Flávio Sanson Fogliatto
UFRGS
Suzan L. Albin
Ana Maria dos Santos Silva
RUTGERS USA
E-mail: sffoglitto@ppgep.ufrgs.br
Abstract: The quality of several industrial products can only be fully measured through sensory
evaluation of some of its properties. That is particularly true in the food and pharmaceutical industry,
where product development technicians are often faced with the task of collecting and processing data
from sensory evaluation panels. In this paper we propose a new method for sensory data collection and
analysis. Our method presents two important features that may appeal to the product development
practitioner: (i) reliable sensory panel data may be elicited from untrained panelists, and (ii) a
consistency index which objectively measures how well panelists perform sensory evaluations is
determined for each panel member. The method we propose is applied in a case study from the food
industry.
Keywords: Indirect pairwise comparison method, sensory evaluation, product development.
1. Introduction
Sensory evaluation methods offer an organized way to collect information on sensory aspects of samples as
perceived by the human senses. These methods are applied in product development and reformulation, on-line
and off-line quality control, and for marketing purposes (monitoring competition, for example).
Samples evaluated in sensory panels correspond to different outcomes of an experiment, i.e., different
formulations of a product, manufacturing setups, and so forth. Ideally, sensory evaluation results should allow
the analyst to relate the factors varied when preparing samples to their corresponding sensory impact. In other
words, results are transformed into mathematical models used to predict sensory outcomes according to the
combination of factors chosen when preparing samples. Such models may only be determined when sensory
evaluation of samples yields quantitative data, one or more measurement per sample.
In this paper, we propose Indirect Pairwise Comparison (IPC) method for sensory data collection and
analysis. Our method is based on a family of psychophysical scaling methods introduced in the 50's
(STEVENS, 1957), and usually denoted by magnitude estimation (ME). Originally, ME was used to
determine mathematical relationships between physical intensities of an attribute and corresponding
subjective intensities, as perceived by panelists. Although mostly applied in psychophysical studies, ME is
also reportedly efficient when used in the sensory evaluation of products (see, LEIGHT &
WARREN, 1988).
We use the central idea of ME in the IPC method: measuring the intensity of an attribute as perceived from
different samples using ratios (i.e., describing the intensity of pairs of samples using a ratio of intensities). We
also incorporate to our method the use of graphic rating scales to measure responses, as suggested in
Quantitative Descriptive Analysis techniques (MEILGAARD et al., 1991). Data analysis in our method uses
analytic tools from SAATY's (1977) Analytic Hierarchy Process, a methodology used in decision making for
selecting the best among a set of alternatives, given some criteria
