Diagnosing faulty performance deviations of electro-mechanical systems can be difficult, given the multitude of components and features which could contribute as root causes. Yet this is often a problem in manufacturing, where only some of the units built do not meet performance requirements only some of the time. In this context, product and process simulation studies can aid in diagnosis. This paper aims to develop a practical workflow and toolchain to guide use of uncertainty quantification and sensitivity analysis methods for root cause analysis of manufacturing processes. This approach offers more rapid diagnosis than the typical approach using some form of iterative experimentation such as Red-X, fault tree analysis and when in high volume production, statistical analysis and potentially machine learning. Here, part processes, features and assembly deviations are used as inputs to product performance simulation to understand their detrimental impact. The large set of possible process inputs can be systematically varied and contributions to system performance deviation computed. To do this simply using uncertainty quantification and sensitivity analysis is impractical, as the problem is too large. Rather, a sequential refinement workflow is developed to define the problem and possible causes, understand ability model causes, screen causal variables, and then apply quasi-Monte-Carlo uncertainty quantification sampling and global sensitivity analysis. This provides computational guidance to ascertain which manufacturing process inputs are more likely causes of performance deviations on manufactured units.