In the last 100 years, statistical sampling theory for particulate materials has developed to a stage where sampling systems can be confidently designed to provide assay results of known and controlled variance for a chosen time period. The two independent sources of sampling variance, namely intrinsic and distributional heterogeneity are defined and explained quantitatively. Intrinsic heterogeneity is characterised by the sampling constant for the material and distributional heterogeneity is quantified by the process steam variogram or covariance function. The means by which sampling systems can be designed to an accuracy specification is described.