The geometrical quality of a welded assembly is to some extent depending part positions before welding. Here, a design of experiment is set up in order to investigate this relation using physical tests in a controlled environment. Based on the experimental results it can be concluded that the influence of part position before welding is significant for geometrical deviation after welding. Furthermore, a working procedure for a completely virtual geometry assurance process for welded assemblies is outlined. In this process, part variations, assembly fixture variations and welding induced variations are important inputs when predicting the capability of the final assembly.