Bayesian Networks (BNs) are increasingly being used to model coastal processes. BNs are probabilistic graphical models that are able to represent complex physical systems with the benefits of very low computational cost, intrinsic handling of uncertainty and error, and explicit description of causation and relationships between variables within the system. BNs can be used for both predictive and diagnostic inference, and are particularly suitable for application to management tools, as they are explicit in uncertainty, give outputs in probability distributions, and are relatively straight-forward to integrate within existing risk analysis and decision-making frameworks. The general steps in developing a Bayesian Network are described. An example application to predict the degree of shoreline erosion caused by coastal storms is then presented, based on a data set spanning 10 years that includes 137 individual storm events. Located at Collaroy-Narrabeen Beach in Sydney, Australia, an optimised BN is shown to correctly predict the extent of storm erosion (‘extreme’, ‘mild’, ‘no-change’) 65% of the time when applied to unseen storm events. Sensitivity analysis shows that incident wave power and pre-storm shoreline position were the most sensitive parameters in the BN for predicting storm erosion at this site. The application of the BN to the June 2016 major ECL storm at Collaroy-Narrabeen shows how this simple approach could potentially be used in a real-world forecast application to inform emergency management decisions and assist with community preparedness. While possible, the development of a more complex BN with higher predictive skill requires more observation data than was available to this study, and is the topic of future work.