This presentation is a survey of subsurface machine learning concepts that have been formulated for unconventional asset development, described in the literature, and subsequently patented. Operators that utilize similar subsurface machine learning workflows and other data modelling techniques enjoy a competitive advantage at optimizing the development of unconventional plays. For example, this advantage has allowed Chevron to book 3.3 MMBOE net resource (P4-P6) adds while saving an estimated $500MM in exploration well costs from 2020-21. The value benefits increase as subsurface machine learning is applied beyond exploration and into development and enhanced oil recovery activities. These workflows typically guide practitioners from data gathering to geospatial assembly, quality control and ingestion, then on through machine-learning feature selection, modeling, validation, and acceptance for results reporting. The ultimate products of these workflows can be visualized in both map or log (depth) space to help identify key regions for well optimization or landing zones, respectfully. Machine-learned algorithms that forecast production from engineering, development, reservoir, and geologic predictors in unconventional plays can be further interpreted to derive optimization practices for well development and operations. The geostatistical, spatial, and non-parametric statistical approaches used in conjunction with subsurface machine learning workflows are discussed for their utility to understand the geographic extent of well-trained production forecasts from data models, including applications for testing the viability of type-curve neighborhoods. These data driven workflows also ultimately serve to characterize often multidimensional trends and non-linear interactions between key predictors that is useful for building a deeper understanding of the critical physical processes active in the subsurface that influence unconventional well productivity and profitability.