Deriving significant experiment-based conclusions on mechanical processing of mixed solid waste is challenging: the input material cannot be downscaled in a way that enables drawing transferable conclusions from lab-scale experiments. Hence experiments need to be conducted in industry-scale, using real waste. Besides the enormous resulting experimental efforts and costs, which economically limit the number of experimental runs, identifying and quantifying significant effects is complicated by the distortion of the data introduced by the waste's variability. The distortion is particularly high for cases where sampling is necessary and in experiments where material cannot be re-used from one run to the next. In the latter case, inter-experimental differences of the waste add to the distortion of the data. In this work, a systematic approach for deriving representative and significant results at the minimum possible effort is described and evaluated, based on the method of Design of Experiments. It is applied to a 32 runs D-optimal industry-scale coarse-shredding experiment with mixed commercial solid waste, based on a reduced cubic design model, examining the influence of the gap width, shaft rotation speed, and cutting tool geometry on the throughput behavior and energy demand. The resulting models are highly significant (model p-values < 0.0001), proving the ability to extract reliable information from industry-scale waste processing experiments. Concerning commercial waste shredding, the models provide new insights into process behavior, for example, the quadratic dependence of the mass flow on the shaft rotation speed, with the highest hourly mass flows at 84% of the maximum shaft rotation speed.