Many important technical innovations occur through changes to existing system architectures. To manage the balance between performance gains by the innovation and the risk of change, companies estimate the degree of architectural change an innovation option could cause due to change propagation throughout the entire system. To do so, they must evaluate the innovation options for their integration cost given the present system architecture. This article presents a new algorithm and metrics based upon eigenvector rotations of the architectural connectivity matrix to assess the sensitivity of a system architecture to introduced innovations, modelled as perturbations on the system. The article presents studies of the impact of changes on synthetic system architectures to validate the method. The results show that there is no single architecture that is the most amenable to introduced innovation. Properties such as the density of existing connections and the number of changes that modify intra- or inter-module connections can introduce global effects that are not known in advance. Hierarchical modular system architectures tend to be relatively stable to introduced innovations and distributed changes to any architecture tends to cause the largest eigenvector rotations.