*This is an abstract only. No full paper is available for this abstract.* Geometallurgy is a cross-disciplinary activity with the primary aim of characterising ore, in terms of critical processing performance, in a quantified and spatially constrained manner. A key objective of geometallurgy is the economic optimisation of mining through reduction of technical risk. Geological variability lies at the heart of geometallurgy. Definition of geological variability can present many challenges, for example, mineralogy. Mineralogy is a fundamental processing characteristic for a given rock mass throughout the mining chain, so understanding the mineralogy of the bulk rock, including both the gangue and ore component (in our example, Cu-bearing minerals), is critical when making predictions on processing characteristics. Definition of geological variability can be improved through development of mathematical modelling and machine learning techniques for determining quantitative mineralogy, using commonly available chemical assay and geological logging data.Calculating modal mineralogy from chemical assay provides a fast and cost-effective alternative method to estimating bulk mineralogy. To demonstrate this we provide an example of the calculation of complex mineralogy for the Productora Cu-Au-Mo deposit, Chile, using 133 963 inductively coupled plasma atomic emission spectroscopy multielement analyses and a training set of 625 quantitative X-ray diffraction analyses. These data sets were used to constrain the ore and alteration mineralogy of the hypogene, supergene and transitional ore at Productora.We also present a new approach to defining oxide, transitional and sulfide material on a sample by sample basis at the same deposit in order to improve the spatial definition of variable Cu speciation. A simple Cu species classification scheme based on sequential leach data and S per cent has been devised to classify oxide, transitional and sulfide Cu, and also to account for non-recoverable Cu, ie non-sulfide Cu which is insoluble in weak acids. Through machine learning, a proxy for this Cu species classification scheme was developed based on: depth, Ca per cent, Cu per cent, Fe per cent, K per cent, Mn ppm, S per cent and Ln(Cu/S) plus logged regolith class, thus allowing classification to be extended to areas of the deposit, where no sequential leach data are available.Results from these two novel approaches to predicting mineralogy and Cu species indicate that geochemistry and geochemical proxies can be used successfully, resulting in a high density of deposit wide data and increased orebody knowledge at low cost. This new data has been used to generate deposit-wide 3D models for mineralogy, including total quartz + feldspar per cent and total pyrite per cent, and Cu species classification that can be used in geometallurgy studies.CITATION:Escolme, A J, Berry, R F, Hunt, J and Cooke, D R, 2017. Novel approaches to geometallurgy using geochemistry - same data, new tricks!, in Proceedings Tenth International Mining Geology Conference 2017, pp 453-454 (The Australasian Institute of Mining and Metallurgy: Melbourne).