Identifying preferences in a jury study using clustering techniques

In the study of product sound quality using paired-comparison methods, it is generally assumed that preference within the jury population is universal. This assumption often holds true, but there are often scenarios where preference varies significantly with demographics. Identifying the presence of and partitioning the jury members with these varying criteria for preference is seldom a trivial exercise. This article presents an unsupervised clustering technique that can be applied to jury-paired preferences in an attempt to infer the number of subgroups in a jury pool. The same algorithm can then be used to classify jury members into appropriate subgroups. In addition, the approach is applied to the results of a small-engine jury study of more than 80 members. Two subgroups, corresponding to American and European jurors, were successfully identified and classified.