How reliable are the consumers? Comparison of sensory profiles from consumers and experts
Introduction
In sensory analysis, one of the most important tools is the quantitative characterization of the perceivable product attributes. In the literature, this tool is referred to as “descriptive analysis”, or “profiling” (two frequently used profiling methods are quantitative descriptive analysis (QDA®, Stone, Sidel, Oliver, Woosley, & Singleton, 1974) and Spectrum™ (Meilgaard, Civille, & Carr, 2006)). These methods use trained or expert panels. Because of their routinely use of the type of products in question, and because of dedicated training sessions, these panels seem to be more able to characterize products in an accurate way than naïve consumers. On the other hand, hedonic questions are also of great importance and most practitioners use consumers for hedonic tasks. So trained panels are required for sensory profiles and consumers are required for hedonic profiles. In the literature, many warnings are given concerning the use of consumers for profiling:
- –
“…as with any untrained panel, beyond the overall acceptance judgment there is no assurance that the responses are reliable or valid” (Stone & Sidel, 1993)
- –
“…consumers can only tell you what they like or dislike” (Lawless & Heymann, 1999)
According to these practitioners, profiling results from consumers lack two essential qualities: consensus between respondents and reproducibility.
Moreover, it has also been shown, that asking consumers liking and intensity questions in the same test can return an unwanted halo effects (Earthy, MacFie, & Hedderley, 1997). The potential impact of the attribute questions on the hedonic ratings is a key point in the objection of the use of getting sensory information from consumers. Since the aim of this paper is to compare experts’ and consumers’ sensory profiles, this point is not studied here.
In market research, most companies need quick answers about their products. Hence, they do not always have the possibility to train panels (which is time consuming). Profiles obtained with consumers can be a good alternative, depending on the type of tests one is interested in, especially in view of the fact that consumers’ profiles also meet the requirements discrimination, panelists’ consensus and reproducibility (Husson, Le Dien, & Pagès, 2001). Moskowitz also showed that consumers can be used to assess the sensory descriptions of sauces, and hence “refutes the notion that consumers are incapable of validly rating the sensory aspects of products” (Moskowitz, 1996).
Because of these two notions (training panels takes time, and consumers are not allowed to profile products), a number of faster methods for collecting sensory data have been developed. Among them is free-choice profiling (Williams & Langron, 1984), Flash profiling (Sieffermann, 2000, Siefferemann, 2002), and ultra flash profiling (Perrin et al., 2008). These methods have in common that they avoid training sessions beforehand, and that they use naïve consumers (Gazano et al., 2005, Nestrud and Lawless, 2008). Paradoxically, it is well accepted that consumers can be used for profiling products using these methods, but the use of consumers with standard QDA® type profiling using a fixed, predefined vocabulary is still subject to criticism. The question remains: How reliable are consumers? To answer this question, classical sensory profiles, obtained from an expert and a consumer panel on the same products, are compared.
Section snippets
Data
The datasets provided here concern 12 luxurious women perfumes. The list of the perfumes is given Table 1. These 12 perfumes were profiled by an expert and a consumer panel.
The expert panel was run in Agrocampus Ouest (Rennes, France) with 12 persons (eleven students and one teacher) from the Chantal Le Cozic School (esthetic and cosmetic school in Rennes). First, two focus groups, with two moderators, were conducted. Then, a summary discussion was conducted, and a list of 12 attributes was
Unidimensional aspects
To measure the quality of the two panels, the following unidimensional measures have been computed: the product discrimination and the panel reproducibility through ANOVA and the panelists’ consensus through the correlations between each panelist and the average of the panel without that panelist.
Multidimensional aspects
For each panel, the product space is computed by Principal Components Analysis on the products’ profile, where one profile is a table crossing the products (i) in rows and the attributes (k(panel), with k(expert) = 12 and k(consumer) = 21) in columns, and the cell (i, k(panel)) is the average score for the product i and the attributes k(panel).
In order to compare the results given by the experts with those given by the consumers, the two products spaces are submitted to multiple factor analysis (
Conclusions and comments
Both panels are similar in terms of discriminatory ability and reproducibility. By looking at the pair comparisons given by the confidence ellipses, the results are close, even though some specificity for each panel can be observed. In terms of panelists’ consensus, the experts show more consistencies than the consumers: this might be due to the fact that they have a better knowledge about this type of product (through their experience and the training sessions). Moreover, they defined their
Software
The analyses were done with R 2.8.0 (R Development Core Team, 2008), with the packages SensoMineR v1.08 (Lê & Husson, 2006) and FactoMineR v1.10 (Husson, Lê, Josse, & Mazet, 2007), and with Senstools.
Acknowledgements
The authors would like to thank M. Cousin, M. Penven, M. Philippe and M. Toularhoat, students in applied statistics in Agrocampus Ouest (Rennes), who managed the expert study. They also would like to thank the reviewers for their constructive comments.
References (31)
Assessing panel consonance
Food Quality and Preference
(1995)- et al.
Which value can be granted to sensory profiles given by consumers? Methodology and results
Food Quality and Preference
(2001) - et al.
Confidence ellipse for the sensory profiles obtained with principal component analysis
Food Quality and Preference
(2005) - et al.
Measurement of the reliability of sensory panel performances
Food Quality and Preference
(2006) - et al.
The ACT (STATIS method)
Computational Statistics and Data Analysis
(1994) - et al.
Methodology for the comparison of sensory profiles provided by several panels: Application to a cross-cultural study
Food, Quality and Preference
(2008) Base size in product testing: A psychophysical viewpoint and analysis
Food Quality and Preference
(1997)- et al.
Perceptual mapping of citrus juices using projective mapping and profiling data from culinary professionals and consumers
Food Quality and Preference
(2008) - et al.
Comparison of three sensory methods for use with the Napping® procedure: Case of ten wines from Loire valley
Food Quality and Preference
(2008) - et al.
Defining the underlying sensory dimensions
Food Quality and Preference
(2000)
The quest for the ideal product: Comparing different methods and approaches
Food Quality and Preference
Effect of question order on sensory perception and preference in central location trials
Journal of Sensory Studies
Analyses factorielles simples et multiples
Le traitement des variables vectorielles
Biometrics
Cited by (125)
Description of virgin olive oil: Descriptive evaluation by trained assessors vs flash profile using panels with diverse degrees of training
2024, International Journal of Gastronomy and Food ScienceScent and sustainability: investigating consumer evaluations of biocatalysis and naturalness in fragrances
2023, Food Quality and PreferenceComparison of descriptive analysis and flash profile by naïve consumers and experts on commercial milk and yogurt products
2023, Food Quality and PreferenceFirst steps towards FAIRization of product-focused sensory data
2023, Food Quality and PreferenceAlternative descriptive methods answered by consumers for the sensory characterization of meat products: Fundaments and methods
2022, Sensory Analysis for the Development of Meat Products: Methodological Aspects and Practical Applications