Geometallurgy is an inter-disciplinary approach to improving decision-making and optimising production in the mining industry by integrating geological and metallurgical data and knowledge. Improvements come from a better understanding of the resources and their metallurgical performance and by optimising over the complete mining value chain. The primary-response rock property framework is an approach to geometallurgy in which primary properties such as grades, lithology, alteration and mineralogy are proxies of important metallurgical responses; for example, recovery, comminution and energy use. Using this framework, primary variables are used to predict key response variables. Primary rock properties, typically geological, geotechnical and structural, are relatively abundant compared with metallurgical response properties, making the integration of the latter more difficult. Moreover, the relationship between geometallurgical input variables and their processing responses is usually complex and the response variables are often non-additive which complicates the prediction process. Consequently, in many cases, traditional multi-linear regression models (MLR) are not good predictors and non-linear models may be a better alternative. Projection Pursuit is a powerful exploratory statistics technique, in which promising directions are found and data are projected on to these directions to reveal underlying relationships. In particular, Projection Pursuit Regression (PPR) finds several directions in which the variance explained by the projected data is maximised, which enables more accurate predictions to be made. In this paper we present a case study in which six geometallurgical response variables are modelled by PPR of primary properties. The results from the proposed PPR models show a significant improvement over those from MLR models. In addition, as the quantification of uncertainty is a key aspect of risk management, the models were bootstrapped to generate a distribution of feasible scenarios. Our results show that PPR is a robust technique for modelling geometallurgical response variables and their uncertainty.