Practical solution of a bulk sampling problem follows a stepwise procedure. Model hypothesis, experiment, test and development of a new hypothesis is vital to the solution. Two case histories demonstrate the procedure.

The first describes a study of ore received in boxcars. Particle size and chemical composition were examined to develop a car acceptance plan. Fifteen samples from each of two cars from two vendors were taken. The variability of samples within cars, among cars from the same vendor, and between vendors was established.

The second describes a study of two thief sampling methods for approving finished hoppers of alumina abrasive. The sampling and testing variation was evaluated. Testing errors introduced by sifting machines, sieves and operators were appraised. Next, a mathematical model associated with the physical grain handling was postulated, tested by experiment, and denied. Another model was hypothesized from the data.