Hedonic scores

Table III presents a typical training panel s hedonic scores for the 17 words discussed previously. Although the panel was asked to use "L" and "D" to denote like and dislike, the scale is actually positive numbers for like and negative numbers for dislike. Our experience has shown that the panelist can use L and 0 with much less difficulty than plus and minus.

CATA questions can be used concurrently with hedonic scores with the aim of understanding consumer preferences and identil g recommendations for product reformulation (Stone and Sidel, 2004). However, including questions about specific sensory characteristics can be a source of bias on hedonic scores (Popper et al 2004 Prescott ct a/., 2011). 

On the left side, the sensory profiles of the products in the middle, the ideal profiles on the right side, the vector of hedonic scores. Here, j indicates the consumer, p the product and a the attribute. 

At the end of the test, each consumer has thus generated three blocks of data (Table 14.1) the sensory profiles of the products (noted the ideal profiles (noted Zjpa) and the vector of hedonic scores (noted hp). 

In IPM, the sensory profiles of the products, the information regarding the consumers ideals, and the hedonic scores are collected directly for a set of products. Once the test is finished, the data are immediately actionable for product optimization and/or product development. The only missing parameter at this moment is the coefficient used to estimate the weight of each attribute on liking. Since the consumers have provided information on how they perceived the products and on how they liked them, the weights b can be computed. 

SO, the correlation is calculated across attributes between the corrected ideal scores of a consumer and the correlation coefQcient defining drivers of liking. In the first case, the corrected ideal score of an attribute is positive if the consumer wants more of that attribute in the second case, the correlation between the hedonic scores and the perceived intensity of an attribute is positive if the attribute is a driver of liking. For a consistent consumer, such correlation is expected to be high and positive. 

Consumers are consistent from a sensory point of view if they rate their ideal product with similar sensory characteristics to the product they like most. The evaluation of sensory consistency (at the panel level) is done by evaluating whether the ideal information provided is making the link between the perception and the appreciation of the products. Such evaluation is done by double projection as supplementary of the sensory profiles (supplementary entities) and the hedonic scores (supplementary variables) within the ideal space (Fig. 14.3). 

Figure 14.3 Evaluation of the sensory consistency of the ideal products at the panel level. Panel (a) represents the consumer ideal space with projection as illustrative of the products profiles (in grey). Panel (b) shows the corresponding variables representation with projection as illustrative of the hedonic scores (in grey).

Thanks to proper ANOVA, we checked that the presence of co-pilots in the car, the weather conditions and the order of presentation did not bias mean hedonic scores. The 151 drivers significantly discriminated the seven minivans. They preferred the... 

A Hierarchical Ascendant Classification (Euclidean distance, Ward s criterion) was performed on the hedonic scores of the entire population four clusters can be distinguished (Fig. 20.12). Even if the number of consumers was low for Clusters 1 and 2, it seemed of interest to consider these drivers separately because they have separate opinions on RC3-P2. Cluster 1 differentiated the cars according to their seat, regardless of the chassis. Cluster 2 liked RC3-S3 and rejected RC3-P2. Cluster 3 differentiated the cars according to their chassis and not their seat. Finally, Cluster 4 differentiated the cars according to their chassis and then to their seat. 

Figure 20.11 Hedonic scores of the consumers study (ANOVA F(6,1050) = 40.452, p < 0.001 - Turkey s honest significant difference (HSD) 5%).

The hedonic scores of 90 consumers are better modelled by the sensory profile provided by the eight assessors who differentiate the cars with the roUing chassis ... 

---