Project Topics
Get Complete Project Material File(s) Now! »
Sensory analysis with consumers using Free-Comment
To the best of our knowledge, the first reported study using Free-Comment (FC) in the context of sensory analysis with consumers was that of ten Kleij and Musters (2003). The motivations of these authors lied in the fact that free responses “are not often used for detailed analyses” while they “undoubtedly contain very rich information” with the additional benefit of being “stated in consumer language”. Their study confirmed that FC indeed provides rich information able to characterize a product space with a “striking” agreement to conventional sensory profiling.
The main practitioners’ motivations for using FC rather than other descriptive methods with consumers are that FC does not rely on a pre-established list of sensory descriptors and it is based on a natural descriptive presence/absence principle. This presence/absence principle is easier and faster for consumers than ratings and rankings because it is cognitively lighter. Further, because FC does not rely on a pre-established list of sensory descriptors, it provides less biased descriptive sensory information than CATA, which is the list-based presence/absence method.
Avoiding the limitations from lists of sensory descriptors
Using a pre-established list of sensory descriptors induces several biases, thus not using such a list might be the most important benefit of FC.
Lists of sensory descriptors are likely to steer consumers in some directions and suggest to them sensory descriptors they would not have thought without the list (Coulon-Leroy, Symoneaux, Lawrence, Mehinagic, & Maitre, 2017; Kim, Hopkinson, van Hout, & Lee, 2017; Krosnick, 1999; Reja, Manfreda, Hlebec, & Vehovar, 2003; Schuman & Presser, 1979; Züll, 2016). On the contrary, FC enables the gathering of spontaneous unbiased descriptions (Lebart & Salem, 1994) that are not influenced by the practitioners and their preselection of possible applicable sensory descriptors (Foddy, 1993; Reja et al., 2003). In extreme cases of influence, the descriptive sensory information gathered by list-based methods could simply be the confirmation of practitioners’ expectations (Züll, 2016). This occurs when the list of sensory descriptors is not properly established and thus it does not let the opportunity to consumers to disagree with the practitioners. Further, if the list does not enable consumers to report what they indeed perceive, the descriptive sensory information gathered is inevitably biased by the dumping effect (Campo, Ballester, Langlois, Dacremont, & Valentin, 2010; Coulon-Leroy et al., 2017; Krosnick, 1999; Varela et al., 2018). This dumping effect occurs when consumers cannot report what they perceive because it does not belong to the proposed sensory descriptors. In those situations, consumers report the sensory descriptors they judge the closest to what they perceive. Depending on the difference between the perception and the sensory descriptors of the list, the dumping effect can lead to strong misinterpretations. To avoid the previous limitations, an “other” option might be included in the list. This additional option would aim to invite consumers to volunteer their own sensory descriptors if the ones proposed in the list do not appear relevant to them for describing the products under interest. However, it is unlikely to conduct to the expected results as these options are generally ignored (Castura, 2009). Consumers are likely to restrict themselves to the sensory descriptors listed, even if the most appropriate sensory descriptors to describe their perception are not included in the list (Krosnick, 1999; Reja et al., 2003; Schuman & Presser, 1979; Schuman & Scott, 1987). This might result in missing some information. On the contrary, FC reduces the risk of missing some key information as the consumers are somewhat forced to volunteer their own sensory descriptors without the possibility of taking refuge in those of the list (Reja et al., 2003; Schuman & Presser, 1979).
Limitations of Free-Comment
The main limitation of FC is that it requires a relatively extensive pretreatment to establish an a posteriori list of sensory descriptors. Since it exists several ways of conveying the same descriptive sensory information with possible typing errors, this pretreatment is time-consuming and cumbersome relatively to list-based methods that do not require such pretreatment (Hanaei, Cuvelier, & Sieffermann, 2015; Payne, 1980; Reja et al., 2003; Sheatsley, 1983; Symoneaux, Galmarini, & Mehinagic, 2012; ten Kleij & Musters, 2003).
Another limitation of FC is that some consumers, generally the less educated ones, might encounter some difficulties to verbalize properly their perception (Krosnick, 1999; Reja et al., 2003). This might result in some broad and general sensory descriptors in the a posteriori list that bring only overall and imprecise descriptive sensory information (Schuman & Presser, 1979). Further, some consumers might provide only hedonic information. On the contrary, when a priori lists of sensory descriptors are properly established, they produce only informative descriptive sensory information since practitioners do not include uninformative sensory descriptors or hedonic ones in the lists (Reja et al., 2003; Schuman & Presser, 1979), at least when they are solely interested in sensory descriptive information. Further, using a pre-established list of sensory descriptors enables to provide the consumers a definition of each of these descriptors to minimize the risk of misinterpretations of the products’ characterizations. While this can be tedious and does not entirely erase any risk because of individual interpretations of the definitions, this offers an opportunity to render list-based methods less subject to misinterpretations than FC.
FC instructions require being very precise in the way they are stated as well as the most possible focused on a single aspect (Reja et al., 2003; Symoneaux et al., 2012; Züll, 2016). If they are not, the descriptions might be uninformative regarding the aspect investigated. On the contrary, list-based methods do not necessitate such preciseness in their instructions since the list of sensory descriptors guides consumers on the aspect investigated. Further, list-based methods enable to render relevant and applicable sensory descriptors considered as trivial and/or obvious by consumers while FC might miss this information due to consumers not mentioning these sensory descriptors (Lebart & Salem, 1994).
Popularity of Free-Comment
The other descriptive methods of sensory analysis with consumers but FC and CATA are based on rating or ranking products under interest regarding several sensory descriptors. Since this is relatively cognitively heavy and difficult for consumers and further that rating is sometimes criticized, using a presence/absence principle appears the best practice for sensory characterization with consumers. In this context, the popularity of FC is only compared to that of CATA.
Pretreatment of Free-Comment data
Some limitations regarding the reported approaches for pretreating Free-Comment (FC) data presented in chapter I can be mentioned. Most of the time, the pretreatment of FC data was reported to be manually conducted and thus time-consuming (Ares et al., 2010; Hanaei et al., 2015; Symoneaux et al., 2012; ten Kleij & Musters, 2003). The fact that negations seem not to be taken into account systematically can lead to huge misinterpretations since for example strong and not strong is “a big difference!” (ten Kleij & Musters, 2003). Grouping the terms with similar meanings into a single term adds subjectivity in the procedure because a single term is selected more or less arbitrarily to represent all its synonyms. Further, it renders the grouping procedure unclear and it discards nuances of the terms provided by the consumers, which results in losing a part of the richness of the FC method. Grouping the terms with similar meanings only based on semantic considerations can be quite subjective and further time-consuming due to the need for considering every possible grouping and due to the resulting arbitration. Finally, using an overall threshold of citations for a term to be considered in statistical analyses is suboptimal since it does not guarantee any consensus from the consumers while repartition-dependent thresholds (a certain percentage of consumers for at least one same product) do to some extent.
Table of contents :
CHAPTER I: INTRODUCTION
A. BASICS OF SENSORY ANALYSIS
B. SENSORY ANALYSIS WITH CONSUMERS
C. SENSORY ANALYSIS WITH CONSUMERS USING FREE-COMMENT
D. AIMS AND STRUCTURE OF THIS MANUSCRIPT
CHAPTER II: GATHERING AND PRETREATMENT OF FREE-COMMENT DATA
A. GATHERING OF FREE-COMMENT DATA
B. PRETREATMENT OF FREE-COMMENT DATA
CHAPTER III: STATISTICAL ANALYSES OF FREE-COMMENT DATA
A. CONTEXT AND CONTENTS
B. ACCOUNTING FOR THE DIMENSIONALITY OF THE DEPENDENCE BETWEEN PRODUCTS AND SENSORY DESCRIPTORS IN ANALYSES OF FREE-COMMENT DATA
C. A MULTIPLE-RESPONSE CHI-SQUARE FRAMEWORK FOR THE ANALYSIS FREE-COMMENT DATA
CHAPTER IV: PERFORMANCES OF FREE-COMMENT AS COMPARED TO CHECK-ALL-THAT-APPLY
A. CONTEXT AND CONTENTS
B. DISCRIMINATION AND CHARACTERIZATION OF THE PRODUCTS
C. STABILITY OF THE PROVIDED DESCRIPTIVE SENSORY INFORMATION
CHAPTER V: EXTENSIONS OF FREE-COMMENT
A. CONTEXT AND CONTENTS
B. TEMPORAL SENSORY ANALYSIS: FREE-COMMENT ATTACK-EVOLUTION-FINISH
C. DRIVERS OF LIKING IDENTIFICATION AND IDEAL PRODUCT CHARACTERIZATION: IDEAL-FREE-COMMENT PAIRED WITH LIKING SCORING
CHAPTER VI: DISCUSSION AND PERSPECTIVES
A. BENEFITS AND LIMITATIONS OF FREE-COMMENT
B. GATHERING AND PRETREATMENT OF FREE-COMMENT DATA
C. STATISTICAL ANALYSES OF FREE-COMMENT DATA
D. PERFORMANCES OF FREE-COMMENT AS COMPARED TO CHECK-ALL-THAT-APPLY
E. EXTENSIONS OF FREE-COMMENT
CHAPTER VII: CONCLUSION
REFERENCES
