When serious material flow disruptions occur in supply networks, complex interdependencies can lead to unpredictable effects which propagate through the entire network and endanger the existence of participating enterprises. Practitioners seek risk analysis methods that quantify risk propagation in the network as realistically and extensively as possible and present the results with sufficiently clear key figures and diagrams. Existing approaches, however, reveal their weaknesses precisely in these areas. While existing Bayesian network approaches only model the results of forward propagative effects through subjectively estimated and assumed conditional probabilities, current Agent-based Models provide insight into the dynamic interactions for specific research purposes, but fail to present the results of propagations concisely and to consider the network in total. This paper highlights the synergy effects of a combination of Agent-based Modeling and Bayesian networks not previously considered in the literature. A risk analysis methodology is presented that captures the complex interrelationships of various risk scenarios in disrupted supply networks. It offers both forward and backward risk propagations and direct and indirect risk consequences in a way that is easily digestible by management. For this purpose, three novel risk metrics for Bayesian networks are presented that quantify the exact risk propagation pattern in the network and provide a more accurate planning basis for mitigation strategies. Our methodology is applied to an excerpt of a supply network in the consumer goods industry and illustrates the approach and its benefits.