An experimental pilot plant was constructed, commissioned and operated at a major porphyry copper mine to understand the challenges of microwave infrared thermal (MW-IRT) sorting at scale and to compare batch laboratory performance with pilot-scale continuous sortability performance. A method was developed to define the 95% confidence intervals on pilot plant operating windows from experiments on 50-150 fragments performed in a laboratory based replica of the pilot scale microwave treatment system. It appeared that the laboratory testing methodology predicted the sortability of the ores fairly well. For the 11 ore types and three size classes (-76.2 + 50.8 mm, 50.8 + 25.4 mm and 25.4 + 12.7 mm) tested over 233 pilot plant experiments, approximately 42% of the better optimised pilot plant runs predicted copper recovery to within +/- 5% copper recovery and approximately 84% of the runs to within +/- 10%. These "figures were improved to approximately 50% predicted to within +/- 5% and approximately 90% to within +/- 10% if the -25.4 + 12.7 mm size class was omitted. It was demonstrated that laboratory testing better predicted pilot plant sorting performance and provided a narrower operating window when a larger sample size (>50 fragments) was considered due to improved representivity. It is, therefore, fully expected that better predictions would result from larger laboratory sample sizes than those tested during any future testing campaigns. To date, approximately 15,500 tonnes of ore has been processed through the pilot-scale test facility, generating significant engineering know-how and demonstrating MW-IRT sorting at a scale in the order of that required by the mining industry. (C) 2016 Elsevier Ltd. All rights reserved.