Selection of “matched” areas for test marketing is an important undertaking if reliable comparisons among markets are to be made. This usually has been done on a rather arbitrary basis, largely because of the large number of market characteristics on which markets can be viewed as similar or different. The authors suggest a numerical procedure—cluster analysis—for matching prospective test markets on the basis of a large variety of characteristics which could affect test marketing results. In this way, markets can be pre-selected so as to reduce undesired variability among test areas.