Squeak and rattle are annoying sounds that are often regarded as failure indicators by car users. Geometric variation is a key contributor to the generation of squeak and rattle sounds. Optimization of the connection configuration in assemblies can be a provision to minimize this risk. However, the optimization process for large assemblies can be computationally expensive. The focus of this work is to propose a two-stage evolutionary optimization scheme to find the fittest connection configurations that minimize the risk for squeak and rattle. This was done by defining the objective functions as the measured variation and deviation in the rattle direction and the squeak plane. In the first stage, the location of the fasteners primarily contributing to the rattle direction measures is identified. In the second stage, fasteners primarily contributing to the squeak plane measures are added to the fittest configuration from phase one. It was assumed that the fasteners from the squeak group plane have a lower-order effect on the rattle direction measures, compared to the fasteners from the rattle direction group. This assumption was falsified for a set of simplified geometries. Also, a new uniform space filler algorithm was introduced to efficiently generate an inclusive and feasible starting population for the optimization process by incorporating the problem constraints in the algorithm. For two industrial cases, it was shown that by using the proposed two-stage optimization scheme, the variation and deviation measures in critical interfaces for squeak and rattle improved compared to the baseline results.