Capital expenditure on sampling and blending is often very difficult to justify. The costs are easy to estimate but the benefits likely to be obtained by constructing good sampling and blending facilities are generally difficult to estimate. For instance, improving blending of the ore feed to a smelter would improve throughput, yield and controllability ... but by how much? Similarly, better sampling would allow processing conditions to be dynamically adjusted and would improve extraction ... but by how much? Predicting benefits often requires the use of simulation models which carefully distinguish between true grades, predicted grades and observed grades. The parameters of these simulation models must often be estimated from rather inadequate information. These difficulties are illustrated using three scenarios.