The balance shaft is a countermeasure to reduce the 1st order excitation caused by the crankshaft in inline three-cylinder engine. Its design consists of two main variables, balance mass and angle. These variables need to be tuned so the engine pitch and yaw moments are minimized, which can be achieved by solving a deterministic design optimization problem. However, due to manufacturing tolerances of connecting rods, pistons, and balance shaft itself, the actual vibration level of the optimized balance shaft could fluctuate, and often becomes sub-optimal and even unacceptable. One way of addressing the influence of tolerances on vibration is to conduct a Taguchi parameter design, to pick the optimal settings of balance mass and angle, so the final designs are the least subject to the influence of manufacturing tolerances. Another approach of tackling the robust design issue is to use Robust Design Optimization (RDO), empowered by optimization algorithms, to search for the best nominal values of balance mass and angle. Furthermore, a Reverse-RDO can achieve designs with similar performance as the solutions of RDO but with much looser tolerance values. In this paper, Taguchi parameter design approach is compared with RDO, mainly in terms of overall complexity of procedures, turnaround time, optimal solution quality and efficiency, and easiness of being generalized to other engineering problems. Then the Reverse-RDO is conducted to find alternative optimal designs with much looser manufacturing tolerances. This paper concludes the Multi Objective Robustness Design Optimization (MORDO) and Reverse-MORDO, augmented by Response Surface Models, is preferred and recommended to tackle robustness in competitive product designs.