The k-Visit Method of Consumer Testing

When a pair of products, A and B, is being consumer-tested for preference, it is often found that the results of a single interview are insufficient to determine the attitude of the consumer population with respect to the products. At the interview it is usual to ask the consumer to indicate preference for one of the two products, or to declare no preference. The proportion of no preference votes is particularly dependent on the phrasing of the questionnaire or emphasis of the question. In addition, the information contained in the no preference (or tied) votes has often been wasted, misused, or misinterpreted, in distributing these votes in order to reduce data to two figures: the preferences for A and B. The problem is formulated below as a problem in estimation. A model for the behavior of the consumer population is proposed, and the parameters of the proposed model are estimated. The variancecovariance matrix of the estimators is obtained. A check on the appropriateness of the model is indicated, and some common departures from the model are mentioned. Finally, the method is illustrated with some numerical examples.