Food Quality and Preference 22 (2011) 581591

Contents lists available at ScienceDirect

Food Quality and Preference

journal homepage: www.elsevier.com/locate/foodqual

Identifying ideal products using three different consumer proling

methodologies. Comparison with external preference mapping
Gastn Ares a,, Paula Varela b, Germn Rado a, Ana Gimnez a
a
Seccin Evaluacin Sensorial, Departamento de Ciencia y Tecnologa de Alimentos, Facultad de Qumica, Universidad de la Repblica (UdelaR),
General Flores 2124, C.P. 11800 Montevideo, Uruguay
b
Instituto de Agroqumica y Tecnologa de Alimentos, Avda Agustn Escardino 7, 46980 Paterna, Valencia, Spain

a r t i c l e

i n f o

Article history:
Received 5 October 2010
Received in revised form 8 April 2011
Accepted 11 April 2011
Available online 14 April 2011
Keywords:
Consumer studies
Product development
Optimization
Check-all-that-apply
Projective mapping

a b s t r a c t
The aim of the present work was to identify consumers ideal product by three consumer proling techniques and to compare the agreement among the three techniques. Two studies were carried out in
which consumers evaluated seven samples of orange-avoured powdered drinks. In the rst study 108
consumers scored their overall liking, whereas in the second one three groups of 50 consumers evaluated
the sensory characteristics of the samples using projective mapping, a check-all-that-apply question or
intensity scales. After completing the task consumers were asked to identify their ideal product.
The different approaches yielded similar information regarding the sensory characteristics of the products and consumers ideal product, providing similar recommendations for product improvement. However, they differed in the position of consumers optimum product within the sensory space dened by
the sensory characteristics of the evaluated samples. Projective mapping identied the consensus position of the ideal product within the range of sensory characteristics of the evaluated samples, providing
similar results than external preference mapping. Differences and similarities between the methodologies are discussed, as well as potential applications.
2011 Elsevier Ltd. All rights reserved.

1. Introduction
Despite the fact that new food product innovation and development is necessary for food companies to survive in todays highly
competitive market, the vast majority of new food products fail
(Stewart-Knox & Mitchell, 2003). The low innovation and high failure rate of new food products suggests that the methodology used
for new food product development is seriously awed (Rudolph,
1995; Stewart-Knox & Mitchell, 2003). Thus, product development
teams need more efcient product development techniques.
One of the most important steps of new product development
process is product optimization. This step aims to identify ideal
products, i.e. products that maximize consumers acceptance, usually measured as liking (Lagrange & Norback, 1987). The most common approach to product optimization is to ask consumers to rate
their overall liking of a large set of products and characterize the
sensory properties of those products using a trained assessors
panel. Then, both data sets are combined using regression analysis
to identify the sensory characteristics of the ideal product (van
Trijp, Punter, Mickartz, & Kruithof, 2007). One of the most common
methodologies during this step of product optimization is external
Corresponding author. Tel.: +598 2 9245735; fax: +598 2 9241906.
E-mail address: gares@fq.edu.uy (G. Ares).
0950-3293/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.foodqual.2011.04.004

preference mapping (Arditti, 1997; Carroll, 1972; McEwan, 1996;

Schlich, 1995).
External preference mapping refers to a series of approaches to
relate consumers overall liking scores to a conguration of products determined by their sensory characteristics, i.e. the products
conguration in a sensory space (Carroll, 1972; Meullenet, Lovely,
Threlfall, Morris, & Striegler, 2008b). Individual consumer scores
are tted into the product conguration in the sensory space using
a regression model (Danzart, 1998). After this, individual models
are overlapped to create a density plot of consumer acceptance
and the points of maximum density are considered the ideal products (Danzart, Sieffermann, & Delarue, 2004).
Another tool frequently used for the understanding of the
descriptive sensory attributes that drive consumer preferences is
internal preference mapping. In this methodology the sensory prole of the product is related to preference ratings from a representative sample of consumers, using only consumer data to
determine consumer preference patterns. Afterwards, the sensory
description is linked by regressing it onto the consumer map (Lawlor & Delahunty, 2000). Alternative internal preference mapping
approaches were also developed, such as Landscape Segmentation
Analysis (LSA), which unfolds an acceptance space and estimates
individual ideal locations in that space; vectors of descriptive attributes (measured by a panel) are then overlaid on the acceptance

582

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

space (Ennis, Palen, & Mullen, 1988; Ennis & Rousseau, 2004; Rousseau
& Ennis, 2008; Thompson, Drake, Lopetcharat, & Yates, 2004).
In these approaches consumers are only asked about their liking, and therefore information about how they perceive the sensory characteristics of the products is not gathered. However,
trained assessors could describe the product differently or take
into account attributes that may be irrelevant for consumers (ten
Kleij & Musters, 2003). Besides, since liking data is regressed onto
principal components, it could be difcult to translate preference
directions into product sensory attributes (Guinard, Uotani, & Schlich, 2001). Moreover, during food product development the application of quantitative descriptive analysis with trained assessors
remains a very time-consuming approach since the vocabulary
and associated training must be adapted to each product (Faye
et al., 2006).
Another possible approach for product optimization would be
asking consumers to describe the sensory characteristics of the
products and also to identify their ideal product. According to Risvik, McEwan, and Rodbotten (1997) the best way to understand
consumer preferences is consumer data. In particular, getting consumer feedback about sensory description of products has become
of great interest in the last decade. In this context, several consumer proling methodologies have increased popularity, among
which intensity scales, check-all-that-apply questions and projective mapping have had a central role.
Asking consumers to rate the intensity of different sensory attributes using scales has been reported to be a good alternative to the
classical sensory prole provided by trained assessors (Husson, Le
Dien, & Pags, 2001; Worch, L, & Punter, 2010). Husson et al.
(2001) and Worch et al. (2010) reported that sensory proles obtained with consumers using intensity scales meet discrimination
and reproducibility requirements, being an interesting alternative
when companies face difculties to use trained assessors panels.
Moskowitz (1996) also concluded that consumers are able to assess the sensory characteristics of products using this type of
methodology.
Check-all-that-apply questions (CATA) consist of a list of words
or phrases from which respondents should select all the words
they consider appropriate to describe a product. This type of question has been used in consumer studies to determine which sensory attributes consumers perceive in different food products
(Adams, Williams, Lancaster, & Foley, 2007; Ares, Deliza, Barreiro,
Gimnez, & Gmbaro, 2010; Meullenet, Lee, & Dooley, 2008a).
Compared to just-about-right or intensity questions, CATA questions seem easier and more natural for consumers (Adams et al.,
2007).
Another alternative for consumer proling is the application of
projective mapping to quantify individual perception of overall
similarity and dissimilarity between products (Risvik, McEwan,
Colwill, Rogers, & Lyon, 1994). In this methodology consumers
are asked to provide a two dimensional projection of a group of
samples, according to their own criteria (Risvik et al., 1997). This
technique might be a useful and simple way to evaluate consumer
perception of food products.
Few studies have been published reporting the identication of
ideal products based on consumers perception of the sensory characteristics of food products. Epkong, Ngarmsak, and Winger (2006)
reported that consumers and trained assessors provided similar
and suitable information for the optimization of mango gel snacks.
Furthermore, van Trijp et al. (2007) used intensity and just-aboutright scales to optimize the sensory characteristics of natural yogurts, reporting that both methodologies yielded valid and similar
results.
In this context, the aims of the present work were to: (a) study
the ability of three consumer proling techniques (intensity scales,
check-all-that-apply questions and projective mapping) to identify

consumers ideal product, and (b) to compare the results among

the three techniques to those obtained using external preference
mapping.
2. Materials and methods
2.1. Powdered juice samples
The powdered drinks market in Latin America is currently one
of the most dynamics in terms of new avor developments and
new launches. In 2008, more than 62 million liters of powdered
beverages were drunk in Uruguay, a country with a population just
above 3 million. Regarded somehow as healthier than carbonated
beverages, and costing a fraction of their natural counterparts,
powdered drinks are chosen as an economic and healthier soft
drink (Varela, Ares, Gimnez, & Gmbaro, 2010). Therefore, it is
interesting to gather information about consumers perception of
the sensory characteristics of this type of product.
Seven samples of orange-avoured powdered juice drinks
(named samples 17), corresponding to different brands available
in the Uruguayan market were evaluated in the present study.
Samples were prepared by diluting the powders with tap water,
as recommended in the packages, adding 25% extra water (e.g.
samples for dilution in 1 L were diluted to 1.25 L), based on usage
and attitude data from previous consumer studies in the region.
Twenty-ve milliliters of each drink were served to consumers at
20 C in plastic containers. The seven samples were presented to
consumers coded with 3-digit random numbers, following a balanced rotation (multiple orthogonal Latin square). Mineral water
was available for rinsing between samples.
2.2. Consumer sample
Two-hundred and fty-eight people, all regular consumers of
the category participated in the present study. Participants ranged
in age from 18 to 60 and were 63% females and 37% males. They all
consumed orange-avoured non-carbonated drinks at least once a
week. Consumers were recruited from University campuses and
public places, based on their interest and availability to participate
in the study. Consumer tests were carried out in a sensory laboratory that was designed in accordance with ISO 8589 (ISO, 1988).
Evaluations were performed under articial daylight type illumination, temperature control (between 22 and 24 C) and air circulation. Data were collected in paper ballots, through self
administered questionnaires previously explained to each consumer in a 1:1 basis.
Two consumer studies were carried out on two consecutive
days. In the rst one, 108 consumers evaluated the overall liking
of the seven drinks. In the second study 150 consumers were randomly divided into 3 groups of 50 people, each of which evaluated
the samples using a different consumer proling methodology:
projective mapping, check-all-that-apply questions and intensity
scales.
2.2.1. Overall liking
In the rst study consumers were asked to try each of the drinks
and to evaluate their overall liking using a 9-point hedonic scale.
2.2.2. Projective mapping
For the projective mapping task, also known as Napping (Pags,
2005), consumers were rst asked to try each of the drinks and to
provide up to four words they consider appropriate to describe
each of them. After this, consumers were asked to place the samples on an A3 white sheet (60  40 cm), according to the similarities or dissimilarities between them. Consumers were explained

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

that they had to complete the task according to their own criteria
and that there were no right or wrong answers. They were also explained that two samples close together on the sheet would correspond to very similar samples and that if they perceived two
samples as very different they had to locate them very distant from
each other. After completing the task, consumers were asked to position their ideal drink on the sheet, relatively to the actual products they have just evaluated.
For each consumer map, the X and Y coordinates of each sample
were determined, considering the left bottom corner of the sheet
as origin of the coordinate system.
2.2.3. Check-all-that-apply
Consumers had to answer a check-all-that-apply (CATA) question with 19 hedonic and sensory attributes to describe the drinks.
The CATA question was composed of the following terms: very
sweet, sweet, not much sweet, bitter, not much orange avour,
off-avour, dark colour, light colour, diluted, yummy, disgusting,
orange avour, articial avour, natural, rough, soft, intense avour and aftertaste. These words were selected based on results
from a previous study in which consumers provided words to describe orange-avoured powdered drinks using an open-ended
question. Consumers were explained that they had to check all
the terms they considered appropriate to describe each of the
drinks. After evaluating the six samples consumers were asked to
check all the terms they consider appropriate to describe their
ideal orange-avoured powdered drink.
2.2.4. Intensity scales
Consumers had to rate the intensity of seven sensory attributes
using unstructured 10-cm scales. The attributes considered were
the following: colour, sweetness, orange avour, acidity, bitterness,
articial avour and off-avour. These attributes are the ones commonly used in quantitative consumers tests of the category when
gathering information about attribute levels for guiding product
development. After evaluating the seven samples consumers were
asked to rate the intensity of each of the evaluated attributes for
their ideal orange-avoured powdered drink. This procedure is also
known as the Ideal prole method.
2.3. Data analysis
2.3.1. Overall liking data
Analysis of variance (ANOVA) was performed on consumer
overall liking scores from consumers considering consumer and
sample as sources of variation. Mean ratings were calculated and
honestly signicant differences were checked using Tukeys test
(p < 0.05).
Internal preference mapping was carried out using a principal
component analysis on the correlation matrix of consumer individual liking data.
Hierarchical cluster analysis was performed to identify groups
of consumers with different preference patterns. This analysis
was performed on standardized liking scores, considering Euclidean distances and average linkage as agglomeration criterion.
2.3.2. Projective mapping
Data from the projective mapping task consists on the X and Y
coordinates of the drinks in the sheet, for each consumer. This data
was analyzed using Multiple Factor Analysis (MFA), as suggested
by Pags (2005).
In order to incorporate consumers descriptions in the samples
map obtained using MFA, the frequency table containing consumers descriptions was considered as a set of supplementary variables in MFA. This implies that this frequency table did not
participate in the construction of the MFA factors but consumers

583

descriptions were projected in the sample space by calculating

their correlation coefcients with the MFA factors (Pags, 2005).
Consumers ideal product was considered as supplementary individual in the analysis.
For the construction of the frequency table with consumers
descriptions a qualitative analysis was carried out. First, a search
for recurrent terms was performed in which terms with similar
meaning were grouped into different categories within each drink.
This classication was performed independently by two researchers considering personal interpretation of the meaning of the
words and word synonymy as determined by a Spanish dictionary.
After individually evaluating the data, a meeting of the researchers
was undertaken in order to check the agreement between their
classications. The nal categories and their names were determined by consensus between the two researchers, considering
their two independent classications in an open discussion. Categories mentioned by more than 10% of the consumers were considered. Frequencies in each category were determined by counting
the number of consumers that used those words to describe each
of the drinks.
2.3.3. Check-all-that-apply question
The frequency of mention of each term from the check-allthat-apply question was determined by counting the number of
consumers who used that term to describe each drink. In order
to evaluate if the check-all-that-apply question was able to detect
differences in consumers perception of the evaluated drinks,
Friedmans test was carried out for each of the terms, considering
sample and consumer as sources of variation.
Multiple Correspondence Analysis (MCA) was performed on the
frequency table containing responses to the CATA question, considering the ideal product as supplementary individual.
2.3.4. Intensity scales
Descriptive statistics (mean, standard deviation, minimum and
maximum score) were calculated for each of the descriptors evaluated using intensity scales. Then, an analysis of variance (ANOVA)
was carried out on data from intensity scales considering consumer and sample as xed sources of variation. Mean ratings and
honestly signicant differences were calculated using Tukeys test,
and were considered signicant when P 6 0.05.
A Principal Component Analysis (PCA) was performed on the
means of consumers data considering consumers ideal product
as supplementary individual.
2.3.5. Hierarchical cluster analysis
Hierarchical cluster analysis was performed to identify groups
of samples with different sensory characteristics. This analysis
was performed on samples coordinates in the rst and second
dimensions of the multivariate analysis used for the evaluated consumer proling methodologies (MFA Projective mapping-, MCA
CATA- or PCA intensity scales-), considering Euclidean distances
and Wards aggregation criterion.
2.3.6. Comparison of the methods
Hierarchical Multiple Factor Analysis (HMFA) (Le Dien & Pags,
2003; Perrin et al., 2008) was carried out to compare the product
positioning generated by the three consumer proling methodologies, for both the evaluated samples and consumers ideal product.
This analysis was performed on a table composed of seven rows,
corresponding to the seven evaluated drinks, and three groups of
columns, corresponding to the data from the three consumer proling methodologies considered. HMFA rst split the variables into
three groups in order to compare the three methods (Projective
mapping, CATA question and intensity scales). The second level
split the coordinates of the projective mapping task, equilibrating

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

consumers coordinates together. The advantage of this methodology is that it allows comparing the samples prole in the different
methodologies, in this case projective mapping, CATA and intensity
scales (Le Dien & Pags, 2003). Consumers ideal product was considered as supplementary individual and projected on the sensory
space dened by the evaluated products.

(a) 1

2.3.7. External preference mapping

External preference mapping was used to link consumer overall
liking scores and data from consumer proling methodologies. This
analysis was carried out as proposed by Danzart et al. (2004) using
SensoMineR (L, & Husson, 2008) in R language (R Development
Core Team, 2007). In order to determine the area of the map that
maximizes consumers liking, a quadratic response surface was
computed per consumer. The area of maximum liking for each consumer was determined as the area where predicted liking scores
were higher than the mean score of all the evaluated samples for
that consumer. Then, preference zones were delimited and nally
superimposed for the whole consumer sample. This analysis was
carried out considering samples coordinates in the HMFA performed on data from the three evaluated consumer proling methodologies. It is important to highlight that a quadratic model was
applied on data for only seven samples, which indicates that only
one degree of freedom was used to estimate the models
parameters.
All statistical analyses were performed using R language (R
Development Core Team, 2007). FactoMineR was use to perform
the multivariate analysis (MFA, MCA, PCA, and HMFA) (Husson,
Josse, L, & Mazet, 2007; L , Josse, & Husson, 2008).

D im 1 ( 1 7 . 7 %)

584

-1
-1

3.1. Consumers liking scores

As shown in Table 1, highly signicant differences were found
in consumers overall liking of the drinks (F = 29.3, p < 0.001, for
the product effect in the ANOVA). This suggests that consumers reacted differently to the sensory characteristics of the drinks. Overall liking scores ranged from 2.7 to 5.5, suggesting that samples
caused different affective reactions to consumers. Samples 3 and
5 were the most liked, whereas sample 4 was the most disliked. Except for sample 4, the overall liking of most of the samples was
positioned in the middle part of the scale, corresponding to products that were classied as indifferent or slightly disliked in the
9-point hedonic scale.
Internal preference map of consumers overall liking scores for
the seven evaluated samples is shown in Fig. 1. The rst two principal components explained 51.3% of the variability of the experimental data. As shown, most consumers were located at positive
values of the rst preference dimension, revealing that consumers

Table 1
Mean overall liking scores for the seven evaluated orangeavoured powdered drinks.
Sample

Overall liking*

1
2
3
4
5
6
7

4.2c
4.5b,c
5.2a,b
2.7d
5.5a
4.2c
4.3c

Different superscripts within a column indicate signicant differences according to Tukeys test (p 6 0.05).
*
Evaluated in a structured 9-point hedonic scale.

(b) 10

3
1

Dim 2 (17.7%)

3. Results

D im 1 ( 3 3 .6 %)

0
7
5
-5

6
-10
-15

-10

-5

10

15

Dim 1 (33.6%)
Fig. 1. Internal preference map of consumers liking scores of the seven evaluated
orange-avoured drink powders: (a) consumers representation and (b) samples
representation.

preferences were not highly heterogeneous. Most consumers liked

samples 3 and 5, in agreement with the fact that these samples
showed the highest mean overall liking scores. Furthermore, hierarchical cluster analysis did not show a clear segmentation in consumers responses (data not shown).
3.2. Projective mapping
Multiple factor analysis was carried out to get a consensus conguration of the samples. This analysis equilibrates the conguration given by all the consumers that participated in the study
together, providing a unique conguration.
Fig. 2 shows the representation of the drinks in the rst two
dimensions of the MFA of the projective mapping data. The rst
and second dimension of the MFA accounted for 28.6% and 23.8%
of the total variance of the experimental data, respectively. Cluster
analysis indicated that samples were separated into four groups in

585

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

2

(a)
5

Dim 2 (23.8%)

Ideal

1
0
-2

-1

-1

-2

Dim 1 (28.6%)
1

Yummy
Concentrated

(b)

Sweet
Very sweet

Dark colour

Artif icial f lavour

Dim 2 (23.8%)

Natural

Orange f lavour
Of f -f lavour
Acid
Not very sweet

0
0

-1

With pulp

Light colour

Disgusting

Bitter
Sof t

Not much f lavour

Diluted
-1

Dim 1 (28.6%)
Fig. 2. Representation of: (a) the seven orange-avoured powdered juice drinks and consumers ideal product and (b) the terms used to describe the samples, in the rst two
dimensions of the MFA plot of data from projective mapping.

the rst two dimensions of the MFA of the projective mapping task
(Fig. 2). One group composed of samples 2 and 7, another one of
samples 1, 3 and 5, whereas two other groups composed of only
one sample (samples 6 and 4) were located apart from the rest.
Respondents provided between one and six terms to describe
each of the evaluated drinks. Table 2 shows the terms used by consumers to describe the samples. Eighteen terms were mentioned
by more than 10% of the consumers. These terms were related to
hedonic and sensory characteristics of the drinks, particularly
appearance and avour attributes. The most frequently used terms
were not much orange avour, sweet and acid. This suggests that orange avour, sweetness and acidity might be among the most relevant sensory characteristics of orange-avoured powdered drinks
for consumers.
As shown in Fig. 2, sample 6 was described by consumers as disgusting and having orange pulp. Sample 4 was located at negative
values of the second dimension, being described as diluted as having not much orange avour. Meanwhile, samples 2 and 7 were perceived as bitter, acid and not very sweet, showing a clearly different
sensory prole. Finally, samples 1, 3 and 5 were located in the rst

Table 2
Terms used by consumers for describing the drinks before the projective mapping
task and number of mention for each of the samples.
Term

Not much orange avour

Sweet
Acid
Diluted
Articial avour
Yummy
Concentrated
Soft
Very sweet
Light colour
Disgusting
Not very sweet
Bitter
Off-avour
With pulp
Natural
Dark colour
Orange avour

Sample
1

8
11
4
3
6
6
2
3
1
1
0
1
1
1
0
0
3
0

4
1
17
4
2
2
1
4
0
5
1
4
2
0
0
0
0
2

0
10
2
1
2
5
7
1
8
1
2
0
0
2
0
1
1
3

21
0
0
13
2
0
0
5
1
0
1
1
2
0
0
0
0
0

1
7
5
0
5
6
6
3
5
0
0
3
1
1
0
1
2
0

8
1
1
6
4
2
3
1
2
1
6
1
2
3
7
2
0
0

1
8
8
4
3
3
4
2
0
5
3
3
4
5
0
2
0
0

586

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

scores. Therefore, considering the consensus position of the consumers ideal product in the projective mapping task enabled the
identication of zones of maximum liking in the sensory map, as
well as the sensory attributes responsible for consumers liking.

quadrant, being described as yummy, concentrated, very sweet and

having dark colour.
Apart from being able to identify differences in consumers perception of the sensory characteristics of the evaluated drinks, projective mapping also enabled the identication of consumers
drivers of liking and disliking for the evaluated products. The hedonic term yummy was highly correlated to the sensory terms sweet
(r = 0.87) and concentrated (r = 0.78), and negatively correlated to
the terms not much avour (r = 0.76) and diluted (r = 0.85). This
indicates that consumers want orange-avoured powdered drinks
to be sweet and have high avour intensity.
The consensus position of the ideal product in the rst two
dimensions of the MFA veries the importance of sweetness and avour intensity in consumers hedonic impression of this type of
product (cf. Fig. 2). Besides, the ideal product was located very close
to samples 3 and 5, which were those with the highest overall liking

3.3. Check-all-that-apply-question
Consumers used between 1 and 11 terms to describe the drinks
in the check-all-that-apply question. The most frequently used
term was articial avour, followed by light colour, not much orange
avour and dark colour. Meanwhile, the least used term was bitter,
followed by very sweet, rough and off-avour. Table 3 shows the
number of consumers who used each of the terms of the CATA
question to describe the seven evaluated drinks. Signicant differences between the samples were found in the frequency in which
the 19 terms of the CATA question were used to describe the

(a)
2

Dim 2 (26.1%)

7
Ideal

0
0

-1

-1

Dim 1 (55.0%)
1

(b)
Acid
Light colour

Dim 2 (26.1%)

Bitter

Orange flavour
Yummy
Natural Intense flavour

Off-flavour

Aftertaste

0
-1

10

Not very sweet

Diluted

Disgusting
Not much orange
flavour

Rough
Artificial flavour

Sweet

Very sweet

Soft

Dark colour

-1

Dim 1 (55.0%)
Fig. 3. Representation of: (a) the seven orange-avoured powdered juice drinks and consumers ideal product and (b) the terms used to describe the samples, in the rst two
dimensions of the MCA plot of data from the CATA question.

587

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

with the terms intense avour (r = 0.78), sweet (r = 0.87), orange avour (r = 0.82), suggesting that these sensory attributes might be
drivers of liking. These results are in agreement with those from
the projective mapping task.
As shown in Fig. 3, the ideal product was located on the positive
side of the rst dimension, close to samples 3 and 5. The location of
the ideal product suggests that consumers expect sweet and intense orange-avoured powdered drinks.
The ideal product was located further apart from all the evaluated samples than in the map from projective mapping (Fig. 2).
This suggests that according to data from the CATA question the
samples were less similar to consumers ideal when compared to
data from projective mapping. This result could be related to the
fact that counts for the ideal product clearly showed that consumers expect powdered drinks to be sweet, yummy with natural and
intense orange avour. However, none of the evaluated samples
showed a sensory prole close to this, since most of them were
perceived by consumers as articial.

samples (cf. Table 3), suggesting that this type of question was able
to detect differences in consumers perception of the drinks, particularly differences in their sensory and hedonic characteristics.
The rst and second dimension of the Multiple Correspondence
Analysis (MCA) calculated from CATA counts accounted for 55.0%
and 26.1% of the variance of the experimental data, respectively.
According to hierarchical cluster analysis, results from MCA performed on CATA questions sorted the drinks into four groups
(Fig. 3). Sample 4 was sorted apart from the rest of the samples,
being described using the terms disgusting, not very sweet, diluted
and not much orange avour. Meanwhile, samples 2 and 7 were described with the terms bitter, light colour, and acid. Samples 3 and 5
were grouped together, being described with the terms sweet, very
sweet, yummy, orange avour, aftertaste and intense avour. Finally,
samples 1 and 6 were located in an intermediate position.
The correlation of these hedonic terms with sensory attributes
from the CATA question enabled the identication of drivers of liking for this product. The term yummy was positively correlated

(a)

2
1

Dim 2 (33.5%)

7
6

Ideal

0
-2

-3

-1

3
1

-1

Dim 1 (46.9%)
1

(b)

0.5

Bitter

Acid

Off-flavour

Dim 2 (33.5 %)

Orange flavour

Sweetness

-0.5

Colour intensity

Artificial flavour

-1
-1

-0.5

0
Dim 1 (46.9 %)

0.5

Fig. 4. Representation of: (a) the seven orange-avoured powdered juice drinks and consumers ideal product and (b) the terms used to describe the samples, in the rst two
dimensions of the PCA plot of data from intensity scales.

588

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

Fig. 5. Superimposed representation of the samples in the HMFA plot performed on

data from the three methodologies. Each sample is represented using four points
corresponding to each of the three evaluated methodologies and the consensus
representation.

3.4. Intensity scales

Table 4 shows average scores and standard deviations for the
evaluated sensory attributes of the seven evaluated drinks. For

all the evaluated attributes standard deviations were high, ranging

from 1.2 to 3.4, which indicates large variability in consumers
evaluations. However, all the attributes were able to discriminate
between the evaluated drinks. According to ANOVA, highly significant differences (p < 0.001) between the samples were found for
the seven sensory attributes evaluated using intensity scales.
Principal component analysis (PCA) was carried on consumers
average scores for the seven evaluated sensory attributes in order
to get a consensus representation of the samples. The rst and second dimensions of the PCA accounted for 46.9% and 33.5% of the
total variance of the experimental data, respectively.
According to the results from the cluster analysis, the drinks
were separated into four groups in the PCA of data from intensity
questions (Fig. 4). Sample 4 was located apart from the rest of
the samples, being perceived as having the lowest sweetness and
orange avour intensity, and the highest articial avour intensity.
Meanwhile, samples 1, 3 and 5 were located at positive values of
the rst dimension and negative values of the second dimension,
corresponding to samples perceived as sweet, with intense colour
and high orange avour intensity. Samples 2 and 7 were located up
in the second dimension, corresponding to the most acid and bitter
samples, whereas sample 6 was mainly perceived as bitter and
having high off-avour intensity.
The abovementioned results suggest that even though consumers evaluations using intensity scales were highly heterogeneous,
considering average scores provided useful information regarding
consumers perception of the sensory characteristics of the drinks.
The different panels of consumers described their ideals in a
similar way. As shown in Table 4, using intensity scales, consumers
described their ideal product as having a relatively high sweetness,
high orange avour and color intensity, intermediate acidity and
low bitterness. Meanwhile, their ideal product was described as

Fig. 6. Representation of the attributes evaluated using a CATA question (in gray) and intensity scales (in black) in the HMFA plot performed on data from the three evaluated
methodologies.

589

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

not having articial or off-avour, as expected. It is interesting to

notice that standard deviations of consumers ideal product were
smaller than those from the drinks evaluations, suggesting homogeneity in consumers ideal product.
Furthermore, none of the evaluated drinks had a sensory prole
close to consumers ideal product (cf. Table 4). This could explain
the fact that the ideal product was clearly separated from all the
evaluated samples in the PCA of data from intensity scales
(Fig. 4). This result suggests that according to this methodology
the sensory prole of none of the samples could be regarded as
close to that of consumers ideal.
3.5. Comparison between the methods based on Hierarchical Multiple
Factor analysis
HMFA was used to compare results from the three consumer
proling methodologies for the evaluated samples and consumers
ideal product in the three cases. The superimposed representation
of the samples according to each of the evaluated methods allows
the evaluation of their agreement for each of the evaluated samples. As shown in Fig. 5, points corresponding to the three methods
were close for the evaluated samples, suggesting that sample congurations were very similar. Distances between samples in the
sensory maps were similar regardless the methodology considered.
On the other hand, the differences in the positioning of the ideal
product were larger between the methods. According to projective
mapping the ideal product was located close to samples 3 and 5,
whereas when Intensity scales or the CATA question were considered, it was located far from the sensory space dened by the evaluated samples.
This suggests that differences between the methodologies were
larger for the ideal product than for the evaluated samples. In this
case, the data from the methodologies based on pre-selected attributes were highly correlated to each other and provided an ideal
product that was located outside the preselected sensory space,
whereas projective mapping, which is based on consumers holistic
perception of the products, provided an ideal within the sensory
space dened by the evaluated samples. These differences could
be also observed through the RV coefcients between the congurations of the different methodologies. The RV coefcient between
congurations from CATA question and intensity scales was 0.93,
whereas the RV coefcients of these congurations with that from
projective mapping were 0.79 and 0.82, respectively. The relatively
high values of these coefcients highlight the similarities between
the three congurations.
Apart from comparing samples conguration, HMFA also enabled to visualize the correlation between terms evaluated by
the CATA question and intensity scales. As shown in Fig. 6 all the
attributes evaluated by consumers using intensity scales were
highly correlated to similar terms evaluated using the CATA question, suggesting an agreement between the perception of both consumer groups.

Fig. 7. External preference mapping performed on samples coordinates in the

HMFA plot performed on data from the three evaluated methodologies (projective
mapping, CATA and intensity scales).

consensus position of the ideal sample in the projective mapping

task (Fig. 2).
3.7. General discussion
Projective mapping, CATA questions and intensity scales provided very similar product spaces, yielding similar information
regarding the sensory characteristics of the evaluated samples.
This suggests that product congurations were stable, although
different consumer panels and methodologies were considered.
Overall liking scores, terms with hedonic connotation and the
position of the ideal products for all the consumer groups indicated
that samples 3 and 5 were the most preferred. Hence, the preference patterns of the different consumer panels were alike.

Table 3
Results of the check-all-that-apply question. Number of consumers who used each
term to describe the seven evaluated orange-avoured powdered drink samples.
Term

Very sweet***
Sweet***
Not very sweet***
Bitter***
Not much orange avour***
Off-avour***
Dark colour***
Light colour***
Diluted***
Yummy***
Disgusting***
Orange avour***
Articial avour***
Natural***
Acid***
Aftertaste**
Rough***
Soft***
Intense avour***

3.6. External preference mapping

Fig. 7 shows the results of the external preference mapping carried out on consumers individual overall liking scores, considering
samples coordinates in the rst two dimensions of the HMFA of
data from projective mapping, CATA counts, and intensity scales.
According to the external preference map, the area in which samples 3 and 5 were located was close to that of maximum liking. This
result was in agreement with the fact that these samples showed
the highest overall liking scores.
Besides, it is interesting to notice that results from external
preference mapping were very similar to those derived from the

**
***

Sample
1

Ideal

15
17
9
2
22
8
15
17
16
8
12
12
36
5
9
7
4
13
7

6
12
17
10
17
11
3
43
8
8
11
13
29
5
28
4
7
7
12

22
22
4
1
19
6
24
10
2
13
6
14
33
5
4
7
10
15
19

0
6
36
4
39
7
3
26
46
0
32
3
30
1
1
1
5
20
0

13
27
6
0
11
7
38
4
4
26
6
17
32
8
9
8
13
13
15

3
8
22
8
23
20
29
2
13
8
29
8
38
6
11
9
16
9
7

6
15
16
6
14
11
2
32
13
15
13
19
23
5
22
10
12
10
8

0
44
8
0
0
0
29
12
0
49
0
50
0
50
12
1
2
32
13

Indicates signicant differences at p 6 0.01 according to Friedmans test.

Indicates signicant differences at p 6 0.001.

590

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

Table 4
Average scores and standard deviations (between brackets) for the sensory attributes of the seven orange-avoured powdered drinks evaluated by consumers using intensity
scales.
Attribute*

Sweetness
Orange avour
Colour
Articial avour
Off-avour
Acid
Bitter

Sample
1

Ideal

6.0c (2.5)
3.5b (1.9)
5.4b (1.9)
7.0a,b (1.7)
3.8a,b (2.9)
3.4b (2.3)
2.4a,b (2.2)

4.1b (2.4)
4.7b (2.5)
2.9a (2.1)
6.4a,b (2.7)
4.6a,b,c (3.2)
6.4c (2.6)
3.3a,b (2.6)

6.7c (2.3)
4.3b (2.5)
6.0b,c (1.6)
6.8a,b (2.0)
4.6a,b,c (3.1)
3.5b (2.4)
2.3a,b (2.0)

1.9a (2.2)
1.1a (1.2)
3.1a (1.8)
7.6b (2.5)
4.5a,b,c (3.4)
1.0a (1.2)
2.5a,b (2.9)

6.2c (2.1)
4.8b (2.8)
7.6d (1.5)
6.2a (2.3)
2.9a (2.8)
3.2b (2.3)
2.1a (2.2)

3.9b (2.4)
3.6b (2.7)
6.6c,d (1.4)
7.2a,b (2.3)
5.6c (3.2)
3.9b (2.9)
3.5b (2.8)

4.3b (2.4)
4.3b (2.5)
3.5a (1.9)
6.4a,b (2.6)
4.0a,b (3.4)
5.6c (2.9)
3.1a,b (2.6)

5.8
8.6
6.3
0.0
0.0
4.6
1.6

(1.6)
(1.0)
(1.4)
(0.0)
(0.0)
(1.8)
(1.4)

Different letters within a row imply signicant difference according to Tukeys test (p < 0.05).
Evaluated in 10 cm unstructured intensity scales.

Therefore, the different approaches yielded similar information

regarding the sensory characteristics of consumers ideal orangeavoured powdered drink, providing similar recommendations
for product development. Consumers wanted sweeter beverages,
with more concentrated/intense orange avour. However, the
evaluated methodologies differed in the coordinates of consumers
optimum product within the sensory space dened by the sensory
characteristics of the evaluated samples.
CATA questions and intensity scales showed to be very sensitive
to the inclusion of attributes that are not wanted by consumers in
orange-avoured powdered drinks, such as articial avour. This
could be related to the fact that these attributes were strongly negatively correlated to overall liking and therefore have a great inuence in the samples position in the sensory space. Added to this,
the number of attributes utilized was lower in these two methods,
particularly for the intensity scales, compared to the descriptions
utilized in the projective mapping exercise. Thus, the weight of
the negative attributes in the resulting biplots might have been
higher. All the samples were described by consumers as having
articial avour (cf. Tables 3 and 4). However, consumers desire
a product without this avour, which is not feasible for this type
of product from a technological point of view. The ideal product
in the PCA of intensity scales data was located far from the
evaluated samples, in agreement with the fact that the ideal sensory prole was quite different than that of the commercial
samples.
On the other hand, projective mapping seemed more adequate
to identify consumers ideal products within a set of samples with
specic sensory characteristics. This could be explained by the fact
that when consumers are asked to include their ideal product in
their A3 sheet they necessarily have to place it within the space dened by the real samples, as they already occupied the map with
the evaluated samples. In the particular example of the orange
powdered juices, consumers might like, for instance, something
highly natural and with fruit pulp, but that does not belong to
the category in study, and the researcher might want to restrict
their options to choosing the best liked product within the boundaries of the category. In the case of CATA and intensity scores, the
ideal product could potentially be placed outside the map, as consumers could theoretically check/measure whatever they want for
the ideal. In the case the sample set was not wide enough, one of
these two methods might be better than Napping. This, however,
could provide an interesting application for projective mapping,
since it could enable the identication of liked products within
the sensory space dened by commercial products within a category. This can be very helpful in the case of technological or economic restrictions in the development. Anyhow, it has to be
pointed out that for this approach to be useful in product development guidance, the sensory space to be considered should be wide
enough, not to miss opportunities within the range of the category.

The challenge, then, would be the selection of the samples. Furthermore, asking consumers to rate or check multiple attributes
on an ideal product might be a hard task, which might induce an
attempt to be rational, while consumer decisions/choices tend to
invoke more emotional responses. This can be another point in favour of Napping, as it might be a more intuitive, easier task for
consumers based on their holistic perception of the samples, simplifying the nding of the ideal point in direct comparison with the
real products, rather than thinking about individual attributes. The
consensus position of the ideal sample in the MFA of the projective
mapping task was very close to the zone of maximum liking in
external preference mapping and indicated that the best products
were those with the highest overall liking scores.
It is important to take into account that in the present study,
consumers preference patterns were not highly heterogeneous,
which allowed working with the consensus ideal product. In future
work it would be necessary to study the variability in consumers
ideal products and to investigate the applicability of this approach
to products with a highly segmented market.

4. Conclusions
Projective mapping, check-all-that-apply questions and intensity scales were able to detect differences in consumers perception
of the sensory characteristics of the evaluated orange-avoured
powdered drinks. The three methodologies provided similar product spaces, suggesting that the three consumer panels were consistent in terms of their description of the products.
Good correlations between the methods in terms of the sensory
maps were obtained, the description of the samples and the identication of drivers of liking were found. Thus, choice between the
three evaluated methodologies for consumer proling might depend on practical considerations such as ease of use and consumers understanding.
However, differences between the methodologies were identied. Projective mapping/Napping enabled the identication of
an ideal product within the range of sensory characteristics of the
evaluated samples, providing very similar results to those of external preference mapping. Meanwhile, considering data from checkall-that-apply questions and intensity scales the ideal product was
located outside the sensory space dened by the evaluated samples.
This could be explained by the fact that several terms with strong
hedonic connotation were considered and that none of the samples
were close to the sensory prole of the ideal product, particularly in
terms of their articial avour. It is interesting to notice that orange-avoured powdered drinks would probably never be perceived as having natural or non-articial avour. Therefore, for
this category of products, CATA and intensity scales might not be
appropriate in this context since consumers indicate their ideal,

G. Ares et al. / Food Quality and Preference 22 (2011) 581591

which is not possible to reach from a technological point of view.

Further studies should be carried out to evaluate the capability of
CATA and intensity scales, as compared to Napping to identify consumers ideal product in other types of products.
Despite the fact that the proposed approach seems promising
and interesting, it needs further exploration. The rst issue to be
addressed is the application of this approach to other products,
particularly products with known differences in their sensory characteristics. In further research the methodology should be applied
to more complex products which cover a wide range of sensory
characteristics. Moreover, another relevant issue to investigate is
the sensitivity of these methodologies to products with a highly
segmented market. It would be interesting to perform studies with
a large number of consumers, and to apply segmentation techniques to identify clusters with different preference patterns based
on their responses when asked about their ideal product. Further
research it is also needed to investigate the variability in consumers ideal products.
Acknowledgements
We are grateful to the anonymous reviewers, whose comments
and suggestions have contributed to improve the quality of our
manuscript.
References
Adams, J., Williams, A., Lancaster, B., & Foley, M. (2007). Advantages and uses of
check-all-that-apply response compared to traditional scaling of attributes for
salty snacks. In: 7th Pangborn Sensory Science Symposium. Minneapolis, USA, 12
16 August, 2007.
Arditti, S. (1997). Preference mapping: A case study. Food Quality and Preference, 8,
232327.
Ares, G., Deliza, R., Barreiro, C., Gimnez, A., & Gmbaro, A. (2010). Comparison of
two sensory proling techniques based on consumer perception. Food Quality
and Preference, 21, 417426.
Carroll, J. D. (1972). Individual differences and multidimensional scaling. In R. N.
Shepard, A. K. Rommey, & S. B. Nerlove (Eds.). Multidimensional Scaling: Theory
and Applications in the Behavioral Sciences (Vol. 1, pp. 105155). New York:
Seminar Press.
Danzart, M. (1998). Quadratic model in preference mapping. In: Proceedings of the
4th Sensometrics Meeting. Copenhagen, Denmark, 68 August, 1998.
Danzart M., Sieffermann J.M., & Delarue J. (2004). New developments in preference
mapping techniques: nding out a consumer optimal product, its sensory
prole and the key sensory attributes. In: Proceedings of the 7th Sensometrics
Meeting. Davis, USA, 2730 July, 2004.
Ennis, D. M., Palen, J., & Mullen, K. (1988). A multidimensional stochastic theory of
similarity. Journal of Mathematical Psychology, 32, 449465.
Ennis, D. M., & Rousseau, B. (2004). Motivations for product consumption:
Application of a probabilistic model to adolescent smoking. Journal of Sensory
Studies, 19, 107117.
Epkong, A., Ngarmsak, T., & Winger, R. J. (2006). Comparing sensory methods for the
optimisation of mango gel snacks. Food Quality and Preference, 17, 622628.
Faye, P., Brmaud, D., Teillet, E., Courcoux, P., Giboreau, A., & Nicod, H. (2006). An
alternative to external preference mapping based on consumer perceptive
mapping. Food Quality and Preference, 17, 604614.
Guinard, J. X., Uotani, B., & Schlich, P. (2001). Internal and external mapping of
preferences of commercial lager beers: comparison of hedonic ratings by
consumers blind versus with knowledge of brand and price. Food Quality and
Preference, 12, 243255.

591

Husson, F., Josse, J., L, S., & Mazet, J. (2007). FactoMineR: factor analysis and data
mining with R. R package version 1.04, URL http://cran.R-project.org/
package=FactoMineR.
Husson, F., Le Dien, S., & Pags, J. (2001). Which value can be granted to sensory
proles given by consumers? Methodology and results. Food Quality and
Preference, 12, 291296.
ISO. (1988). Sensory analysis: General guidance for the design of test rooms, ISO
8589. Geneva: International Organization for Standardization.
Lagrange, V., & Norback, J. P. (1987). Product optimization and the acceptor set size.
Journal of Sensory Studies, 2, 119136.
Lawlor, J. B., & Delahunty, C. M. (2000). The sensory prole and consumer
preference for ten specialty cheeses. International Journal of Dairy Technology,
53, 2836.
Le Dien, S., & Pags, J. (2003). Hierarchical multiple factor analysis: Application to
the comparison of sensory proles. Food Quality and Preference, 14, 397403.
L, S., & Husson, F. (2008). SensoMineR: a package for sensory data analysis. Journal
of Sensory Studies, 23, 1425.
L, S., Josse, J., & Husson, F. (2008). FactoMineR: An R package for multivariate
analysis. Journal of Statistical Software, 25(1).
McEwan, J. A. (1996). Preference mapping for product optimization. In T. Naes & E.
Risvik (Eds.), Multivariate analysis of data in sensory science (pp. 7180).
Amsterdam: Elsevier.
Meullenet, J.F., Lee, Y., & Dooley, L. (2008a). The application of check-all-that-apply
consumer proling to preference mapping of vanilla ice cream and its
comparison to classical external preference mapping. In: Proceedings of the
9th Sensometric Meeting. St. Chaterines, Canada, 2023 July, 2008.
Meullenet, J. F., Lovely, C., Threlfall, R., Morris, J. R., & Striegler, R. K. (2008b). An
ideal point density plot method for determining an optimal sensory prole for
Muscadine grape juice. Food Quality and Preference, 19, 210219.
Moskowitz, H. R. (1996). Experts versus consumers: A comparison. Journal of
Sensory Studies, 11, 1937.
Pags, J. (2005). Collection and analysis of perceived product inter-distances using
multiple factor analysis: Application to the study of 10 white wines from the
Loire valley. Food Quality and Preference, 16, 642649.
Perrin, L., Symoneaux, R., Matre, I., Asselin, C., Jourjon, F., & Pags, J. (2008).
Comparison of three sensory methods for use with the Napping_procedure:
Case of 10 wines from Loire valley. Food Quality and Preference, 19, 111.
R Development Core Team. (2007). R: A language and environment for statistical
computing. ISBN 3-900051-07-0. Vienna: R Foundation for Statistical
Computing.
Risvik, E., McEwan, J. A., Colwill, J. S., Rogers, R., & Lyon, D. H. (1994). Projective
mapping: A tool for sensory analysis and consumer research. Food Quality and
Preference, 5, 263269.
Risvik, E., McEwan, J. A., & Rodbotten, M. (1997). Evaluation of sensory proling and
projective mapping data. Food Quality and Preference, 8, 6371.
Rousseau, B., & Ennis, D. M. (2008). An application of Landscape Segmentation
Analysis_to blind and branded data. IFPress, 11, 23.
Rudolph, M. J. (1995). The food development process. British Food Journal, 97, 311.
Schlich, P. (1995). Preference mapping: Relating consumer preferences to sensory or
instrumental measurements. In P. Etivant & P. Schreier (Eds.), Bioavour95.
Analysis/Precursor Studies/Biotechnology (pp. 231245). Versailles, France: INRA
Editions.
Stewart-Knox, B. J., & Mitchell, P. (2003). What separates the winners from the
losers in new food product development? Trends in Food Science & Technology,
14, 5864.
ten Kleij, F., & Musters, P. A. D. (2003). Text analysis of open-ended survey
responses: A complementary method to preference mapping. Food Quality and
Preference, 14, 4352.
Thompson, J. L., Drake, M. A., Lopetcharat, K., & Yates, M. D. (2004). Preference
mapping of commercial chocolate milks. Journal of Food Science, 69, 406413.
van Trijp, H. C. M., Punter, P. H., Mickartz, F., & Kruithof, L. (2007). The quest for the
ideal product: Comparing different methods and approaches. Food Quality and
Preference, 18, 729740.
Varela, P., Ares, G., Gimnez, A., & Gmbaro, A. (2010). Inuence of brand
information on consumers expectations and liking of powdered drinks in
central location tests. Food Quality and Preference, 21, 873880.
Worch, T., L, S., & Punter, P. (2010). How reliable are the consumers? Comparison
of sensory proles from consumers and experts. Food Quality and Preference, 21,
309318.