Discriminators vs. Guessers or "Do they really care or see a difference?"

In the early experimental stage of brand development such things as sensory cues and aesthetics are researched to reinforce positioning or enhance the potential success of the brand. In many cases the choices are made via paired comparison blind testing. The results of these tests can look obvious when in many cases they are deceptive. There are two common problem areas with the paired comparison blind test method. First, the results are generally interpreted "as the product with the most votes is the appropriate candidate for the brand." This is not necessarily so, especially if the brand has a unique position in the product category. Consider the fact that a pleasant odor is preferred for a cleaning product. However, would you attempt to sell "Janitor in A Drum" with a pleasant or floral odor? I doubt it. The appropriate test for determining the feature of a brand must be assessed in context with the theme or concept of the brand. Some time in the future I will outline my thoughts on "Acceptability and Theme Support" an ideal protocol for addressing this problem.

The second problem area and the one I want to focus on at this time deals with the interpretation of the paired comparison results. That is, does a 50/50 split mean that no one cares about the two variables being measured? Or does it mean that the variables are really important and that there are two equally large segments of the population differing about which is best? To reconcile this question, I propose a "repeat paired comparison" test. This method will identify the proportion of preference discriminators and non-discriminators. The basic premise of this design is that half the guessers will reverse their preference on the second pair, thereby yielding an estimate of the proportion of non-discriminators. The protocol is as follows:

The respondent tests the first pair (taste, smell, feel, etc.)

The respondent does some other task.

A minimum of 5 minutes later, the respondent repeats the pair test. The stimulus is presented in the same order as in the first test.

The preference results are put into a 2 X 2 table. The table might look like the following:

                                        First Test Preference
                                        A        B

     Second Preference          A       32%     18%
                                B       22%     28%

In a future Views I will go beyond preference discrimination and cover overall non-discriminator analysis.