with Cus-
tomers: High-Variety Strategies,” Journal of the Academy of
Marketing Science, 26 (Winter), 45–53.
Kamakura, Wagner A. and Michel Wedel (1995), “Life-Style Seg-
mentation with Tailored Interviewing,” Journal of Marketing
Research, 32 (August), 308–317.
120 / Journal of Marketing, September 2009
Kaplan, Andreas M., Detlef Schoder, and Michael Haenlein
(2007), “Factors Influencing the Adoption of Mass Customiza-
tion: The Impact of Base Category Consumption Frequency
and Need Satisfaction,” Journal of Product Innovation Man-
agement, 24 (March), 101–116.
Knox, Simon and David Walker (2001), “Measuring and Manag-
ing Brand Loyalty,” Journal of Strategic Marketing, 9 (2),
111–28.
Kotha, Suresh (1995), “Mass Customization: Implementing the
Emerging Paradigm for Competitive Advantage,” Strategic
Management Journal, 16 (Summer), 21–42.
Kramer, Thomas (2007), “The Effect of Measurement Task Trans-
parency on Preference Construction and Evaluations of Person-
alized Recommendations,” Journal of Marketing Research, 44
(May), 224–33.
———, Suri Spolter-Weisfeld, and Meneesh Thakkar (2007),
“The Effect of Cultural Orientation on Consumer Responses
to Personalization,” Marketing Science, 26 (March–April),
246–58.
Lemon, Katherine N., Tiffany Barnett White, and Russell S. Winer
(2002), “Dynamic Customer Relationship Management: Incor-
porating Future Considerations into the Service Retention
Decision,” Journal of Marketing, 66 (January), 1–14.
Maxwell, Scott E. and Harold D. Delaney (2004), Designing
Experiments and Analyzing Data: A Model Comparison Per-
spective, 2d ed. Mahwah, NJ: Lawrence Erlbaum Associates.
Morrison, Donald G. and Jorge Silva-Risso (1995), “A Latent
Look at Empirical Generalizations,” Marketing Science, 14
(Part 2 of 2), 61–70.
Morrison, Pamela D., John H. Roberts, and Eric von Hippel
(2000), “Determinants of User Innovation and Innovation Shar-
ing in a Local Market,” Management Science, 46 (December),
1513–27.
Oppenheim, Abraham N. (2000), Questionnaire Design, Inter-
viewing and Attitude Measurement. London: Continuum.
Pascarella, Ernest T. and Patrick T. Terenzini (1991), How College
Affects Students: Findings and Insights from Twenty Years of
Research. San Francisco: Jossey-Bass.
Pedhazur, Elazar J. and Liora Pedhazur Schmelkin (1991), Mea-
surement, Design, and Analysis: An Integrated Approach.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Peppers, Don and Martha Rogers (1993), The One to One Future.
New York: Currency-Doubleday.
——— and ——— (1997), Enterprise One to One. New York:
Currency-Doubleday.
Pine, B. Joseph (1993), Mass Customization: The New Frontier in
Business Competition. Boston: Harvard Business School Press.
Podsakoff, Philip M., Scott B. MacKenzie, Jeong-Yeon Lee, and
Nathan P. Podsakoff (2003), “Common Method Biases in
Behavioral Research: A Critical Review of the Literature and
Recommended Remedies,” Journal of Applied Psychology, 88
(October), 879–903.
Polanyi, Michael (1983), The Tacit Dimension. Gloucester, MA:
Peter Smith Publisher.
Randall, Taylor, Christian Terwiesch, and Karl T. Ulrich (2005),
“Principles for User Design of Customized Products,” Califor-
nia Management Review, 47 (Summer), 68–85.
———, ———, and ——— (2007), “User Design of Customized
Products,” Marketing Science, 26 (March–April), 268–80.
Sawyer, Alan G. (1975), “Demand Artifacts in Laboratory Experi-
ments in Consumer Research,” Journal of Consumer Research,
1 (March), 20–30.
Schlosser, Ann E. and Sharon Shavitt (2002), “Anticipating Dis-
cussion About a Product: Rehearsing What to Say Can Affect
Your Judgments,” Journal of Consumer Research, 29 (June),
101–115.
Schoder, Detlef, Stefan Sick, Johannes Putzke, and Andreas M.
Kaplan (2006), “Mass Customization in the Newspaper Indus-
try: Consumers’ Attitudes Toward Individualized Media Inno-
vations,” International Journal on Media Management, 8 (1),
9–18.
Schreier, Martin (2006), “The Value Increment of Mass-
Customized Products: An Empirical Assessment,” Journal of
Consumer Behaviour, 5 (July–August), 317–27.
Schwarz, Gideon (1978), “Estimating the Dimension of a Model,”
The Annals of Statistics, 6 (March), 461–64.
Sheth, Jagdish N. and Rajendra S. Sisodia (1999), “Revisiting
Marketing’s Lawlike Generalizations,” Journal of the Academy
of Marketing Science, 27 (Winter), 71–87.
———, ———, and Arun Sharma (2000), “The Antecedents and
Consequences of Customer-Centric Marketing,” Journal of the
Academy of Marketing Science, 28 (January), 55–66.
Simonson, Itamar (2005), “Determinants of Customers’ Responses
to Customized Offers: Conceptual Framework and Research
Propositions,” Journal of Marketing, 69 (January), 32–45.
Slovic, Paul (1995), “The Construction of Preference,” American
Psychologist, 50 (May), 364–71.
Smith, Wendell R. (1956), “Product Differentiation and Market
Segmentation as Alternative Marketing Strategies,” Journal of
Marketing, 21 (July), 3–8.
Sonnac, Nathalie (2000), “Readers’ Attitudes Toward Press Adver-
tising: Are They Ad-Lovers or Ad-Averse?” Journal of Media
Economics, 13 (4), 249–59.
Surowiecki, James (2004), Why the Many Are Smarter Than the
Few and How Collective Wisdom Shapes Business, Economies,
Societies and Nations. New York: Doubleday.
Syam, Niladri B., Ranran Ruan, and James D. Hess (2005), “Cus-
tomized Products: A Competitive Analysis,” Marketing Sci-
ence, 24 (Fall), 569–84.
Uncles, Mark and David Lee (2006), “Brand Purchasing by Older
Consumers: An Investigation Using the Juster Scale and the
Dirichlet Model,” Marketing Letters, 17 (1), 17–29.
Varki, Sajeev and Roland Rust (1998), “Technology and Optimal
Segment Size,” Marketing Letters, 9 (2), 147–67.
Vermunt, Jeroen K. and Jay Magidson (2002), “Latent Class Clus-
ter Analysis,” in Applied Latent Class Analysis, Jacques A.
Hagenaars and Allan L. McCutcheon, eds. Cambridge, UK:
Cambridge University Press, 89–106.
Voelckner, Franziska (2006), “An Empirical Comparison of Meth-
ods for Measuring Consumers’ Willingness to Pay,” Marketing
Letters, 17 (April), 137–49.
Von Hippel, Eric (1988), “Lead Users: A Source of Novel Product
Concepts,” Management Science, 32 (July), 791–805.
——— (1994), “Sticky Information and the Locus of Problem
Solving: Implications for Innovation,” Management Science,
40 (April), 429–39.
——— (2001), “User Toolkits for Innovation,” Journal of Product
Innovation Management, 18 (July), 247–57.
——— (2005), Democratizing Innovation. Cambridge, MA: MIT
Press.
——— and Ralph Katz (2002), “Shifting Innovation to Users via
Toolkits,” Management Science, 48 (July), 821–33.
Wedel, Michel and Wagner A. Kamakura (2000), Market Segmen-
tation: Conceptual and Methodological Foundations, 2d ed.
Boston: Kluwer Academic Publishers.
Wertenbroch, Klaus and Bernd Skiera (2002), “Measuring Con-
sumers’ Willingness to Pay at the Point of Purchase,” Journal
of Marketing Research, 39 (May), 228–41.
West, Patricia M., Christina L. Brown, and Stephen J. Hoch
(1996), “Consumption Vocabulary and Preference Formation,”
Journal of Consumer Research, 23 (September), 120–35.
Wind, Jerry and Vijay Mahajan (2001), Digital Marketing: Global
Strategies from the World’s Leading Experts. New York: John
Wiley & Sons.
Yoon, Song-Oh and Itamar Simonson (2008), “Choice Set Config-
uration as a Determinant of Preference Attribution and
Strength,” Journal of Consumer Research, 35 (August),
324–36.
Testing the Value of Customization / 121
Zaichkowsky, Judith L. (1985), “Measuring the Involvement Con-
struct,” Journal of Consumer Research, 12 (December),
341–52.
Zipkin, Paul (2001), “The Limits of Mass Customization,” MIT
Sloan Management Review, 42 (Spring), 81–87.
