Geometallurgy is at the core of life-of-mine value chain optimisation, with the aim of integrating geoscientific disciplines along with mining engineering and minerals processing. The objective is to link comprehensive geological, geochemical, mineralogical and geotechnical information with metallurgical and mining variability - based on spatially distributed samples. The spatial coverage is a crucial element in this process. Geometallurgy samples are used for metallurgical testing in the service of plant and process design and optimisation. To avoid discrepancies between the expected and actual process performance, geometallurgical test work must be based on representative samples collected and processed in compliance with the Theory of Sampling (TOS). However, even if samples are initially collected to populate a multivariate block model, most of TOS’ recommendations for estimating sampling protocols and sample representativeness is univariate. While the univariate approach is sufficient when a sample must be representative for one property only e.g., for grade estimation, it fails to properly qualify representativeness of a sample which must be representative for multiple properties such as for geometallurgical purposes. Indeed, a geometallurgy sample is considered representative sensu stricto only if its metallurgical behaviour is representative of that of the full zone of the orebody it represents. This can only be achieved if-and-when geo-metallurgical samples are representative for the full set of ore properties that influence process performance. The critical success factor of multivariate representativeness can be assessed using multivariate approaches, such as the multi-variogram, which allow us to summarise the global variability of multiple properties into a single characteristic function. This approach could be optimised by using downstream results from geo-metallurgical process modelling, to select or weight, the individual property contributions to the multi-variogram according to their importance, thereby allowing to optimise a specific geometallurgical sampling procedure in terms of sampling mode, sampling frequency and the number of increments involved according to the overall process performance.