Currently there is considerable interest in developing fuel-efficient engine oils. However, this development is hampered by the fact that the variability of automobile fuel economy tests is often as large as the fuel economy benefit being measured. It is essential, therefore, that tests of fuel-efficient engine oils be statistically designed and properly analyzed. In this paper, we discuss the strengths and weaknesses of various designs and describe the proper ways of analyzing such data. In addition, general guidelines are presented for determining the test fleet size and amount of replicate testing required.