This paper describes a method for using data from a series of parallel line assays to determine the optimal design for a future series of routine, as opposed to research, assays. Data are presented for a series of 34 assays of an antibiotic. The data are used to estimate the within and between assay variation in the slope and intercept of the Standard Preparation regression line. These estimates are substituted into expressions for the Mean Squared Error of the log potency estimate for several designs (one, two and three point). The design with smallest Mean Squared Error over future routine assays is chosen as optimal, even though it would not be optimal for a single research assay.