On the basis of the information developed from the previous steps, management is in a position to undertake research that will provide insights into whether they have been successful in developing the desired product. In many ways the types described in the preceding section overlap with the types used to test the product under actual use conditions. The discussion here will center on the research problems encountered when the objective is to test the product under “live use” conditions.
Management will, of course, test the product intensively in the laboratory whenever it is possible to simulate real world usage conditions. This is particularly true for products that can be submitted to lab tests to determine reliability and performance over time. For some industrial goods, it is feasible to have customers try the product under a set of specified conditions. It must be remembered that real world tests are designed primarily to determine whether the product’s physical attributes lead to satisfaction and preference.
Another way to test new products is through the use of a panel of experts. This is often the case with foods where experts evaluate “taste” and “aroma” to make sure the new product lives up to its benefits offered claims. In other cases different recipes can be tested to determine which is best. The advantage of such teats is their relatively low cost and the speed with which they can be done. The problem is that the experts may not accurately reflect the views of the actual consumers.
Paired Comparison Tests:
When the new product is designed to replace an existing one, or when the product’s competitors can be readily determined, blind paired comparison tests (where the consumer does not know the brands involved) can be employed. Such tests are relatively simple in concept but difficult to implement. They typically involve two variations of the same product, variations that often differ in ways not easily identified by consumers. Consumers who are members of the target market are asked to use both products and then to choose the one they like most. If it is desired to test several different product variations, a number of paired comparison tests must be run. Each variation must be tested against each other variation. Respondents must have only two different products to compare at any time; but they may have different pairs at different times, or different groups of consumers may be used for different pairs. If respondents are asked to compare several different product designs at one time, the results obtained may be misleading. For example, assume that three product designs, A, B, and C are to be tested.
Each respondent tests all three and is asked to pick the one preferred. The replies might be distributed as follows:
Product Design Percent of Respondents Preferring
At first glance it appears that design A should be the one selected for marketing. But is this conclusion valid? The 30 percent of respondents who voted for B might have preferred C over A if they could not have B. If this were true, 60 percent of the respondents would actually prefer design C over A. This vote splitting produces ambiguity, which is not found if only paired comparison are used (e.g. in the above situation, three different paired comparison tests would have to be run – A with B, A and C, and B with C).
From the above, it is likely that the results of blind paired comparison tests will be difficult to analyze. First and foremost, these tests cannot replicate the choices available in the market place. Second, the use situation is artificial in that the house wife tries two products one after the other. Third, the statistics obtained from the tests may be misleading or ambiguous. Assume B is preferred by only 25 percent what does this mean? If the 25 percent very strongly preferred the product over the alternative product then product A may be a real “winner” And finally, in selecting A over B a respondent may be stating preferenceor, if unable to discriminate, the choice made may be chance. But if A and B are relatively similar, as they are in many product tests, then the problem of random choices by non-discriminating subjects is very real.