The authors present and test a pathfinder screening algorithm (PSA) that can be used to identify which logging data is useful for ML modelling. They have applied the technology on three different deposit types: (1) gold modelling at a Western Australian orogenic lode gold deposit; (2) copper modelling at a Chilean manto-type iron oxide copper gold deposit; and (3) sulfur modelling at a Papua New Guinean low sulfidation epithermal gold deposit. In addition, the authors also test different methods of integrating geological logging into machine learning, quantify accuracy improvements, and establish a set of best practices.