There is no generally accepted methodology to determine the energy requirements of regrind circuits that is independent of the milling technology. The Bond ball mill work index test is not suitable for regrind circuit design due to the fineness of the feed materials and the change in log-log exponent when grinding fine enough, so regrind mill vendors have developed their own proprietary methodologies for estimating full scale performance of their specific mills. These disparate methods and the resulting scale-up procedures have led to estimates of required specific energy varying by 100–300 per cent even for similar mills in the same duty (Larson et al, 2011). To assess the variance between the different vendor methods, the authors compiled test results from Ausenco’s in-house database of regrind test work, where more than one stirred mill technology was considered. This paper presents the regrind test work methods used by different regrind mill vendors and compares typical test conditions based on the data set of stirred milling test work results. A review of the test work results is provided, highlighting issues that metallurgists face when analysing test work performed using different technologies such as coarse particle accumulation, differences in power measurement, variable viscosity and different media size. The objective was to identify a method that allows a direct assessment of test results. Various methods were tested and two are discussed: signature plots (grinding size versus specific energy) and the size specific energy (SSE) (specific energy versus additional fines production) (Mokken, 1978; Ballantyne and Giblett, 2019). Both models were found to produce useful insights for unit performance evaluation as well as identification of deviations from expected performance in the presence of test work issues. The signature plot is a simple model which can be a fitted to the data for estimating the grinding energy of a unit. However, there have been issues observed and discussed regarding this model which require consideration. The SSE model is an effective alternative to the signature plot model, as it avoids issues that stem from utilising the P80 to describe the bulk size distribution. Recommendations for regrind test work planning and analysis are provided to aid in explaining the observed variance between tests in future regrind test work endeavours.