Methods for identifying sites with potential for preventing traffic fatalities and injuries have been developed for vehicle-vehicle collisions. This study was funded by the California Department of Transportation (Caltrans) to develop methods for identifying sites where there is potential for significant reductions in pedestrian and bicyclist injury. Data from 1998-2007 from a 16.5-mile section of San Pablo Avenue (SR 123) in the San Francisco East Bay was used as a study area. Several approaches for identifying sites with high potential for reducing pedestrian and bicyclist injury were evaluated and compared, a framework was developed for conducting benefit-cost analyses, and a prototype was developed for a training protocol for conducting analyses of pedestrian and bicyclist safety in a corridor or network. The basic principle followed is that sites with the most potential for reducing injury are those sites where the most injuries can be prevented per dollar spent. Everything else being equal, these sites are the ones with the highest expected number of injuries if nothing is done. Prior history is typically used to make this estimate, but this may not be sufficient, especially if the underlying rates are low. Several approaches to developing statistically stable estimates are explained and compared: (i) extend the number years for both the baseline and follow-up periods, (ii) expand the size of the target sites considered, and (iii)apply Bayesian methods to include a modeled estimate of risk in the calculation. Strengths and weaknesses of each of these approaches are discussed with illustrations from the study area.