Independent Variable

Ingredient A Ingredient B Ingredient C Ingredient D Ingredient E Cook-process Cost-of-Goods





















Quality Control6










I/Pieces-surface I/Color I/NaCl I/Acid












"C = Consumer rating. Like = liking. Sen = sensory.

Segment = sensory preference segments (showing different patterns of liking).

4QC = Quality Control Panel. CI = Instrumental measure. dE = Expert panel (QDA Method).

studies involve fewer variables (often no more than 2-3), and may use only one panel (eg, consumers from a research guidance panel). However the data becomes more variable once the data set is augmented to comprise consumer and instrumental data.

Panel Selection

Depending upon when the research is undertaken (early versus late in the development cycle) the researcher will work with different consumer panels. In the earliest development stages researchers work with an in-house group of consumers or experts. Often, however, it is necessary to obtain ratings from users of different and competing products in the category. It is also instructive to run the study in different geographical areas (or even in different countries). The researcher hires field services in the different markets and "tailors" a consumer panel with specific demographic and usage characteristics. Fewer consumers participate in these guidance tests than in subsequent large scale market tests. This study tested products among 120 female consumers, 40 consumers in 3 markets, (in order to ensure representativeness). Optimization studies are usu ally of this magnitude, with panel sizes ranging from 50 to 200 participants.

Panels for optimization studies usually comprise users of the category, occasionally (and more specifically) users of the product being optimized and/or users of competitive brands. Here half the panel comprised consumers who used the manufacturer's product most often, and the other half comprised consumers who used competitor products. By incorporating different groups of consumers in the same panel, (all of whom test the same array of products), the investigator can compare ratings assigned to the same products by the two groups. The comparison can be used to develop optimal products which satisfy one target group, or both groups simultaneously.

Activities During the Evaluation

Depending upon the product being tested, panelists may either participate in a supervised session in a central location, or test the products at home, without supervision.

Panelists followed the protocol presented in Table 15. A central location setup makes it easy to monitor the quality of the data as the panelists assign their ratings. Close supervision of the evaluation ensures that panelists remain alert and motivated. Neither the interviewers nor the panelists "know the correct answer." Panelists must, perforce, answer honestly because there are no other cues to aid them other than their sensory perceptions. The ongoing question and answer dialogue between the panelist and the interviewer (after each product is rated) maintains motivation. For home use tests (appropriate when panelists prepare the product themselves, or use the product as an ingredient), other procedures are necessary. These procedures include giving the panelist one or two products to use, instructing the panelist to rate each product just after eating it, and then requiring the panelist to return to the test site where the panelist is questioned about the ratings that he or she assigned. Upon returning to the test site the panelist receives the next set of products, and follows the same protocol. Control in a home use test is maintained by

Table 15. Protocol for the Sauce Study

Panelists prerecruited to participate, by a telephone call.

Panelists are screened to be appropriate (viz, category users, interested in the concept which an interviewer reads to them).

Panelists show up, prepared for a 4-hour session. Panelists show up, in groups of 15-20 (randomly recruited, so that the panelists do not know each other).

Panelists are oriented in scaling by a short practice exercise. Panelists rate water and cracker on sensory and liking attributes.

Panelists try the first product, rating it on all attributes. Products are randomized to reduce order bias. Products are rated on attributes as they appear (viz, appearance first, then aroma, then texture). Liking ratings and sensory ratings are interdigitated.

Ratings checked by an attending interviewer on a panelist-by-panelist basis, to ensure panelist comprehension.

Panelist waits 15 minutes and goes to the next product.

After panelist finishes, the panelist is paid and dismissed.

Each panelist rates 11 products (randomized from the full set).

the careful allocation of products to panelists, along with an orientation session at the start of the study (before actual home use). Motivation is maintained by payment for the successful completion of the test and by the ongoing question and answer interchanges between panelists and the interviewers.

Initial Data Analysis—Product X Attribute Matrix

The initial data comprises a matrix of mean ratings of each product by each attribute. Table 16 shows part of this matrix. Statisticians may prefer to adjust the means in Table 16 to account for the fact that each panelist evaluated only a randomized subset of products. These "adjusted means" rather than the actual means, would appear in Table 16.

Table 16 alone teaches a lot about the products because the researcher can compare products, on the different attributes. Even without an experimental design several instructive analyses (correlation, regression, and histogram plotting) highlight key dimensions of the category.

Questions investigators can answer with Table 16 (in its entirety) include:

1. For a specific attribute do the products differ a great deal or relatively little? What is the range of scores? By looking at the highest and lowest means across products on the attribute, and graphing the distribution of the means within that range the investigator sees the "spread" of products. If the raw data is also available, then a 1 way analysis of variance (ANOVA) provides a measure of discrimination, (viz, the F ratio). The F value measures the "signal to noise" ratio, or variability due to product differences versus variability due to random error (eg, panelist differences).

2. What is the distribution of the mean liking ratings across products? If competitor products score in the 50s and 60s (on a 0-100 point scale), whereas most of the prototypes (from the experimental design) score in the 30s and 40s, then the optimized product will probably not score higher than the competitors. The only way that one could develop a superb product would be if the optimal formulation lies in a region of ingredient values not tested in the original experimental design. (In that case the optimal solution would be suspect).

3. Which liking attributes covary with overall liking? This analysis reveals important attributes. Figure 2 shows a schematic plot of overall liking (ordinate) versus attribute liking (abscissa). The correlation between the two attributes describes the strength of a linear relation between attribute liking and overall liking, but does not show the slope. A straight line shows the quantitative nature of the relation. The steeper the slope of the straight line the more small changes in attribute liking covary with large changes in overall liking (and therefore the more important the attribute).

4. Which sensory attributes covary with overall liking? As a formula variable or sensory attribute changes, liking also changes. Often this relation exhibits an

Table 16. Partial Data Base for Sauces


Berry Boosters

Berry Boosters

Acai, Maqui And Many Other Popular Berries That Will Change Your Life And Health. Berries have been demonstrated to be some of the healthiest foods on the planet. Each month or so it seems fresh research is being brought out and new berries are being exposed and analyzed for their health giving attributes.

Get My Free Ebook

Post a comment