Determining Sample Size for Pretesting Comparative Effectiveness of Advertising Copies

This paper considers the problem of determining optimal sample sizes in advertising pretests, where two or more copies are compared for their relative advertising effectiveness measured on a dichotomous (0 or 1) scale. As the sample size is increased, sampling variations decrease so that the pretest has a better chance of identifying the truly best ad. Consequently, increasing the sample size decreases the opportunity costs of not selecting the best ad which, however, have to be balanced against the increased sampling costs. Taking these two considerations into account, three approaches (indifference zone, minimizing maximum loss and Bayesian) for determining sample size are discussed. The minimizing maximum loss approach seems to offer the best compromise in terms of realistic modelling and practical implementability. Extensions of the approaches to the case where advertising effectiveness is measured on an interval scale are outlined.