CONSUMER PRODUCT TESTING 775 tested and that they do not vary significantly. After they have been placed, other samples are periodically checked in the laboratory so that any unanticipated changes occurrinK during the testing period will be known. Interpretation of Consumer Responses Since in most of this work the response sheets are a series of short essays, each one is read for meaning and perspective. The respondent, whether interviewer or consumer, assumes this will be done. For ex- ample, for Soup A the consumer may have written that its unfavorable feature was "not salty enough," and for Soup B, "good rich flavor." Since the Flavor Profiles showed that both soups had the same salt level, she is not saying she would like to taste salt. She is stating that Soup A's flavor needed something and to improve it she would have added salt. If her descriptions for each soup had been read separately, Soup A might have been tallied under "salt level low" this would be inaccurate and misleading. The value of use profiles for interpretation is illustrated by the fol- lowing example. In a series of paired comparisons involving six bal- anced panels, the objective was to select the best flavored product of three. Each product had been tested against the other and also against a control. The six sets of response sheets were read separately, and it was found that Product B evoked a seemingly different reaction when tested against Product A than against the control. Against Product A it was called flat against the control it was called pleasant tasting. Without the use profiles of the products, one could not have interpreted these findings. But they dovetailed nicely with the use profiles of A rs. B, A rs. control, and B rs. control, which showed each pairing had differ- ent flavor parameters. Product A's over-all flavor was stronger, more identifiable but less appropriate the control's flavor had several nega- tive components Product B's flavor was weaker than A's, stronger than the control's, and more appropriate than either. In other words, the test situation was different with each pairing, and the use profiles had defined the differences. A research study of an oral product is a final example of relating con- sumer findings to product properties. This study was aimed at defining the product's important flavor elements and the ranges wherein these elements could make positive and negative contributions to preference. First, the consumer panelists isolated the important elements. Then, in a series of paired comparison tests conducted over a two-year period, the

776 JOURNAL OF THE SOCIETY OF COSMETIC CHEMISTS intensity of each flavor element was varied separately, each time relating the consumer findings back to the test products' Flavor Profiles. After only six tests the study furnished the fl avonring formulators with a Profile- blueprint. This tabulation defined the flavor character notes, their upper and lower intensity limits, and their order of sequence. The formulator with his expert knowledge of flavoring materials could then create new models which he was able to evaluate at the bench. Thus, it is possible to relate consumer findings to laboratory measure- ments in order to draw up a product's specifications in meaningful techni- cal terms. If the basic principles of good testing are followed when dis- criminating consumers make up the test panel, a product that fails to pass a consumer product test can be purposefully modified and a prod- uct that passes is ready to be turned over to the marketing people.