During the vehicle development process in the premium and luxury automotive segments an important task is the refinement of noise, vibration and harshness. Along with other attributes such as styling, drivability and vehicle dynamics it strongly influences the overall perception of the vehicle. At the same time the automotive manufacturers need to release more products faster to the market using shorter vehicle development time with reduced cost. Altogether this has increased the use of virtual models and decreased the number of test vehicles in the programs. When assembling vehicles in production there will be a natural variation, which will result in a spread in the attribute performances. When shifting towards virtual models and reduced numbers of physical test vehicles there is a higher risk that the variations in production will be neglected, leading to more customer complaints. The question is if the production variability could be predicted before start of production by taking the component variability and assembly process into account early on in the vehicle development phase. This paper demonstrates a methodology how this could be performed for an all-wheel drive system. The all-wheel drive system imbalance can give vibrations in the steering wheel and seats and also induce a low frequency noise. This is a typical case where a robust solution needs to be considered. The imbalance could be reduced by balancing components and/or the assembly in the production. However, this will increase the cost and therefore the benefit of these solutions must be quantified before they are introduced. The paper will demonstrate that using a Design For Six Sigma, DFSS, approach the vehicle robustness can be predicted. Before their introduction, the effects of two proposals for reducing imbalance of the all-wheel drive system have been quantified. The results have also been verified by testing 30 cars in running production.