A Proactive Reconciliation System at the Cuiabá Mine: Quantification of Random Errors in Gold Grade Determination Across the Production Chain

Establishing a proactive reconciliation chain is essential for understanding the gold production system from the geological model to the final product. Monitoring this chain enables the identification of process deviations and sources of variation, providing a robust management tool for quality control. Although the major process steps are well known, the sources and magnitudes of measurement errors are often unknown. Quantifying errors at each node of the reconciliation system allows estimation of the statistical confidence of each dataset and its impact across the entire chain. This study presents the reconciliation chain implemented at the Cuiabá Mine, owned and operated by AngloGold Ashanti, identifies the principal random errors, as defined by geostatistics and Pierre Gy’s Theory of Sampling (TOS), and quantifies the statistical reliability of the data at key stages. This work provides one of the first quantitative integrations of TOS-based sampling errors and geostatistical uncertainty across a full industrial reconciliation chain. Results show that sampling-related uncertainties are dominated by the Fundamental Sampling Error, reaching 24.84% in channel samples and 22.24% in plant feed samples, whereas the Heterogeneity Fluctuation Error contributes 2.96%, and the geological model uncertainty reaches 5.80%. These results highlight the critical influence of sampling protocols on reconciliation performance and demonstrate the value of integrating statistical error quantification into mine production management and decision-making.