A variety of vehicle controls, from active safety systems to power management algorithms, can greatly benefit from accurate, reliable, and robust real-time estimates of vehicle mass and road grade. This paper develops a parallel mass and grade (PMG) estimation scheme and presents the results of a study investigating its accuracy and robustness in the presence of various noise factors. An estimate of road grade is calculated by comparing the acceleration as measured by an on-board longitudinal accelerometer with that obtained by differentiation of the undriven wheel speeds. Mass is independently estimated by means of a longitudinal dynamics model and a recursive least squares (RLS) algorithm using the longitudinal accelerometer to isolate grade effects. To account for the influences of acceleration-induced vehicle pitching on PMG estimation accuracy, a correction factor is developed from controlled tests under a wide range of throttle levels. The estimation approach is applied to data collected while driving on public roads under a variety of driving conditions, replicating several noise factors associated with daily driving. Thresholds are developed to isolate driving events that are likely to result in quick and accurate mass estimates, and an averaging filter is then applied to these converged values. The algorithm is shown to rapidly deliver estimates within 3% of true vehicle mass.