The data improvement, as achieved by optimising a mass balance of inconsistent experimental data, was quantified statistically using a Monte Carlo technique. It was found that an increase in the number of levels in the mass balance problem produced further increases in the data improvement. However, the improvement was lost when, for example, the solids split value was numerically in the vicinity of the flow split value. Furthermore, accuracy in the data at one level of the mass balance was found to be transferred to only one adjacent level of the mass balance. For example, accurate assay data produced improvement in the solids and assay split values, but not in the pow split values. Similarly, accurate pulp density data produced improvement in the flow and solids split values, but not in the assay split values. And, accurate flow data produced no significant improvement in the solids split or assay split values. In general, the transfer of improvement only occurred in the direction of lower level data (eg., from the assay to the solids level, or from the solids to the pow level).