Modeling joint dynamics is the bottleneck for precise predictive models of machines. Bolted joints are especially relevant due to their common use in mechanical engineering. Stiffness and damping properties of bolted joints are highly influenced by contact parameters (geometry, surface treatment, preload, …). For high amplitude vibrations, when non-appropriate joint closure is present, even frictional effects can play a role. Substructuring techniques offer a lean solution for isolating dynamical or quasi-static components of joint dynamics. It may be used for an identification of linearized contact parameters for multiple degrees of freedom (dofs). The main limitation of the method is finding a robust workflow to guarantee a proper controllability and observability of the desired joint dynamics, as well as dealing with disturbance/noise in the measurements. In this contribution a robust procedure for the identification of a bolted joint using frequency-based substructuring is presented for a contact in an experimental scenario. The dynamics of the joint are isolated from the assembled system using different substructuring techniques treating the joint as a quasi-static or dynamic component. A simple physical model of the joint is parametrized from the experimental joint dynamics. A validation of the methodology is given by using the identified joint parameters on a modified assembled system. •Robust procedure for the multi-dof identification of bolted joints.•Theoretical and experimental comparison of quasi-static and dynamic decoupling in frequency domain.•Practical experimental application of the proposed joint identification.•Inverse substructuring seems to be a good fit for bolted connections.