Early-stage mineral exploration, commonly involves surface soil, lag or chip sampling and analysis of elements. This has not changed significantly over the past decades: that is, digest a sample and, subsequently, analyse the solution, examine key target and pathfinder elements, then rinse and repeat at the next tenement package. as an industry, we need to change the way we explore to improve success, especially in covered terrains. The CSIRO research team has developed tools that fundamentally change the soil analysis and interpretation approach in australia with the development of UltraFine+® (soil analytical technique) and LandScape+ (a machine learning workflow) as outcomes of two major R&D projects with ~40 industry and government collaborators. In this presentation we highlight briefly the evolution of the analytical method and data processing that assist exploration for multiple commodities. More importantly, we focus on the machine learning analytics and the approach of using spatial data to generate landscape types (with dimensionality reduction and clustering methods). We highlight the new web-based software application that makes it easy for any explorer to generate a first-pass analysis and interpretation of their surface geochemistry for areas of up to 2000 km2. While landscape maps are available for some regions, using machine learning to derive landscapes using remotely-sensed spatial data allows this approach to be employed in most regions even where traditional map products are not available or only at a coarse resolution (as is often the case when explorers advance into true greenfields settings). These data driven landscapes can accelerate interpretation and reduce risk by understanding where cover is thicker, or surface sampling may be problematic. There are limitations to the unsupervised machine learning landscape models and these will be discussed along with the value of identifying coherent anomalies in landscape settings that are notoriously difficult to explore.