Using meaningful reconciliation information to evaluate predictive models

Reconciliation of production information is critical to evaluating the effectiveness of predictive models that would allow for the optimization of mining operations. Whether mining open pit or underground, mine-to-mill reconciliations can be one of management's better tools for the performance of proper accounting. Such reconciliations can also be a very useful tool in evaluation and optimization. This paper proposes a set of criteria and objectives for a typical reconciliation program. This paper also presents a stepwise, logical approach to performing reconciliations and discusses the benefits associated maintaining good information. It also highlights some of the potential pitfalls involved and the methods used to avoid collecting misleading information. These issues are illustrated with an example of a model-to-mine-to-mill reconciliation program that is being implemented at an operating mine. The program involves multiple predictive models (long-term and short-term block models), feed from three open pits and one underground mine, in addition to intermediate stockpiling, for a total of 23 mill streams. This paper presents a summarized version of this reconciliation program and its initial results as implemented at Minera Michilla S.A.