Documente Academic
Documente Profesional
Documente Cultură
a r t i c l e
i n f o
Article history:
Received 24 November 2012
Received in revised form 17 January 2013
Accepted 29 January 2013
Available online 8 February 2013
Keywords:
Polarized sensory positioning
Sensory characterization
Consumer proling
Powdered drinks
Make-up foundations
a b s t r a c t
Polarized sensory positioning (PSP) is a novel methodology for sensory characterization which is based on
the comparison among samples and a set of reference products, called poles. Its main advantage over
other holistic methodologies is that results from different sessions can be compared. This is particularly
useful when working with samples with persistent avours, when evaluating a large number of samples
over a long period of time, or during quality control. Although the key step of this methodology is the
selection of the products to be considered as poles, no studies have been published addressing this issue.
In this context, the aim of the present work was to contribute to bridging this gap by studying the inuence of the poles on results from sensory characterization of two widely different products (make-up
foundations and orange-avoured powdered drinks) using Polarized Sensory Positioning with consumers. Results from both studies provided preliminary evidence about the stability of sensory spaces
obtained with different set of poles (RV coefcients higher than 0.908). However, differences in conclusions regarding similarities and differences among samples were identied between the sets of poles in
both studies, suggesting the need to carefully select stable poles for PSP tasks.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Sensory descriptive analysis is a very powerful tool for the
industry since it provides valuable information about the sensory
characteristics of their products (Lawless & Heymann, 2010). Using
this methodology, qualitative and quantitative information about
human perception when using a product can be obtained, which
allows informed decisions when developing new products or reformulating existing ones with the purpose of having better consumer
acceptance or reaching a wider market (Stone & Sidel, 2004).
Sensory descriptive analysis methodologies can be basically divided into classic and novel techniques (Varela & Ares, 2012).
Within the rst group, Quantitative Descriptive Analysis (QDA) is
the most popular methodology (Murray, Delahunty, & Baxter,
2001). It involves the selection, training and maintenance of a panel that usually requires from 8 to 20 assessors (Lawless & Heymann, 2010). This panel needs very specic training in each
product in order to provide very detailed, reliable, robust, consistent, reproducible results and stable in time. Furthermore, panel
maintenance is required and this represents high inversions for
companies in terms of time and resources since training can be relatively long as it must be detailed and extensive (Varela & Ares,
2012). The length of training usually ranges between 10 and
120 h, depending on the sensory complexity of the product
Corresponding author. Tel.: +598 29248003; fax: +598 292419906.
E-mail address: gares@fq.edu.uy (G. Ares).
0950-3293/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.foodqual.2013.01.009
(Lawless & Heymann, 2010; Murray et al., 2001). Lastly, the panel
is trained in one specic product. In the context of todays highly
competitive markets the above-mentioned requirements might
hinder the use of this methodology since companies demand quick
responses as well as reliable results. The time and economic resources needed for these methods for small companies, along with
the large variety of products larger companies might have, require
different techniques that allow them to get similar results without
the drawbacks associated with classic descriptive analysis
(Moussaoui & Varela, 2010).
In this context novel techniques are particularly useful
(Valentin, Chollet, Lelivre, & Abdi, 2012; Varela & Ares, 2012). They
all have the advantage of being less time and money consuming, as
both semi trained assessors and nave consumers can be used (Ares,
Deliza, Barreiro, Gimnez, & Gmbaro, 2010c; Cadoret, L, & Pags,
2009; Cartier et al., 2006; Chollet, Lelivre, Abdi, & Valentin, 2011;
Nestrud & Lawless, 2008; Perrin & Pags, 2009). They are also highly
reliable and simple, which makes them very useful with consumers
because information can be gathered in their own terms (Ares,
Gimnez, Barreiro, & Gmbaro, 2010b; ten Kleij & Musters, 2003;
Veinand, Godefroy, Adam, & Delarue, 2011). Identifying consumers
vocabulary could provide very valuable information for new product developers in order to align as much as possible new products
with consumers preferences, and to design marketing or communication campaigns (Varela & Ares, 2012).
There are many novel techniques for product characterization,
including sorting (Lawless, Sheng, & Knoops, 1995; Schiffman,
26
Table 1
Description of the make-up foundation samples
used in study 1.
Sample
Texture characteristics
A
B
C
D
E
F
G
H
Fluid
Mousse
Cream
Cream
Fluid
Mousse
Fluid
Fluid
27
Table 2
Description of the eight orange-avoured powdered drinks considered in study 2.
Sample
Market
segment
Main characteristics
Premium
B
C
D
E
F
G
H
Economy
Economy
Premium
Premium
Economy
Medium
Premium
Table 3
Example of the data matrix used for analyzing data from Polarized Sensory
Positioning using multiple factor analysis. Each couple of columns RA, RB, and RC
represent the degree of difference between a sample and each of the poles (RA, RB and
RC respectively) for each assessor.
Sample
1
2
...
X
Assessor 1
Assessor 2
Assessor n
RA
RB
RC
RA
RB
RC
...
RA
RB
RC
1.4
0.2
...
3.2
8.3
9.8
...
4.8
0.9
7.8
...
6.4
1.0
8.9
...
5.8
7.6
5.6
...
9.9
8.5
2.4
...
1.4
...
...
...
...
1.3
3.4
...
8.9
7.8
6.5
...
7.5
4.4
6.7
...
1.4
28
Fig. 1. Representation of eight make-up foundations in the rst and second dimensions of the multiple factor analysis performed on data from Polarized Sensory Positioning
performed using different sets of poles: (a) Set 1 (samples D, E and F were used as poles), and (b) Set 2 (samples A, B and C were used as poles). Dotted ellipses show clusters
identied using Hierarchical cluster analysis.
and D). Except for samples C and D (samples with a cream-like texture), the discriminative capacity of the methodology within each
group of samples was low since their condence intervals
overlapped.
In the case of the second set of poles, the rst two dimensions of
the MFA explained 59.7% of the variance of the experimental data
(Fig. 1(b)), similar to the explained variance when the rst set of
poles was considered. As shown, samples were sorted into four
main groups. Samples B and F (corresponding to make-up foundations with a mousse-like texture), were located at negative values
of the rst dimension and positive values of the second dimension,
being clearly different from the rest (Fig. 1(b)). Samples A, E, G and
H (with a liquid texture) were located at positive values of the rst
dimension, being close to sample D, which had a cream-like texture. Finally, sample C was sorted apart from the rest of the samples, being more similar to samples F and B.
Pole A
(mousse-like texture) %
Pole B
(cream-like texture) %
Pole C
(liquid texture) %
Set 1
Set 2
77
73
67
67
83
80
29
Fig. 2. Representation of nine orange-avoured powdered drink samples in the rst and second dimensions of the multiple factor analysis performed on data from Polarized
Sensory Positioning performed using two sets of poles: (a) Set 1 (samples C, D and E were used as poles), and (b) Set 2 (samples F, G and H were used as poles). Dotted ellipses
show clusters identied using Hierarchical cluster analysis.
30
Table 5
Percentage of consumers who correctly identied each one of the three poles when
they were presented as blinded samples when the two sets of poles were considered
in the polarized sensory positioning task of study 2.
Set of
poles
Pole A
(economy drink) %
Pole B
(without sugar) %
Pole C
(with sugar) %
Set 1
Set 2
76
70
70
78
56
59
Consumer group
Set 1
3 poles (n = 11)
2 poles (n = 13)
1 or 0 poles (n = 6)
Whole consumer sample (n = 30)
3 poles (n = 13)
2 poles (n = 10)
1 or 0 poles (n = 7)
Whole consumer sample (n = 30)
Set 2
signicant at p 6 0.05.
signicant at p 6 0.01.
ns
not signicant (p > 0.05).
**
1
0.87
0.43ns
0.94
1
0.88
0.53
0.96
1 or 0
1
0.54
0.93
0.64
1
0.57
0.93
1
0.61
Consumer group
Set 1
3 poles (n = 16)
2 poles (n = 18)
1 or 0 poles (n = 12)
Whole consumer sample (n = 46)
3 poles (n = 18)
2 poles (n = 17)
1 or 0 poles (n = 11)
Whole consumer sample (n = 46)
Set 2
*
**
1
0.81
0.64
0.96
1
0.87
0.79
0.95
1 or 0
1
0.50
0.90
0.68
1
0.64
0.87
1
0.83
signicant at p 6 0.05.
signicant at p 6 0.01.
signicant at p 6 0.001.
***
consumers. It is important to highlight that in the original application PSP was used with 15 trained assessors for sensory characterization of water.
In both studies the majority of consumers correctly identied
the poles when they were presented as blind samples. The average
percentage of correct identications was 75% for the make-up
foundation samples and 68% for the powdered drink samples.
The ability to correctly identify the three poles was not signicantly affected by the different sets of poles.
A possible alternative for obtaining better discrimination in PSP
would be to analyze results considering only data from those
assessors who correctly identied two or three poles. However, results from the present work showed that removing assessors with
poor performance did not markedly modify sample congurations
(cf. Tables 6 and 7). This suggests that sample congurations were
mainly determined by data from those consumers who showed a
good performance (i.e. those who correctly identied two or three
of the poles when presented as blind samples), so removing consumers with poor performance seems not to be relevant. However,
considering the similarity of the congurations provided by consumers who identied 2 and 3 poles (Table 2), identifying at least
2 of the poles would be a good criterion for selecting the best
assessors for sensory characterizations from PSP.
The key step of PSP is the selection of the poles. Therefore, research is needed to understand how the nature and number of
products selected as poles affect results. The present work contributes to bridging this gap by comparing results from sensory characterization of two widely different products (make-up
foundations and orange-avoured powdered drinks) using PSP
with two different sets of poles. In both studies the sensory spaces
were similar regardless the set of poles considered (RV coefcient
higher than 0.908), suggesting that conclusions regarding the main
similarities and differences among samples were not affected by
the selection of poles. Besides, changing the poles did not largely
affect agreement among consumers, as denoted by the size of the
condence ellipses (cf. Figs. 1 and 2).
However, in both studies some differences were also identied.
In the make-up foundation study results from the PSP task with the
rst set of poles better reected the expected differences among
samples than results obtained with the second set. With the rst
set, samples C and D were regarded as similar (Fig. 1), as expected
by their similar cream-like texture (cf. Table 1). Meanwhile, in the
powdered drink study, results obtained with the second set of
poles were more in agreement with what was expected according
to product formulation and market positioning data (cf. Table 2)
than results obtained with the rst set of poles. Sample E, which
was a premium drink formulated with sugar was expected to be
31
32
Cartier, R., Rytz, A., Lecomte, A., Poblete, E., Krystlik, J., Belin, E., et al. (2006). Sorting
procedure as an alternative to quantitative descriptive analysis to obtain a
product sensory map. Food Quality and Preference, 17, 562571.
Chrea, C., Teillet, E., & Navarro, S. (2011). Application of the polarized sensory
positioning in the cosmetic area. 9th Pangborn Sensory Science Symposium.
Toronto, Canada, 48 September, 2011.
Dairou, V., & Sieffermann, J.-M. (2002). A comparison of 14 jams characterized by
conventional prole and a quick original method, ash prole. Journal of Food
Science, 67, 826834.
ISO (2007). Sensory analysis: General guidance for the design of test rooms, ISO 8589.
Geneve: International Organization for Standardization.
Josse, J., Husson, F., & Pags, J. (2007). Testing the signicance of the RV coefcient.
In IASC 07, August 30th - September 1st, 2007, Aveiro, Portugal.
Lawless, H. T., & Heymann, H. (2010). Sensory evaluation of food. Principles and
practices (2nd edition). New York: Springer.
Lawless, H. T., Sheng, N., & Knoops, S. S. C. P. (1995). Multidimensional scaling of
sorting data applied to cheese perception. Food Quality and Preference, 6, 9198.
L, S., Josse, J., & Husson, F. (2008). FactoMineR: An R package for multivariate
analysis. Journal of Statistical Software, 25(1), 118.
Moussaoui, K. A., & Varela, P. (2010). Exploring consumer product proling
techniques and their linkage to a quantitative descriptive analysis. Food
Quality and Preference, 21, 10881099.
Murray, J. M., Delahunty, C. M., & Baxter, I. A. (2001). Descriptive sensory analysis:
Past, present and future. Food Research International, 34, 461471.
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., & Pags, J. (2009). Construction of a product space from the ultra-ash
proling method: Application to 10 red wines from the Loire Valley. Journal of
Sensory Studies, 24, 373395.