Subsurface uncertainty is one of the main challenges in using reservoir models to predict field performance for development and depletion planning purposes. The importance of reliable characterization of subsurface uncertainty and its impact on reservoir performance predictions is increasingly recognized as essential to robust decision making in the upstream industry, which is especially true for large projects in complex geologic settings. However, despite recent advances in reservoir modeling and simulation, reliable quantification of the impact of subsurface uncertainty remains difficult in practice. Many factors lead to this state of affair; technically, a fundamental difficulty is that reservoir heterogeneity at multiple scales may have strong effect on fluid flows. This lecture presents an analysis of the challenge and possible resolutions. Indeed, relying on computing power alone may not address the challenge. Instead, we must look at reservoir modeling and performance prediction holistically, from modeling objectives to appropriate techniques of incorporating reservoir heterogeneity into the models. We present a goal-driven and data-driven approach for reservoir modeling with the theoretical reasoning and numerical evidence behind them, including real field examples. The one idea that participants of this lecture should take away is that appropriate parameterization of multi-scale reservoir heterogeneity that is tailored to the business questions at hand and available data are essential for addressing the challenge of subsurface uncertainty.