*This is an abstract only. No full paper is available for this abstract.* Investment decisions in mining invariably require an assessment of confidence related to the underlying mineral resource estimates. Mining companies generally address this by requiring projects to adhere to standardised resource confidence profiles at different phases of project development. In practice, this is done by considering a distribution of Measured, Indicated and Inferred Resources through the production plan. Unfortunately, resource categories are generally defined on the basis of drill spacing without considering appropriate measures of uncertainty related to the estimates.As accounting for uncertainty requires its accurate quantification, the technical ability to model uncertainty quantitatively with as much accuracy as possible is of paramount importance. The current industry practice to characterise geological uncertainty consists of generating multiple realisations of the deposit that honour the data at their location and reproduce first- and second-order statistics, as inferred from the available data. This presentation presents a case study, developed with a real mined out deposit, showing that the ability to account for high-order statistics can improve the accuracy of uncertainty assessment and therefore render models that can better support investment decisions. The results of the case study also highlight the shortcomings of using traditional resource classification to support investment decisions.CITATION:Godoy, M and Meagher, C, 2014. Improving the accuracy of uncertainty models through high-order conditioning, in Proceedings Orebody Modelling and Strategic Mine Planning Symposium 2014, pp 377-378 (The Australasian Institute of Mining and Metallurgy: Melbourne).