The design of an innovative rear-end collision-warning system was evaluated for effectiveness. The crash scenario involves a lead vehicle not moving (LVNM) in one lane of a straight, dry, paved arterial road and a following vehicle approaching in the same lane. The LVNM has a rear-facing sensor and is equipped with the rear crash-warning system, which allows the LVNM to flash its brake lights or its center high-mounted stop lamp, warning the following vehicle that it is approaching too rapidly. Because this problem is complex, the scope was narrowed, and it was assumed that the driver of the following vehicle always detected the warning after a response time lag and then applied hard braking. The warning algorithm (i.e., selecting the most appropriate warning distance for each approaching vehicle speed) was designed based on tradeoffs to maximize the capability of preventing crashes, reduce the frequency of nuisance alarms, and minimize the severity of crashes. The overall measures of effectiveness of the warning design were then evaluated for a vehicle speed distribution that represented a suburban arterial road. The findings suggest that the warning system should be very effective in preventing crashes. The expected number of nuisance alarms was small, and for the very small percentage of vehicles that would crash, the expected crash severity was negligible. Experimental studies and field operational tests would be required to obtain more accurate numerical values for the design parameters.