The Lower Cretaceous McMurray Formation of Alberta, Canada, contains vast quantities of bitumen. Oil sands hosting this bitumen were deposited and reworked by tidal and fluvial currents, resulting in heterogeneous sediment distributions comprising cross-stratified and massive sands, breccias, and low permeability sand- to mud-dominated inclined heterolithic stratification (IHS). Conventional reservoir characterization workflows suggest that reservoir permeability correlates most closely to volume of shale (vShale). One aspect of such workflows is derivation of an empirical correlation between permeability obtained from core analyses and calculated vSh from wireline logs. While there may be numerous direct measurements from core samples, uncertainty in oil sands reservoir permeability is known to result from preferential sampling of clean sands versus muddy IHS and from scale differences between core sample and well log resolution. This uncertainty is particularly significant in facies with permeability < 500 millidarcy. We integrated multi-scale exploration data from a producing McMurray Formation reservoir to demonstrate the impact of centimeter-scale reservoir architecture on permeability. The study uses high-resolution 3D models of the near-wellbore region to characterize observed variability in percentage, orientation and dimension of mud laminae, breccia clasts and trace fossils. We created multiple realizations to capture the variability in model simulation parameters. High-resolution models of several million cells were upscaled at the resolution of the vSh log using a flow simulation-based numerical method. Upscaling results are estimates of effective directional permeability (kx, ky, kz). The flow-based permeability results better define the interrelationship of permeability and vShale within a reservoir. Our results demonstrate a modeling workflow that provides more precise permeability estimates at wireline log resolution. This workflow has the potential to improve history matching in flow simulations and improvements in well placement and performance forecasting.