Shallow foundations derive their support from near surface soils and are extensively used to support structures of all sizes. Existing methods to estimate bearing capacity invoke an assumed failure surface developed within rigid-perfectly plastic soil, and require various semi-empirical modifications to the failure surface to account for geometrical effects. However, the global accuracy and uncertainty associated with the general bearing capacity equation and its modifications has not been sufficiently characterized with respect to observed full-scale footing load tests on plastic fine-grained soils. To assess the global accuracy and uncertainty in the general bearing capacity formula, a load test database consisting of 30 full-scale footings was developed from loading tests reported in the literature. The computed bearing capacity was compared to the capacity extrapolated from the load displacement curves in the database, then statistically characterized using the bias, or the ratio of the extrapolated and computed capacities. On average, the general bearing capacity formula under-predicted the extrapolated bearing capacity with a mean bias of 1.25 and exhibited a moderate to significant amount of variability (i.e., COV = 37%). In order to provide a reliability-based ultimate limit state model, varying levels of uncertainty associated with the undrained shear strength were used in the development of resistance factors for use with AASHTO load statistics. The use of risk-informed design approaches, such as those presented herein, should help to increase more efficient, and therefore sustainable, engineering practices.