AbstractUnder pressure for cost reduction, a semi‐automated grey iron foundry was experimenting with alternative supply sources of the moulding sand additives, bentonite and coal dust. During a period when increasing quantities of possible replacement materials were being blended with current supplies it was noticed that the previously high or erratic level of rejects in a small group of products was progressively improving. In the absence of a real understanding of the reason for the improvement and in an endeavour to ensure that adequate safeguards were in place to avoid any relapse, a statistical examination was undertaken of the green sand and metal subsystems.The behaviour of the variables in these subsystems was analysed for a datum period before and after the decision to change the sources of supply. Using existing process records and applying appropriate statistical techniques a clear indication was obtained that factors outside those of bentonite and coal dust were also involved in the improvement.The use of factorial analysis demonstrated the relative significance of the sand and metal variables and the existence of any interactions. The use of box and whisker plots gave a good indication of the desirable control limits for the key variables. Analysis of the follow‐up period revealed a reduction of 90 per cent in the level of rejects for one high volume l