Soil sampling has long been advocated as a means of improving the efficiency of fertility management decisions by better matching fertilizer applications with crop nutrient requirements and nutrient availability. Advances in mapping and sensing technologies have renewed interest in soil sampling as a means of moving to variable rate technologies (VRT) whereby a farmer varies fertilizer applications across space and/or time. This paper develops a framework for determining the optimal number of soil samples when applying nitrogen fertilizer under a variable rate program.