Transportation planners and engineers need better spatial estimates of walking and cycling traffic to assess exposure to hazards, evaluate infrastructure investments, and locate facilities. Facility-demand models are potentially useful for generating spatial estimates of traffic volumes. Few facility-demand models have explored trade-offs between fully specified (i.e., exploratory) models and reduced-form models easily applied in the field. Presented are facility-demand models based on peak period (4 to 6 p.m.) counts of pedestrian and bicycle traffic in Minneapolis, Minnesota. The count database (n = 954 observations; 471 locations) has sufficient spatial density (∼3 locations km-2) to develop spatially resolved models (i.e., ∼100-m resolution). The modeling approach employs a stepwise linear regression method allowing for varying the spatial scale of independent (land use and transportation) variables. Compared were fully specified (statistically optimal) models and supervised, reducedform models that included fewer variables based on theoretical validity. Reduced-form core models had modest goodness of fit (adjusted R2: ∼.5) and included independent variables with large (industrial area and population density) and small (bicycle facilities, retail area, open space, transit stops) spatial scales. Also developed were reduced-form, time-averaged models for a subset of count sites having multiple observations (n = 84). With the use of reduced-form models (independent variables ranged from four to nine among models), block-level traffic estimates were generated (n = 13,886). Results suggest that reduced-form models perform nearly as well as fully specified models and are easier to apply and interpret. This work could be extended by assessing model performance when estimates of annual average traffic are used in model building.