As quantum systems with increasing numbers of qubits emerge, new methods of characterizing and mitigating noise sources are required in the multi-qubit regime. While methods like randomized benchmarking (RB) and its variations are widely used in single and few-qubit systems to characterize coherent and incoherent noise sources, multi-qubit RB quickly becomes prohibitive at current coherence and control limits. In addition, it has been shown in recent experiments with three qubit RB that single and two-qubit RB fail to capture error sources present in larger qubit systems. On the other end of the spectrum, system-level metrics like quantum volume and other methods exist, but lump coherent and incoherent errors together in their output. We present here a method, based on the output of random quantum circuits, capable of quantitatively discriminating coherent noise from incoherent noise in multi-qubit systems. *This work was supported by ARO under Contract No. W911NF-14-1-012.