You are on page 1of 19

The current issue and full text archive of this journal is available at

www.emeraldinsight.com/0007-070X.htm

Brand familiarity
Brand familiarity and tasting in
conjoint analysis
An experimental study with Croatian beer
consumers 561
Marija Cerjak
Department of Agricultural Marketing,
Faculty of Agriculture, University of Zagreb, Zagreb, Croatia
Rainer Haas
Institute for Marketing and Innovation,
University of Natural Resources and Applied Life Sciences, Vienna, Austria, and
Damir Kovačić
Department of Agricultural Marketing,
Faculty of Agriculture, University of Zagreb, Zagreb, Croatia

Abstract
Purpose – The aims of this paper is to determine, via an empirical study of beer consumers in
Croatia, the influence of tasting on the validity of conjoint analysis (CA) under presence of familiar or
unfamiliar brands.
Design/methodology/approach – The research comprised a face-to-face survey with 403 beer
consumers. The respondents were divided into four groups regarding CA experiment
(familiar/unfamiliar beer brand in combination with presence or absence of beer tasting). CA
validity was measured with five criteria: face validity, convergent validity, internal validity, predictive
validity and subjective evaluation of conjoint task. In addition to the CA experiment, a structured
questionnaire was used consisting of a few questions regarding respondents’ socio-economic
characteristics, beer purchasing, and consuming behaviour.
Findings – The research results confirmed that tasting as an additional presentation method has
significant influence on validity of CA. However, the results of the study indicate that tasting should
be used as a stimulus presentation method for CA with food and beverage products/brands, which are
unfamiliar to the consumers. When testing familiar brands and brands with established perceptions,
simpler and less expensive verbal stimulus presentation can be used.
Practical implications – According to the research results, it could be concluded that when
performing CA with strong familiar brands, it is not necessary to use CA with tasting since tasting
increases research complexity and costs and it does not achieve better results. However, tasting as a
stimuli presentation method gives better results than pure verbal CA in the case of unfamiliar brands.
Originality/value – The paper is one of the first to deal with tasting as a presentation method in
conjoint analysis and its results have direct implications for the future use of CA with food and
beverages.
Keywords Brand awareness, Food and drink products, Beer, Croatia, Sensory perception
Paper type Research paper

1. Introduction British Food Journal


Vol. 112 No. 6, 2010
Conjoint analysis (CA) is a common marketing research method for analysing pp. 561-579
consumer trade-offs (Carroll and Green, 1995; Krapp and Sattler, 2001; Green et al., q Emerald Group Publishing Limited
0007-070X
2001; Brusch and Baier, 2002; Toubia, 2001). It is used so frequently because it DOI 10.1108/00070701011052664
BFJ produces fairly realistic imitations of real market choice (Huber, 1987; Brusch and
112,6 Baier, 2002) and provides good estimations of consumers’ preferences. Furthermore,
“the popularity of conjoint measurement appears to derive, at least in part, from its
presumed superiority in (predictive) validity over less expensive techniques such as
self-explication approaches” (Leight et al., 1984; taken from Krapp and Sattler, 2001).
Because of its practical importance, numerous researchers have shown an interest in
562 enhancing the degree of task realism in the evaluation process during CA (Green and
Helsen, 1989). Research efforts in the past have focused on three main aspects:
(1) Realistic definition of attributes and attribute levels.
(2) Improvement in the evaluation procedure so that it resembles actual purchase
decisions.
(3) More realistic stimulus presentation methods.

In view of the specific characteristics of consumer food and beverage choice, the need
for “holistic” CA methods, meaning more realistic stimulus presentations, is obvious.
To gain a fuller understanding of the consumer choice, market research methods need
to encompass quality attributes of pre-purchase phase (brand, quality label, etc.) and
post-purchase phase (taste, convenience, etc.) in one test design (Grunert, 2002).
Using beer as an example we investigate how brand familiarity combined with
tasting as additional presentation method influences the outcome of CA.

2. Research objectives and questions


CA for food and beverage products is frequently confronted with the problem how to
include taste in the experiment, even though taste is an essential intrinsic quality
dimension. There are three possibilities to deal with this problem. First, researchers
use verbal description of taste in terms such as “quite sweet with a strawberry note”,
without knowing for example what degree of sweetness the test persons relate to the
term “quite sweet”. Second, researchers apply CA without tasting and try to obtain
information about taste from qualitative methods such as focus groups. Third, CA
with tasting as additional product presentation method is performed without
knowing if and how taste influences perception of the other attributes contributing to
a change in their utility functions (Vickers, 1993). Until now, little or no research has
taken place to clarify the question of how taste, as one of the most important food
product attributes, interacts with strong or weak brands (“weak” in the sense of
brands that are unfamiliar to the consumer) during a CA experiment. Could it be that
the familiarity with a brand is so strong that consumers have already established
permanent “taste images”, making it unnecessary to include taste in CA test designs
containing “strong” brands?
Consumers’ choice of food and beverage products is typically considered a low
involvement choice, which means that they apply heuristic models of choice with less
cognitive efforts and more routine- or emotional-based decisions. For low involvement
products consumers often use familiar brand names as information cues, standing for a
“bundle” of product attributes including taste, simplifying their selection process
(Wansink, 2003). It has to be expected that consumers apply similar heuristic during
conjoint task. Therefore it is of crucial importance for product development to
understand the mutual influence of taste and brand.
Furthermore consumers’ choice of food and beverage products is mainly influenced Brand familiarity
by quality expectations, which are derived from extrinsic quality cues (i.e. brand, price,
store in which the product is bought, etc.) and intrinsic quality cues (colour of an apple
as a quality cue for taste or fat content as a quality cue for tenderness and taste of meat
(Grunert, 2002, p. 275)). After purchase, when consumers experience the quality of the
product, hedonic factors such as smell, appearance and taste strongly influence
customer satisfaction and re-purchase intentions (Grunert et al., 2000). To this extent, 563
offering test persons the possibility of tasting sample products during conjoint
measurement establishes a more “holistic” test situation that includes attributes from
the pre-purchase and the consumption situation. Does such a setting lead to higher
validity of the CA outcomes?
Based on this framework we looked for a typical food and beverage product for
which the consumption situation could be easily included into the CA experiment,
meaning that extensive cooking processes should not be required. We selected beer as
a test product because it is widely used and beer consumption does not depend on the
consumer’s age, education, income or social status. Standard beer quality, e.g. its
sensory characteristics, enables consumers to have stable preference structures, which
help them to recognise the product according to taste.
Clearly the selection of beer as a test product limits the degree to which the findings
can be generalised. The taste of different beers is not always as distinct as it is the case
with some other foods and beverages. Using beer reduces generalisability to packed,
ready to consume food and beverage products. But in the category “alcoholic
beverages” beer has the highest share of value in sales worldwide (ACNielsen, 2006, p.
42). Furthermore it is important to keep in mind that in the beer category manufacturer
brands play a more important role in the consumer choice than private labels. It is the
only food and beverage category worldwide in which manufacturer brands grew faster
then private labels from 2005 to 2006. The global share of private labels in the beer
market is around 6 per cent, which is quite low compared for example to the dairy
market, which has a 27 per cent global share of private labels (ACNielsen, 2006, p. 42).
Insofar the beer market represents a food and beverage sector where manufacturer
brands are still strong and consumer loyalty to these brands plays an important role.
Choosing beer as an example, the objective of this article is to clarify the influence of
tasting in CA on consumer preferences in the presence of familiar or unfamiliar brands.
In respect to this objective the two main research questions in this study are: “How
does tasting as an additional presentation method influence the validity of CA with
food and beverage products?” and “How does tasting as presentation method in CA
influence the importance of the product attribute ‘brand’?”

3. Theoretical background
The theoretical background to this study focuses on three topics: first, the importance
of taste and its relation to the attribute “brand” in respect to the food and beverage
choice, second the need to present stimuli in CA as realistically as possible, third, an
explanation of the validity measures applied in this study.

3.1 The influence of taste and brand on the food and beverage choice
Roeber et al. (2002) confirm the result of the Food Marketing Institute study (1999) by
determining that “a majority of consumers consider taste, nutrition, product safety,
BFJ and price as ‘very important’ factors in food selection”. In a US study, Govindasamy
112,6 et al. (1997) showed that consumers rank freshness, taste/flavour, cleanliness, health
value and absence of pesticides among the most important food characteristics.
According to Schutz et al. (1986), food sensory characteristics, nutritional information,
brand name and price are the most important characteristics in food choice. Roininen
(2001) identifies health, energy content, good taste, momentary desire, and price as the
564 most frequently mentioned reasons for choosing either apples or chocolate bars.
Numerous studies have shown that brand name and price are the most important
attributes for rapidly consumed goods (Hensel-Börner and Sattler, 2000a). Thus, it is to
be expected that when choosing beer, brand and price will be more important than
other attributes used in the CA. Furthermore, Daems and Delvaux (1997) mention that
the sensory characteristics of beer are the most important for consumers. Food and
beverage sensory characteristics i.e. their sensory image, are very often associated with
a product name (Bárcenas et al., 2001). For that reason, it could be expected that
respondents in our study would express the importance of beer taste through the
importance of beer brands.
It is important to keep in mind that we use the term “brand” as a multidimensional
concept that goes far beyond the term “trademark”. A trademark can be a word, name,
symbol, colour, scent or sound and represents a legal construct designed to protect
products or services of entrepreneurs from imitators. In the strictest sense a trademark
from the consumer’s point-of-view could be little more than a name without further
meaning:
If a company treats a brand only as a name, it misses the point of branding. Branding is used
to develop a deep set of meanings for the brand (Kotler and Keller, 2005, p. 443).
This “deep set of meaning” can encompass different levels such as attributes, benefits,
values, cultures or personality (Kotler and Keller, 2005, p. 443). In this article the term
“unfamiliar brand” stands for brands that in the mind of the consumer represent
nothing more then a name or symbol without further meaning. The term “familiar
brand” stands for any brand having a “deep set of meaning”, which goes beyond the
simple connotation of a name.

3.2 Influence of different stimuli presentations on CA results


“Endeavoring to make stimuli as realistic as possible is traditionally one of the core
elements of conjoint analysis in practice” (Strebinger et al., 2000a), but also the subject
of numerous methodological studies aiming to improve this method. Vriens et al. (1998)
compared pictorial and verbal stimulus presentation methods. Sattler (1994) compared
verbal and real presentation. Jaeger et al. (2001) conducted research comparing pictorial
and real product presentation, and Ernst and Sattler (2000) and Brusch et al. (2002)
studied multimedia stimulus presentation in comparison to verbal presentation. These
studies show that presentation form does not influence CA. There are possible effects
on internal validity (comparing verbal and real stimulus presentation for the benefit of
verbal presentation) and an influence on the direct or derived results (Brusch et al.,
2002).
Presentation forms that are more complex than pictorial or real presentation
increase the understanding of product attributes while verbal stimulus presentation
facilitates stimulus evaluation and reduces research costs.
Although some studies reveal certain advantages of verbal over pictorial and real Brand familiarity
stimulus presentation (Ernst and Sattler, 2000), various authors call for the use of real
product models when possible (Holbrook and Moore, 1981; Green and Srinivasan,
1990). Strebinger et al. (2000a, b) argue that pictorial and real stimuli as opposed to
purely verbal product descriptions can convey more decision-relevant information and
that this effect differs from product to product. A new technical product concept can be
relatively easily described by means of its technical characteristics. However, the 565
problem emerges with products such as food and beverages, which have intrinsic
quality attributes such as taste or sensory attributes. Jaeger et al. (2001) note:
While a verbal representation may be satisfactory for price information, intrinsic product
characteristics of food products including appearance, taste and texture are probably less
adequately represented.
Holbrook and Moore (1981) state that:
Many products involving aesthetic, sensory, or symbolic benefits must be experienced to be
judged adequately. Music, haute cuisine, or fashion designs must be heard, tasted or seen to
be properly appreciated.
Stimulus presentation as real food or drink products together with the possibility of
tasting them has been used only in a very few previous studies. The sensory
characteristics of the products were simply omitted in CA design and only consumer
preferences towards other product attributes were examined. Exclusion of the intrinsic
attributes such as appearance, flavour or texture from CA or ignorance of them could
lead to distortion of results because these attributes cannot be adequately described
verbally (Jaeger et al., 2001; Brusch et al., 2002). Therefore some researchers call for
tasting as a presentation method in CA with food and beverages (Helgesen et al., 1998;
Vickers, 1993; Cheng et al., 1990).

3.3 Definition of applied validity criteria


In this study we used five criteria to measure the validity of CA: face validity,
convergent validity, internal validity, predictive validity and subjective evaluations of
the test persons concerning the conjoint tasks.
Face validity measures whether the results from the CA (partworths) correspond
with the expectations of experts, for example whether the results make sense in view of
the available theoretical and empirical knowledge about the market place. Numerous
studies have shown that brand name and price are the most important attributes for
short durable consumer goods (Sattler, 1994; Tscheulin, 1991; Hensel-Börner and
Sattler, 2000a). Food and beverage sensory characteristics, e.g. their sensory image, are
very often associated with a product name (Bárcenas et al., 2001). For that reason
respondents in our study could express the importance of beer taste through the
importance of beer brand.
Convergent validity tests whether the utility weights (i.e. parthworths) obtained
from different CA methods are the same. If these weights are equal (statistically
indifferent), equal convergent validity can be expected because all further
computations are based on the estimated partworths. Differences in the preference
structure can be identified by a comparison of the distributions of parthworths
Internal validity measures the degree of appropriateness of the measurement model.
It is measured as a correlation between the input values (preference ranks) and the
BFJ estimated (output) values of the dependent variable. For each card/stimuli the
112,6 partworths are summed and according to that sum, the estimated preference ranks are
calculated and compared with the empirical rank value. Further, for each respondent it
is possible to calculate the rank correlation coefficient (Kendal’s Tau) between
empirical and estimated rank values (Sattler, 1994; Backhaus et al., 1996; Ernst and
Sattler, 2000).
566 Predictive validity measures the extent to which estimated results overlap with real
purchase intentions. The first-choice-hit rate employed in this research is one of the
criteria very often used to measure predictive validity (Vriens et al., 1998; Brusch et al.,
2002; Tscheulin, 1991). The individual partworths for each respondent group are used
to predict choice behaviour in the holdout sample. Holdouts are product stimuli ranked
by the test persons without being used for the estimation of the product attribute
partworths during CA. The percentage of choices correctly predicted in the holdout
sample based on the first-choice rule indicates the predictive validity of this study.
Hensel-Börner and Sattler (2000a) mention that the complexity of the CA experiment
could negatively influence the validity of the results. Therefore we additionally
measured the degree of complexity of the different CA experiment settings (with and
without tasting) by asking respondents for their subjective evaluation (Hensel-Börner
and Sattler, 2000a; Huber et al., 1991; Krapp and Sattler, 2001).

4. Research hypotheses
As several studies have shown, more complex and realistic presentations increase the
understanding of product evaluation, while verbal stimulus presentation facilitates
stimulus evaluation and reduces research costs. Nevertheless several authors have
concluded that in the case of food and beverage products it is not sufficient to describe
the intrinsic quality attribute “taste” only verbally (Jaeger et al., 2001; Holbrook and
Moore, 1981; Helgesen et al., 1998; Vickers, 1993; Cheng et al., 1990). Because of the
difficulty of describing sensory attributes as flavour or texture verbally, we assume
that if test persons taste food and beverage products during CA, the results should be
more valid. We therefore formulate the hypothesis:
H1. CA with tasting has higher validity than CA with only verbal presentation of
stimuli.
The analysis of literature shows that for food and beverage products and especially
beer, the most frequent quoted important product attributes are taste (sensory
characteristics), brand and price. It is interesting to note that food and beverage
sensory characteristics are very often associated with product names (Bárcenas et al.,
2001). We assume that the taste “association” of familiar brands is more stable and
stronger anchored in the mind of consumers compared to unfamiliar brands. This
could mean that consumers pay more attention to the taste of unfamiliar brands,
because there is no a priori knowledge about it. If so, the taste of unfamiliar brands
should have a bigger influence during CA on consumer preferences compared to the
taste of familiar brands. Based on this assumption we phrase H2:
H2. The influence of tasting is stronger in CA with unfamiliar brands than with
familiar ones.
5. Methodology and research process Brand familiarity
5.1 Test-product
Owing to the fact that the experiment took place in Croatian cafe bars a typical
Croatian beer (Ožujsko) and a strong international brand (Stella Artois) were chosen as
well known (familiar) beer brands. Together they represent the market leaders in
Croatia among domestic and imported beers respectively (Rajh et al., 2001, 2003;
Dujmicic et al., 2003). Additionally Austrian beer brands, only covering a small fraction 567
of market share, were used for the beer samples offered to the test persons as
unfamiliar brands.

5.2 Pre-research focus group and survey


In order to determine the beer attributes and their levels used in CA, a two-step
pre-research study was conducted with students from the Faculty of Agriculture
University of Zagreb. A focus group of eight students discussed beer attributes that
consumers usually consider as important for their choice. In a second step, a survey
was carried out with 30 students to detect the importance of selected beer attributes on
a five-point Likert scale (Likert, 1932).
Respondents were additionally asked to evaluate coded samples of beers in respect
to bitterness (weakly, moderate, highly), CO2 content (low, moderate, high) and fullness
(full taste, light taste) on a five-point Likert scale. By applying these verbal attributes
test persons evaluated the same beer products differently. Because of the
contradictional nature of these results we decided to use the attribute taste as a
between-subjects factor.

5.3 Research design


Based on the results of the preliminary study, four beer attributes were selected for CA:
brand, package type, package size and price (see Table I).
In order to control the influence of brand on respondents’ preferences, the CA was
divided into two experiment groups. Half of the respondents had to evaluate familiar

Level
Attribute Group G1 and G3 Group G2 and G4

Brand Ožujsko beer R


Stella Artois S
Package type Draught
Bottle
Package size
0.50 litre

0.33 litre
Price
8 Kunaa (e1.08)

12 Kunaa (e1.62)

16 Kunaa (e2.16) Table I.


Beer attributes and levels
Note: a Kuna is the Croatian currency used in CA
BFJ beer brands and the other half beer without a familiar brand label. Ožujsko beer and
112,6 Stella Artois were chosen as familiar beer brands. As beers without a familiar brand
label, two imported Austrian beer brands with limited market access were used and
they were labelled as “R” and “S”. Despite the fact that single letter brands are not
common on the market, we decided to use them instead of fictitious brand names
because for consumers unfamiliar brands are merely replaceable symbols, words,
568 names, colours or even single letters. Each of these two experimental groups was
further divided into two sub-groups: one with and one without tasting (Figure 1).
The levels of the three other attributes were chosen according to the prevalent beer
choice in popular Zagreb cafe bars. The two levels of package type were draught and
bottled beer. Package size had also two levels: 0.5 and 0.33 litters. Only beer price had
three levels: 8 Kuna, 12 Kuna, and 16 Kuna (for amount in Euros see Table I).
Hence, the main research comprised a survey with four independent respondent
groups coming from the same population, namely beer consumers in cafe bars in
Zagreb and surrounding suburbs. For all four groups verbal stimulus presentation
method was used. The differentiating variables are:
(1) Group G1 – familiar beer brands without tasting.
(2) Group G2 – unfamiliar beer brands without tasting.
(3) Group G3 – familiar beer brands with tasting.
(4) Group G4 – unfamiliar beer brands with tasting.

5.4 Main CA experiment


The main CA experiment was carried out in 2004 with visitors to different cafe bars in
Zagreb (8) and suburbs (2). The sample consisted of 403 beer consumers divided in four
independent sub-groups of the same size depending on the stimulus presentation
method and the beer brands they had to evaluate: group G1 (n ¼ 100), group G2
(n ¼ 100), group G3 (n ¼ 100) and group G4 (n ¼ 103). Bar visitors were chosen to
participate in the survey randomly with an equal number of respondents in each group
surveyed in each bar. The conditions are certainly not comparable to a laboratory
experiment, but the intention was to test under real purchasing and consumption
situations.
In addition to the CA experiment, a structured questionnaire was used consisting of
a few questions regarding respondents’ socio-economic characteristics, beer purchase

Figure 1.
Research design
and beer consumption behaviour, importance of a number beer attributes and Brand familiarity
preferences for these attributes.
Different respondent samples have also been used in comparable studies aiming to
test the validity of various CA methods (Ernst and Sattler, 2000; Brusch et al., 2002;
Pullman et al., 1999). The chi-square test showed that there were no differences
between groups in respect to respondents’ demographic and socio-economic
characteristics ( p . 0.05). Therefore it can be concluded that the four groups were 569
comparable and differences in results would not be influenced by systematic deviation
in the groups (Hensel-Börner and Sattler, 2000b).
Conjoint data were collected by means of the full profile method (the most common
method of data collection in conjoint research (Gil and Sánchez, 1997)). A fractional
factorial design was chosen and ten stimuli were derived by applying the Orthoplan
procedure of SPSS (see SPSS Inc., 1997). Otherwise a complete factorial design of
2 £ 2 £ 2 £ 3 would encompass 24 stimuli: a number, which would make the
experiment for the test persons far too complicated. The statistical software SPSS
Conjoint uses fractional factorial designs, which present an appropriate fraction of the
possible alternatives. Fractional factorial designs are experimental designs consisting
of a carefully chosen subset (fraction) of the experimental runs of a full factorial design.
The subset or fraction is chosen so as to exploit the sparsity-of-effects principle to
access information about the most important features of the problem studied, while
using considerably fewer resources than a full factorial design (SPSS Inc., 1997).
An additive partworth approach was utilised to estimate consumer preferences:
The respondent’s task is to rank or score each profile from most to least preferred, most to
least likely to purchase, or some other preference scale. From these rankings or scores,
conjoint analysis derives utility scores for each factor level. These utility scores, analogous to
regression coefficients, are called part-worths and can be used to find the relative importance
of each factor (SPSS Inc., 1997).
Within the conjoint task, respondents had to rank a set of ten stimuli according to their
preferences. We decided to use ranking because it provides similar results compared to
ratings. Carmone et al. (1978) compared the over-all conjoint model goodness of fit
under several forms of input data (raw data, rankings, and six-point rating scales), and
found conjoint analysis to provide robust results regardless of the type of input data
scales, with superior recoveries for rankings in some cases.
Descriptions of beer concepts (stimuli) were printed on separate, coloured cards. The
card order given to the respondents was randomised. Those respondents who
evaluated CA with tasting had to taste two beers before ranking. The respondents had
the possibility of tasting beers as many times as they wanted during the conjoint task.
The beer was served cold in transparent glasses. To prevent the test persons from
mixing up beers, the brand name (familiar or unfamiliar) was written on additional
cards under each glass. Each conjoint task contained five additional, holdout stimuli.
The holdouts represented beers already existing on the market or possible realistic
constructs of beers. These stimuli had to be ranked according to purchase intention but
were not be used for partworth calculation in CA. To facilitate the respondents’ tasks,
holdout and main cards were differently coloured, as were cards representing familiar
and unfamiliar beers (four colours altogether) (Figure 2). Holdouts served to assess the
predictive validity and were the basis for our hypothesis tests.
BFJ
112,6

570

Figure 2.
An example of CA cards

5.5 Data analysis


CA validity was measured with five criteria: face validity, convergent validity, internal
validity, predictive validity and subjective evaluation of conjoint task.
For face validity we used Mann-Whitney and Kolmogorov-Smirnov test to verify
statistical differences between the importances of single attributes. Convergent
validity was verified by applying ANOVA and Kolmogorov-Smirnov test (H1 and H2).
ANOVA was applied to scrutinise differences between distributions of utilities among
respondents in different groups (H1) (Hensel-Börner and Sattler, 2000b).
To test for convergent validity, partworths were calculated for each respondent
followed by a comparison of distribution of these partworths between tasting and
non-tasting respondents by applying a Kolmogorov-Smirnov test.
To verify differences in internal validity between groups we compared individual
Kendal’s Tau coefficients among groups by means of non-parameter Mann-Whitney
and Kolmogorov-Smirnov tests (similar comparison made Ernst and Sattler, 2000).
Predictive validity was evaluated by measuring the first-hit choice rate. By
applying chi-square test to the first-hit choice rates between respondents groups we
verified differences in respect to predictive validity.
To measure respondents’ experience with the different CA experiment settings
(with and without tasting) we collected their subjective evaluations concerning
complexity and interest of the conjoint task. Using the five-point Likert scale
respondents in this study were asked how difficult (complexity) and how interesting
the conjoint task was to evaluate. Non-parameter Mann-Whitney and
Kolmogorov-Smirnov tests were applied to test for significant differences concerning
the subjective evaluations of CA tasks among different respondents groups.

6. Results
6.1 Sample characteristics
Altogether 403 respondents took part in this study of whom 266 were male (66 per cent)
and 137 female (34 per cent). The number of male respondents is significantly higher
but this is in accordance with the sex structure of beer consumers in general (Pettigrew,
2002). Respondents were aged between 17 and 72, with an average of 28.97 (^ 10.078)
years. The majority of respondents were in the age group between 21 and 35 years, Brand familiarity
which reflects the age of most visitors to cafe bars. The majority of respondents had
finished secondary school (72 per cent), a further 25 per cent had higher education and
3 per cent of respondents had only completed primary school.

6.2 Influence of tasting on validity of CA results


Previous research has shown that brand name plays an important role in the formation
571
of beer preferences (Allison and Uhl, 1964; Guinard et al., 2001) and that beer quality is
often associated with its brand name. Hence, in this study we investigated whether
tasting as a presentation method has an influence on the validity of CA with familiar
and unfamiliar brands. Table II gives an overview concerning the outcomes of the
different validity tests. The results are explained in order of the several validity
measures; each first in respect to differences between CA with tasting compared to CA
without tasting and secondly in respect to differences between the groups G1 to G4.
Concerning face validity Mann-Whitney and Kolmogorov-Smirnov tests confirmed
that there were significant differences between the importance of price and other beer
attributes by all consumers ( p , 0.01). Nevertheless, the difference between brand
importance and the importance of package size and package type was notable only
among respondents who tasted beer ( p , 0.01). The respondents who evaluated beer
concepts only according to their verbal description did not significantly distinguish

CA with tasting versus CA without


Validity measure tasting G1-G4

Face validity Higher face-validity for CA with Smaller face validity in G2


tasting compared to other groups
Convergent validity Almost no convergent validity, e.g. No differences in importance
the methods result in different structure among groups G1, G3 and
preference structures G4
Respondents from G2 have
significantly different preference
structure
Internal validity No statistically significant Kendal’s tau for G1 and G3 (known
differences brands) rather similar
Difference between Kendal’s tau of
G2 and G4 (unknown brands) much
higher
Predictive validity (First- No statistically significant No significant differences between
choice-hit-rate) differences G1 and G3
Significant difference between G2
and G4
Subjective evaluation of No statistically significant No statistical differences between
CA task differences in CA task complexity G1-G4 regarding difficulty of
CA with tasting more interesting conjoint task
compared to verbal CA No statistical differences regarding
interestingness of conjoint task
between G2 and G4 Table II.
G1 consider CA task more Summary table of
interesting than G3 validity tests
BFJ between these attributes. These results suggest that stimulus presentation with tasting
112,6 has higher face validity (H1).
As expected, the most important beer attributes for three respondents groups (G1,
G3 and G4) were brand name and price. However, the respondents who evaluated
unfamiliar beer brands without tasting (G2) considered brand name to be less
important than other beer attributes (Figure 3). This indicates the smaller face validity
572 in G2 group of respondents.
In respect to convergent validity the results of the Kolmogorov-Smirnov test for
partworths between tasting and non-tasting respondents (Table III) revealed a
statistical difference between groups for all attributes (brand, p , 0.01; price, package
type and package size, p , 0.05).
Since the importance weights do not have the same structure in these groups, it may
be concluded that there is almost no convergent validity, e.g. the methods result in
different preference structures.
Statistical tests (ANOVA) used to examine the difference between the partworths of
the four groups showed no differences in importance structures between groups G1, G3
and G4, while respondents from G2 had significantly different preference structures
( p , 0.05). This is a clear confirmation of H2 “the influence of tasting is stronger in CA
with unfamiliar brands than with familiar ones”, because G2 is the group with
unfamiliar brands without tasting.
Internal validity measured as the Kendall tau coefficient produced high values in
both respondents’ samples; a slightly higher value was found for respondents who did
not taste beers (0.929) compared with other respondents (0.857). However, the Kendall
tau differences were not statistically significant.
Kendal tau coefficients for respondents from G1 and G3 (familiar brands) were quite
similar (0.93 and 1.0 respectively). Nonetheless, the difference between Kendal tau
coefficients for G2 and G4 (unfamiliar brands) was much higher; its value was 0.79 for
G2 and 0.93 for G4. This indicator of internal validity confirms the hypothesis that
tasting has a stronger influence on the validity of CA with unfamiliar brands than with
familiar brands.
For predictive validity the first-choice hit rate was applied. The hit rate of the
respondents who evaluated beers according to pure verbal presentation method was 54
per cent compared with 59 per cent for respondents who tasted beer. Although a
statistical difference between these two rates was not confirmed, the results indicate
the tendency towards higher predictive validity for CA with tasting.
The chi-square test showed that there was no significant difference in the
first-choice hit rate between respondents in groups G1 and G3. Conversely, a
significant difference was found between the hit rate of G2 and G4 respondents
(x 2 ¼ 2:837, df ¼ 1, p ¼ 0:061). The first-choice hit rates were 46 per cent for G2 and
58 per cent for G4 (Table IV).
The test of proportions indicated that there were differences between first-choice hit
rate and share of correctly estimated stimuli in a random model in all respondents
groups (G1-G4) ( p , 0.01). The minor improvement compared with the random model
appeared in CA without tasting with unfamiliar brands (32.5 per cent), while the
improvement in all other groups was similar and ranged from 47.3 to 52.5 per cent.
This would indicate that CA without familiar brands delivers higher predictive
validity when the test persons can taste the products.
Brand familiarity

573

Figure 3.
The importance of beer
attributes in different
respondent groups
BFJ Both respondent groups evaluated the conjoint task as similarly complex (average
112,6 evaluation of conjoint task difficulty on the five-point Likert scale was 3.4 and 3.3,
respectively; 1 ¼ not complex, 5 ¼ highly complex). On the other hand, a significant
difference was apparent between respondents’ evaluation of conjoint task
attractiveness. The respondents with pure verbal presentation evaluated the task
with an average grade of 2.8 while the respondents who tasted beers considered the
574 conjoint task to be more interesting (average grade 3.1; 1 ¼ not interesting, 5 ¼ very
interesting, p , 0.05).
There were no statistical differences between respondent groups G1-G4 regarding
difficulty of conjoint task (3.3 for G1, G2, G4 and 3.5 for G3). By contrast, the respondents
who evaluated familiar brands without tasting considered the conjoint task as the least
interesting (average grade 2.6) while all other respondents considered it as more
interesting (G2 3.1; G3 3.0; G4 3.2). The difference between respondents of G1 and G3
concerning interestingness was statistically significant (p ¼ 0:01). Although the
respondents who evaluated unfamiliar beers with tasting gave a slightly higher score for
the conjoint task, there were no statistical differences between groups G2 and G4.

7. Discussion
In this study, the hypothesis was confirmed that tasting has a greater influence on the
validity of CA with unfamiliar product brands. Tasting had no influence on the validity
of CA with familiar beer brands i.e. the results of verbal CA and CA with tasting
performed with familiar brands were similar (equal). One could argue that consumers
had already formed a stable judgment of beers with familiar brands and that tasting
did not influence this perception. These findings are similar to those of Bárcenas et al.
(2001) that consumer expectations, derived from previous experience about a particular
product, influence their preferences and product acceptance. Therefore, it may be

Kolmogorov-Smirnov
CA test
Importance (%) Verbal With tasting Z p
Table III.
The importance of single Brand 25.7 38.5 2.865 0.000
beer characteristics in Package size 20.0 16.6 1.509 0.021
verbal CA and CA with Package type 23.1 16.8 1.442 0.031
tasting Price 31.2 28.1 1.495 0.023

Without tasting With tasting


G1 known G2 unknown G3 known G4 unknown
brands brands brands brands

Number of respondents 100 100 100 103


First-choice-hit-rate (%) 62.0 46.0 59.6 57.8
Percent of improvement compared to
random modela 52.5 32.5 49.5 47.3
Table IV.
a
First-choice-hit-rate in Notes: ¼ 100 £ (% correctly estimated 2 % correctly estimated in random model)/(100 2 %
different groups correctly estimated in random model)
concluded that previous experience with familiar beer brands is sufficient to express Brand familiarity
preferences in a stable way. It is interesting that some respondents asked to taste
familiar beer brands Ožujsko beer and Stella Artois refused to taste the beers, replying
that they were familiar with their taste.
This observation could also be a confirmation of the “quality cues”-theory.
Obviously our respondents have subsumed the quality cue “taste” under the
“umbrella” of the quality cue “brand”. This is in accordance with Verbeke and Ward 575
(2006), who state that uncertainty or perceived difficulty to evaluate quality increases
consumers’ usage of extrinsic quality cues. From that point-of-view the presence of
familiar brand names seems to have weakened or overridden the “taste information”.
There were no differences in validity measures for the two groups with familiar
brands. It can therefore be concluded that when performing CA with familiar brands
with a well defined image, it is not necessary to use CA with tasting since tasting
increases research complexity and costs and does not achieve better results.
By contrast, respondents were unable to consistently evaluate their preferences
towards unfamiliar beers only according to their external attributes. After tasting these
products, respondents’ answers were much more consistent. Therefore, it may be
concluded that the structure of attribute importance and preferences in CA varies
significantly when respondents also taste products unfamiliar with, besides receiving
verbal descriptions. The research results showed that after tasting the expected
structure of attribute importance was obtained, i.e. the same structure as obtained in
CA with familiar brands. On the other hand, respondents who evaluated a purely
verbal description of unfamiliar beers considered brand name less important than
expected. These results are consistent with the results of the study by Arvola et al.
(1999), which revealed that purchase intention could only be well predicted after
respondents had tasted unknown cheeses.
Apart from face validity, predictive validity was higher when respondents tasted
unfamiliar beers before evaluating conjoint stimuli. It may therefore be concluded that
stimulus presentation method tasting gives better results than pure verbal CA in the
case of unfamiliar brands.
Although we expected that the complexity of the conjoint task without tasting
would be more difficult for respondents, the study did not confirm this assumption.
One possible explanation is that because of lack of interest, respondents made little
cognitive effort to solve the conjoint task. CA with tasting of unfamiliar beer brands
was evaluated as most interesting, but the differences between presentation methods
were not statistically validated.

8. Recommendations for future research and CA stimuli presentation


Beer consumers have relatively well defined preferences for brands and very often
associate beer taste with its brand and image. In this study, domestic and market
leaders as familiar brands were investigated. It would be interesting to repeat this
study with beer brands familiar to consumers but consumed less often. Since the study
showed that tasting does not have an influence on the CA validity of familiar product
brands, similar research could be carried out with product without well familiar brands
or products in which brand preferences are not well expressed. It is possible that CA
with products of higher heterogeneity, especially regarding their sensory
characteristics (e.g. fruit yoghurt), could produce different results.
BFJ The results of this study indicate that tasting should be used as a stimulus
112,6 presentation method for CA with food and beverage products/brands, which are
unfamiliar to the consumers. When testing familiar brands and brands with
established perceptions, simpler and less expensive verbal stimulus presentation can
be used. The decision regarding the stimulus presentation method in CA designed to
test the acceptance of a new product should be based on the product development
576 stage. Sattler (1994) states that in the idea screening and concept testing stage, methods
that do not necessarily have very high result validity are acceptable. At that stage, the
use of a verbal presentation form is recommended. At the stage of determining intrinsic
product characteristics (including sensory attributes) it is appropriate to use CA with
tasting. When assessing extrinsic product attributes (e.g. package, appearance, price
etc.) it is not necessary to use CA with tasting.
CA used to explore the market situation (e.g. market share) should be carried out
with the tasting presentation method. This is especially the case when studying
products less known to consumers.
Motivated respondents will produce better research results (Strebinger et al., 2000b).
This study showed that tasting increases the attractiveness of the conjoint task as an
interesting exercise, i.e. motivating respondents to evaluate conjoint stimuli. Therefore
it is recommended to include tasting, as CA can become tedious for the respondents
because of the high number of product cards.

References
ACNielsen (2006), “What’s hot around the globe: insights on growth in food and beverages”,
Executive News Report, December, p. 47.
Allison, R.I. and Uhl, K.P. (1964), “Influence of beer brand identification on taste perception”,
Journal of Marketing Research, Vol. 1, pp. 36-9.
Arvola, A., Lähteenmäki, L. and Tuorila, H. (1999), “Predicting the intent to purchase unfamiliar
and familiar cheeses: the effect of attitudes, expected liking and food neophobia”, Appetite,
Vol. 32, pp. 113-26.
Backhaus, K., Erichson, B., Plinke, W. and Weiber, R. (1996), Multivariate Analysemethoden. Eine
anwendungsorientierte Einführung, Springer, Berlin.
Bárcenas, P., Pérez de San Román, R., Pérez Elortondo, F.J. and Albisu, M. (2001), “Consumer
preference structures for traditional Spanish cheeses and their relationship with sensory
properties”, Food Quality and Preference, Vol. 12 No. 4, pp. 269-79.
Brusch, M. and Baier, D. (2002), “Realitätsnähere Produktpräsentation in der Marktforschung –
Multimedia und Conjointanalyse”, Forum der Forschung, Vol. 13, pp. 86-9.
Brusch, M., Baier, D. and Treppa, A. (2002), “Conjoint analysis and stimulus presentation:
a comparison of alternative methods”, in Jajuga, K., Sokolowski, A. and Bock, H.H. (Eds),
Classification, Clustering, and Analysis, Springer, Berlin, pp. 203-10.
Carmone, F.J., Green, P.E. and Jain, A.K. (1978), “The robustness of conjoint analysis:
some Monte-Carlo results”, Journal of Marketing Research, pp. 300-3, 15 May.
Carroll, J.D. and Green, P.E. (1995), “Psychometric methods in marketing research: Part I,
conjoint analysis”, Journal of Marketing Research, Vol. 32 No. 4, pp. 385-91.
Cheng, H.W., Clarke, A.D. and Heymann, H. (1990), “Influence of selected marketing factors on
consumer response to restructured beef steaks: a conjoint analysis”, Journal of Sensory
Studies, Vol. 4, pp. 165-78.
Daems, V. and Delvaux, F. (1997), “Multivariate analysis of descriptive sensory data on 40 Brand familiarity
commercial beers”, Food Quality and Preference, Vol. 8 Nos 5/6, pp. 373-80.
Dujmičić, K., Renko, S. and Renko, N. (2003), “A case study of the Croatian beer market structure
and performances”, British Food Journal, Vol. 105 No. 3, pp. 193-203.
Ernst, O. and Sattler, H. (2000), “Validität multimedialer Conjoint-Analysen, Ein empirischer
Vergleich alternativer Produktpräsentationsformen”, Marketing ZFP, Vol. 22 No. 2,
pp. 161-72. 577
Gil, J.M. and Sánchez, M. (1997), “Consumer preferences for wine attributes: a conjoint approach”,
British Food Journal, Vol. 99 No. 1, pp. 3-11.
Govindasamy, R., Italia, J. and Liptak, C. (1997), “Quality of agricultural produce: consumer
preferences and perceptions”, New Jersey Agricultural Experiment Station, P-02137-1-97.
Green, P.E. and Helsen, K. (1989), “Cross-validation assessment of alternative to individual-level
conjoint analysis: a case study”, Journal of Marketing Research, Vol. 26, pp. 346-50.
Green, P.E. and Srinivasan, V. (1990), “Conjoint analysis in marketing: new developments with
implications for research and practice”, Journal of Marketing, Vol. 54, pp. 3-19.
Green, P.E., Krieger, A.M. and Wind, Y. (2001), “Thirty years of conjoint analysis: reflections and
prospects”, Interfaces, Vol. 31 No. 3, pp. 56-73.
Guinard, J.X., Uotania, B. and Schlich, P. (2001), “Internal and external mapping of preferences
for commercial lager beers: comparison of hedonic ratings by consumers blind versus with
knowledge of brand and price”, Food Quality and Preference, Vol. 12 No. 4, pp. 243-55.
Grunert, K.G. (2002), “Current issues in the understanding of consumer food choice”, Trends in
Food Science and Technology, Vol. 13, pp. 275-85.
Grunert, K.G., Bech-Larsen, T. and Bredahl, L. (2000), “Three issues in consumer quality
perception and acceptance of dairy products”, International Dairy Journal, Vol. 10,
pp. 575-84.
Helgesen, H., Solheim, R. and Næs, T. (1998), “Consumer purchase probability of dry fermented
lamb sausages”, Food Quality and Preference, Vol. 9 No. 5, pp. 295-301.
Hensel-Börner, S. and Sattler, H. (2000a), “Ein empirischer Validitätsvergleich zwischen der
Customized Computerized Conjoint Analysis (CCC), der Adaptive Conjoint Analysis
(ACA) und Self-Explicated-Verfahren”, Zeitschrift für Betriebswirtschaft (ZfB), Vol. 70
No. 6, pp. 705-27.
Hensel-Börner, S. and Sattler, H. (2000b), “Validity of customized and adaptive hybrid conjoint
analysis”, in Decker, R. and Gaul, W. (Eds), Classification and information processing at the
turn of the millennium, Proceedings of the 23rd Annual Conference of the Gesellschaft für
Klassifikation e.V., Berlin, pp. 320-9.
Holbrook, M.B. and Moore, W.L. (1981), “Feature interactions in consumer judgments of verbal
versus pictorial presentations”, Journal of Consumer Behavior, Vol. 8, pp. 103-13.
Huber, J. (1987), “Conjoint Analysis: How we got here and where we are”, Sawtooth Software
Conference Proceedings, available at: www.sawtoothsoftware.com
Huber, J.C., Wittink, D.R., Fiedler, J.A. and Miller, R.L. (1991), “An empirical comparison of ACA
and full profile judgments”, 1991 Sawtooth Software Conference Proceedings, Ketchum,
ID: Sawtooth Software, pp. 189-202, available at: www.populus.com/tech_papers/
compare_aca&fullprof.pdf
Jaeger, S.R., Hedderley, D. and MacFie, H.J.H. (2001), “Methodological issues in conjoint analysis:
a case study”, European Journal of Marketing, Vol. 35 Nos 11/12, pp. 1217-39.
BFJ Kotler, P. and Keller, L.K. (2005), Marketing Management, 12th ed., Prentice-Hall International,
London.
112,6
Krapp, A. and Sattler, H. (2001), “Rethinking preference measurement”, Proceedings of the
30th Annual Conference of the European Marketing Academy (Bergen, Norwegen),
available at: www.henriksattler.de/publikationen/HS_046.pdf
Leight, T.W., MacKay, D.B. and Summers, J.O. (1984), “Reliability and validity of conjoint
578 analysis and self-explicated weights: a comparison”, Journal of Marketing Research,
Vol. 20, pp. 350-67.
Likert, R. (1932), “A technique for the measurement of attitudes”, Archives of Psychology, Vol. 140,
pp. 1-55.
Pettigrew, S. (2002), “A grounded theory of beer consumption in Australia”, Qualitative Market
Research: An International Journal, Vol. 5 No. 2, pp. 112-22.
Pullman, M.E., Dodson, K.J. and Moore, W.L. (1999), “A comparison of conjoint methods when
there are many attributes”, Marketing Letters, Vol. 10 No. 2, pp. 1-14.
Rajh, E., Vranešević, T. and Tolić, D. (2001), “Tržišna vrijednost maraka u prehrambenoj
industriji Republike Hrvatske”, Zbornik radova XVII. kongresa CROMAR-a Hrvatske:
Marketing države – marketing hrvatske države, Zagreb/Pula, pp. 231-5.
Rajh, E., Vranešević, T. and Tolić, D. (2003), “Croatian food industry: brand equity in selected
product categories”, British Food Journal, Vol. 105 Nos 4/5, pp. 263-73.
Roeber, D.L., Scanga, J.A., Belk, K.E. and Smith, G.C. (2002), “Consumer attitudes and
preferences”, 2002 Animal Sciences Research Report, The Department of Animal Sciences,
Colorado State University, Fort Collins, CO.
Roininen, K. (2001), “Evaluation of food choice behavior: development and validation of health
and taste attitude scales”, academic dissertation, Department of Food Technology,
University of Helsinki, Helsinki, available at: http://ethesis.helsinki.fi/julkaisut/maa/elint/
vk/roininen/evaluati.pdf
Sattler, H. (1994), “Die Validität von Produkttests, Ein empirischer Vergleich zwischen
hypothetischer und realer Produktpräsentation”, Marketing ZFP, Vol. 16 No. 1, pp. 31-41.
Schutz, H.G., Judge, D.S. and Gentry, J. (1986), “The importance of nutrition, brand, cost and
sensory attributes to food purchase and consumption”, Food Technology, Vol. 40 No. 11,
p. 79.
SPSS Inc. (1997), SPSS Conjointe 8.0, SPSS Inc., Chicago, IL.
Strebinger, A., Hoffmann, S., Schweiger, G. and Otter, T. (2000a), “Zur Realitätsnähe der
Conjointanalyse”, Marketing ZFP, Vol. 1 No. 1, pp. 55-74.
Strebinger, A., Hoffmann, S., Schweiger, G. and Otter, T. (2000b), “Verbal versus pictorial stimuli
in conjoint analysis: the moderating effect of involvement and hemisphericity”,
in Gundlach, G.T. and Murphy, P.E. (Eds), Enhancing Knowledge Development in
Marketing, 2000 AMA Educators’ Proceedings, Vol. 11, American Marketing Association,
Chicago, IL, pp. 182-8, available at: www.werbelehrgang.at/service/download/files/
ConjointAMA2000.pdf
Toubia, O. (2001), “Interior-point methods applied to internet conjoint analysis”, Master of
Science in Operations Research, Massachusetts Institute of Technology, Cambridge, MA.
Tscheulin, D.K. (1991), “Ein empirischer Vergleich der Eignung von Conjoint-Analyse und
‘Analytic Hierarchy Process’ (AHP) zur Neuproduktplanung”, Zeitschrift für
Betriebswirtschaft, Vol. 61 No. 11, pp. 1267-80.
Verbeke, W. and Ward, R.W. (2006), “Consumer interest in information cues denoting quality, Brand familiarity
traceability and origin: an application of ordered probit models to beef labels”, Food
Quality and Preference, Vol. 17, pp. 453-67.
Vickers, Z.M. (1993), “Incorporating tasting into a conjoint analysis of taste, health claim, price
and brand for purchasing strawberry yogurt”, Journal of Sensory Studies, Vol. 8,
pp. 341-52.
Vriens, M., Loosschilder, G.H., Rosbergen, E. and Wittink, D.R. (1998), “Verbal versus realistic 579
pictorial representation in conjoint analysis with design attributes”, Journal of Product
Innovation Management, Vol. 15, pp. 455-67.
Wansink, B. (2003), “Measuring consumer response to food products: sensory tests that predict
consumer acceptance”, Food and Quality Preference, Vol. 14, pp. 23-6.

Corresponding author
Marija Cerjak can be contacted at: mcerjak@agr.hr

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com


Or visit our web site for further details: www.emeraldinsight.com/reprints

You might also like