This paper illustrates the use of variography and moving average plots as an aid to analysing variations in the process that can be used to direct the improvement efforts. The hydrometallurgical processes used in manganese metal production consist of sequential unit operations that include milling and reduction of manganese ores. Leaching of the reduced ore, is followed by solid-liquid separation, residue disposal, solution purification and electrowinning of manganese metal completes the process. The nearly 250 integrated process control loops rely on reports of pH, flow pressure, level, temperature and chemical analyses._x000D_
Metallurgical accounting, reconciliation and process control is achieved through monitoring of the chemical and mass balances within the manganese reduction plant, requiring approximately 50 000 analyses of solids, powders and liquids each month. The largest numbers of samples are taken in the cell house._x000D_
Changes and variations monitored by sample statistics ensure that the manganese products comply with predetermined specification limits. Optimising the information-value of the approximately 50 000 samples in the manganese plant is carried out in tandem with ongoing process optimisation. Variographic analysis of daily sampling data suggests that the cyclical behaviour evident in the moving average plots is only a reflection of random residual variability and not a reflection of true underlying variability related to cyclical phenomena. Low resolution in the variogram of daily sampling data masks the short-range variability that emerges when hourly data are analysed. Hourly samples collected over a 39-day period (giving 941 sample data), helped identify short-range cycles in the process. The relative distribution of variability for daily sample intervals, indicate that 57 per cent arises from V[0], 18 per cent from V[process] and 16 per cent from random residual variation. V[0] includes all the uncontrolled sampling error arising from a poor sampling protocol, the fundamental error (FE), the grouping and segregation errors (GSE), the delimitation error (DE), the extraction error (EE), etc. This confirms the suspicions regarding the suboptimality of the sampling protocol and the way it is implemented._x000D_
The absence of long-term trends in the leach suggests that short-term variation (less than one day) dominates the process. While the control charts indicate that the overall processes are maintained above the lower specification limits and the target values, there is nevertheless a considerable amount of variability in the system that overhangs the processes and threatens to impact them negatively. Although the variability in the leach is above the prescribed limits, the volatility is a matter of concern because it seems to be uncontrolled. The comparison of the various sources of variability in data collected on an hourly basis suggests that the short-term random variability (V[0]) accounts for almost 50 per cent of the variability in the process._x000D_
The main source of this variability is due to errors in the sampling protocol and is probably related to the size of the fundamental error (FE), the grouping and segregation errors (GSE), the delimitation error (DE), the extraction error (EE), etc. Remediation of FE and GSE will probably lead to a reduction in V[0], whereas V[process] could be mitigated by reducing the biases associated with DE and EE in particular.