Check of panel consistency and data variability in pre-testing evaluations

Marit Rødbotten; Asgeir Nilsen
Matforsk AS, Norway
 Marit.rodbotten@matforsk.no

Sensory evaluations are based on responses from a given number of subjects. In consumer tests the consumers may very well disagree on the choice of a preferred product. In the situations where we want information about intensities of various attributes of the products in consideration, we also want to know the scientific truth about the result. Since the instrument used in this analysis is a group of human beings, it is often seen that the persons in the “instrument” give scores that indicate disagreement in the perception of one or more of the analysed attributes. When that is the situation, the result of the analysis is less valid than when all the panellists in the “instrument” agree on the sensory description of the products. A disagreement among the panellists can be explained by various reasons, but often the main reason is lack of qualified training.

Before the real trial of a sensory profiling is being conducted, a pre-trial is recommended. In the pre-trial often samples with a certain variation in sensory profile are tested with the purpose to find an agreement on which attributes best describe the sensory variation between the samples, and to ensure that each panellist is able to recognise and evaluate each of the attributes.

This presentation will demonstrate some examples which show how panellists agree or disagree in their evaluations. Plots will illustrate various situations, both when there is a disagreement between the panellists and also when a panellist gives bad replications of his own scoring in cases when samples are blind coded and served more than once. Plots will show if there is one or more of the analysed attributes that needs to be re-tested and discussed in more detail before the real trial can start.
It is also possible to see if one assessor needs more training on a certain attribute to be in agreement with the rest of the panellists.

On  www.matforsk.no/panelcheck you may find the data program where this tool to control the performance of a sensory panel is available for down loading.

Download presentation