presentation parfum.pdf - Senstools
presentation parfum.pdf - Senstools
presentation parfum.pdf - Senstools
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
How reliable are the consumers?<br />
Comparison of sensory profiles<br />
from consumers and experts<br />
(1) OPP Product Research<br />
(2) AAgroCampus C OOuest t<br />
Project 8013<br />
July 2008<br />
WORCH Thierry (1)<br />
LÊ Sébastien (2)<br />
LÊ Sébastien ( )<br />
PUNTER Pieter (1)<br />
mailto: thierry@opp.nl<br />
Senior project manager Pieter Punter<br />
Project manager Thierry Worch
introduction<br />
• in the sensory theory:<br />
• experts p ppanels are used for the pproducts’ description p<br />
• consumers should only be used for the hedonic task<br />
• they lack two essentials qualities for profiling (consensus<br />
and reproducibility)<br />
• there are strong halo effects (Earthy, MacFie & Hedderley,<br />
1997)<br />
• in the sensory practice:<br />
• consumers are sometimes used for both tasks<br />
8013<br />
• it has been proven that consumers’ consumers description show the<br />
required qualities (consensus and reproducibility) (Husson, Le<br />
Dien, Pagès, 2001)<br />
3
problematic<br />
How reliable are the consumers?<br />
8013<br />
4
<strong>presentation</strong> of the studies<br />
• products:<br />
• twelve luxurious women perfumes<br />
(Gazano, Ballay, Eladan & Sieffermann, 2005)<br />
8013<br />
AAngel l L’I L’Instant t t<br />
(Eau de Parfum) (Eau de Parfum)<br />
Cinéma J’Adore<br />
(Eau de Parfum) (Eau de Toilette)<br />
Pleasures<br />
J’Adore<br />
(Eau de Parfum) (Eau de Parfum)<br />
Aromatics Elixir<br />
(Eau de Parfum)<br />
Lolita Lempicka<br />
(Eau de Parfum)<br />
Pure Poison<br />
(Eau de Parfum)<br />
Shalimar<br />
(Eau de Toilette)<br />
Chanel N°5 N 5<br />
Coco Mademoiselle<br />
(Eau de Parfum) (Eau de Parfum)<br />
5
<strong>presentation</strong> of the studies<br />
• expert panel (Agrocampus Rennes)<br />
8013<br />
• twelve persons (11 students and 1 teacher) from the Chantal Le<br />
Cozic school (esthetics and cosmetic school)<br />
• focus group per group of six, with two animators<br />
• generation of a list of twelve attributes<br />
• “Vanille”, “Notes Florales”, “Agrume”, “Boisé”, “Vert”, “Epicé”, “Capiteux”,<br />
“Fruité”, “Fraîcheur Marine”, “Gourmand”, “Oriental”, “Enveloppant”<br />
• training session for the most difficult ones<br />
• the twelve products were tested two times in two one-hour<br />
sessions<br />
6
<strong>presentation</strong> of the studies<br />
• consumer panel (OP&P Product Research, Utrecht)<br />
8013<br />
• 103 naïve Dutch consumers living in the Utrecht area<br />
• the same twelve perfumes were rated on 21 attributes<br />
• “odour intensity”, “freshness”, “jasmine”, “rose”, “camomile”, “fresh lemon”,<br />
“vanilla”, “mandarin/orange”, “anis”, “sweet fruit/melon”, “honey”, “caramel”,<br />
“spicy” spicy , “woody” woody , “leather” leather , “nutty/almond” nutty/almond , “musk” musk , “animal” animal , “earthy” earthy , “incense” incense ,<br />
“green”<br />
• two products (Shalimar and Pure Poison) were duplicated<br />
• the fourteen (12+2) products were tasted in two one-hour sessions (seven<br />
products in each session, <strong>presentation</strong> order was balanced)<br />
7
<strong>presentation</strong> route map<br />
• the consumer and expert data are compared in three different ways<br />
8013<br />
1.Univariate analysis<br />
• analyses of variance<br />
• correlations<br />
22.Multivariate Multivariate comparison<br />
• construction of the two products’ spaces (PCA)<br />
• comparison of the products products’ spaces through GPA and MFA<br />
3.Confidence ellipses<br />
• graphical confidence intervals around the products averaged<br />
over the two panels<br />
• graphical confidence intervals around the products defined<br />
by the different panels<br />
8
8013<br />
Performance of the two panels<br />
(univariate analysis)<br />
9
performance of the panels<br />
• usually, the expert panels should have many qualities:<br />
8013<br />
• discrimination: panelists should be able to detect and describe<br />
the differences existing between the products<br />
• reproducibility: panelists should describe the products in the<br />
same way, when they are repeated<br />
• agreement: panelists should give the same description of the<br />
products as the rest of f the panel<br />
• it can be measured with the correlations (usually, one<br />
panelist pa e s is s co compared pa ed to o the e mean ea oover e the e rest es oof the e<br />
panel)<br />
10
expert panel<br />
• panel performance<br />
• discriminate on 11 out of 12 attributes (“Agrume”, pvalue=0.08)<br />
• reproducible for 11 out of 12 attributes (“Notes Florales”)<br />
• panellist performance (discrimination (discrimination, reproducibility)<br />
8013<br />
• panellists 1, 3 and 12 are very good<br />
• panellists 8, 9 and 10 are not good in discrimination<br />
(discriminate the products on less than 6 out of 12 attributes)<br />
• panellist lli 9 iis also l not good d iin reproducibility d ibili ( (reproducible d ibl on<br />
only 3 out of 12 attributes. “Notes Florales”, “Agrume” and<br />
“Enveloppant”)<br />
11
expert panel (correlations)<br />
• distribution of the correlations (correlation between expert i and the<br />
mean over the (n-1) ( ) others) )<br />
8013<br />
12
consumer panel<br />
• discrimination (on the twelve original products)<br />
• the consumers discriminate the products on all attributes except<br />
“camomile” (pvalue = 0.62)<br />
• NB: the consumers discriminate on “Citrus” (pvalue < 0.001)<br />
• reproducibility (on the two duplicated products only)<br />
8013<br />
• consumers are reproducible on all attributes except one<br />
(“woody”)<br />
13
consumer panel (reproducibility)<br />
Shalimar<br />
Shalimar 2<br />
8013<br />
14
consumer panel (correlations)<br />
• distribution of the correlations (correlation between a consumer i and<br />
the mean over the (n-1) ( ) others) )<br />
8013<br />
15
conclusions on the panel performance<br />
• expert panel<br />
• discriminates between the products<br />
• are reproducible<br />
• high correlations<br />
• consumer panel<br />
8013<br />
• discriminates between the products<br />
• shows reproducibility reproducibility’s s qualities<br />
• lower but still positive correlations (consumers are untrained)<br />
Both panels show the same qualities<br />
16
8013<br />
Products’ spaces<br />
(multivariate analysis)<br />
17
methodology<br />
• products’ spaces<br />
• the products profiles (averaged over the panellists or consumers)<br />
are computed.<br />
• Principal Components Analysis is then run on these product x<br />
attribute matrices<br />
• comparison of the two products’ products spaces (expert and consumer) is a<br />
“multi-table problem”<br />
• comparison through the Procrustean analysis<br />
8013<br />
• comparison through Multiple Factor Analysis ( (MFA) )<br />
• comparison through the confidence ellipses technique<br />
18
Dimennsion<br />
2 (21.87 %) %<br />
3,5<br />
3,0<br />
25 2,5<br />
2,0<br />
1,5<br />
1,0<br />
0,5<br />
0,0<br />
-0,5<br />
-1,0<br />
-1,5<br />
-2,0<br />
-2,5<br />
-3,0<br />
-3,5<br />
-4,0<br />
-4,5<br />
expert panel g p<br />
Pleasures<br />
JAdore_EP JAdore_ET<br />
CocoMelle<br />
PurePoison<br />
LInstant<br />
Cinema<br />
AromaticsElixir<br />
Chaneln5<br />
LolitaLempicka<br />
Shalimar<br />
Angel<br />
-4 -3 -2 -1 0 1 2 3 4<br />
Dimension 1 (64.22 %)<br />
8013<br />
2 (21.87 %)<br />
Dimension<br />
1<br />
0<br />
Notes.florales<br />
Agrume<br />
Fraicheur.marine<br />
Fruité<br />
Vert<br />
-1 Gourmand<br />
Boisé<br />
Vanille<br />
Epicé<br />
Oriental<br />
Capiteux<br />
Enveloppant<br />
-1 0 1<br />
Dimension 1 (64.22 %)<br />
19
Dimennsion<br />
2 (17.97 %) %<br />
6,0<br />
5,5<br />
5,0<br />
4,5<br />
4,0<br />
3,5<br />
3,0<br />
2,5<br />
2,0<br />
15 1,5<br />
1,0<br />
0,5<br />
consumer panel<br />
JAdore_EP<br />
Cinema<br />
0,0<br />
-0,5 JAdore_ET<br />
CocoMelle<br />
-1,0<br />
-1,5<br />
-2,0<br />
-2,5<br />
-3,0<br />
-3,5<br />
-4,0<br />
Pleasures PurePoison<br />
-4,5<br />
-5,0<br />
-5,5<br />
-6,0<br />
-6,5<br />
-7,0 70<br />
LolitaLempicka<br />
LInstant<br />
Chaneln5 Shalimar<br />
AromaticsElixir<br />
Angel<br />
-5 -2.5 0 2.5 5 7.5<br />
Dimension 1 (68.29 %)<br />
8013<br />
Dimensionn<br />
2 (17.97 %)<br />
1<br />
0<br />
-11<br />
sweet_fruit<br />
freshness<br />
green<br />
jasmin<br />
rose<br />
citrus<br />
fresh_lemon<br />
g<br />
vanilla<br />
honey<br />
camomille<br />
caramel<br />
anis<br />
iintensity t it spicy i<br />
nutty<br />
animal<br />
musk<br />
incense<br />
leather<br />
woody<br />
earthy<br />
-1 0 1<br />
Dimension 1 (68.29 %)<br />
20
8013<br />
Multivariate comparison of the<br />
two panels (GPA and MFA)<br />
21
expert vs consumer: Procrustes analysis<br />
GPA consensus space<br />
(coefficient of similarity: 0.93)<br />
8013<br />
Dimm<br />
2<br />
0.4<br />
0.1 0.2 0 0.3<br />
0.0<br />
-0.3 -0.2 - -0.1<br />
JAdore_EP JAdore_ET CocoMelle<br />
Pleasures PurePoison<br />
Cinema<br />
LInstant<br />
LolitaLempicka<br />
Chaneln5<br />
AAngell Shalimar<br />
AromaticsElixir<br />
-0.2 02 00 0.0 02 0.2 04 0.4<br />
Dim 1<br />
22
expert vs consumer: Multiple Factor Analysis<br />
MFA partial points’<br />
re<strong>presentation</strong><br />
(RV coefficient: 0.87)<br />
8013<br />
Dim 2 (19.35<br />
%)<br />
2<br />
1<br />
0<br />
-1<br />
-2<br />
experts<br />
consommateurs<br />
Cinema<br />
LInstant<br />
JAdore_EP JAdore_ETCocoMelle<br />
PurePoison<br />
Pleasures<br />
LolitaLempicka<br />
Chaneln5<br />
Angel<br />
Shalimar<br />
AromaticsElixir<br />
-2 -1 0 1 2 3<br />
I di id lf t<br />
Dim 1 (64.06 %)<br />
23
expert vs consumer: Multiple Factor Analysis<br />
g<br />
MFA variables’<br />
re<strong>presentation</strong><br />
(RV coefficient: 0.87)<br />
8013<br />
(19.35 %)<br />
Dimension D 2<br />
1<br />
0<br />
-1<br />
expert<br />
consumer<br />
citrus<br />
Fruité<br />
sweet_fruit<br />
freshness<br />
green<br />
jasmin<br />
rose<br />
Agrume<br />
Notes.florales<br />
ffresh_lemon h l<br />
Fraicheur Fraicheur.marine marine<br />
Vert<br />
Gourmand<br />
vanilla<br />
honey<br />
Vanille<br />
caramel<br />
anis<br />
camomille<br />
intensity<br />
Boisé<br />
nutty<br />
Enveloppant<br />
animal<br />
musk<br />
woody<br />
leather<br />
incense<br />
earthy<br />
Capiteux<br />
Epicé<br />
Oriental<br />
spicy<br />
-1 0 1<br />
Dimension 1 (64.05 %)<br />
24
8013<br />
Comparison through the<br />
confidence fid ellipses lli ttechnique h i<br />
(Husson, ( , Lê & Pagès, g , 2005) )<br />
(Lê, Pagès & Husson, 2008)<br />
25
confidence ellipses<br />
methodology<br />
1.Compute the product profiles (averaged by product over the judges)<br />
2.Create the products’ space<br />
33.Re-sample R l bby bbootstraping t t i new panels l<br />
4.For each new panel, compute new products’ profiles<br />
5.Project as illustrative the products on the original product space<br />
6.Steps 3 to 5 are repeated many times (i.e. 500 times)<br />
7.Confidence ellipses around the products containing 95% of the data are<br />
constructed<br />
8013<br />
principle<br />
• if ellipses are superimposed, the products are not significantly different<br />
• the size of the ellipses is related to the variability existing around the<br />
products p<br />
26
confidence ellipses<br />
Dim 2 (19. .39%)<br />
2<br />
1<br />
0<br />
-1<br />
8013<br />
Confidence ellipses around the products<br />
JAdore_EP<br />
JAdore_ET<br />
Pleasures<br />
CocoMelle<br />
Cinema<br />
LInstant<br />
PurePoison<br />
LolitaLempicka<br />
Chaneln5<br />
Angel<br />
Shalimar<br />
AromaticsElixir<br />
-3 -2 -1 0 1 2 3 4<br />
Dim 1 (64.02%)<br />
27
confidence ellipses<br />
• as we have two different panels, we can apply this methodology to<br />
both<br />
8013<br />
• creation of confidence ellipses around each product seen by<br />
each panel (24 ellipses are created here)<br />
• comparison of a given product through the two panels (same<br />
colour)<br />
• comparison of f the different ff products within a panel (same ( type<br />
of line)<br />
28
Dim 2 (199.39%)<br />
confidence ellipses<br />
2<br />
1<br />
0<br />
-2 - -1<br />
8013<br />
cons.<br />
expert<br />
Confidence ellipses for the partial points<br />
Cinema<br />
LInstant<br />
JAdore_EP JAdore_ET CocoMelle<br />
PurePoison<br />
Pleasures<br />
LolitaLempicka<br />
Chaneln5<br />
Angel<br />
Shalimar<br />
AromaticsElixir<br />
-4 -2 0 2 4<br />
Dim 1 (64.02%)<br />
(64 02%)<br />
29
confidence ellipses<br />
• partial points<br />
8013<br />
• within a product, the ellipses related to the two panels are<br />
always superimposed (no differences between the panels)<br />
• the sizes of the ellipses are equal<br />
• the higher amount of consumers compensate the higher<br />
variability due to the lack of training for consumers<br />
30
conclusions<br />
• although consumers don’t don t have the habit to describe perfumes<br />
(difficult task), they give the same information as the expert panel (and<br />
it’s identical to the standard description of the perfumes)<br />
• they also have the same qualities (discrimination and reproducibility)<br />
• a difference between consumers and experts panel exists in the<br />
variability of the results (more variability for consumers), but this is<br />
compensated by the larger size of the panel (here 103 vs 12)<br />
• with consumers, co su e s, not o oonly y intensity, e s y, bu but aalso so ideal dea aand d hedonic edo c<br />
questions can be asked in the same time<br />
8013<br />
31
eferences<br />
• Earthy P., MacFie H & Hedderlay D. (1997). Effect of question order on sensory<br />
perception and preference in central locations. Journal of Sensory Studies, vol.12, p215-<br />
237<br />
• Gazano G., Ballay S., Eladan N. & Sieffermann J.M. (2005). Flash Profile and flagrance<br />
research: using the words of the naïve consumers to better grasp the perfume’s perfume s universe universe.<br />
In: ESOMAR Fragrance Research Conference, 15-17 May 2005, New York, NY.<br />
• Husson FF., Le Dien SS. & Pagès J J. (2001) (2001). Which value can be granted to sensory<br />
profiles give by consumers? Methodology and results. Food Quality and Preference,<br />
vol.16, p291-296<br />
• Husson F., Lê S.& Pagès J. (2005). Confidence ellipses for the sensory profile obtained<br />
by principal component analysis. Food Quality and Preference, vol.16, p245-250<br />
• Lê S., Pagès J. & Husson F. (2008). Methodology for the comparison of sensory<br />
profiles provided by several panels: Application to a cross cultural study. Food Quality and<br />
Preference, vol.19, p179-184<br />
8013<br />
32
thank you<br />
• special thanks to<br />
• Melanie COUSIN<br />
• Maëlle PENVEN<br />
• Mathilde PHILIPPE<br />
• Marie TOULARHOAT<br />
students from AgroCampus-Rennes, who took care of the whole<br />
expert panel data.<br />
8013<br />
33
Thank you for your attention!