Dynamic Bayesian network modelling of supply chain resilience: learning from multiple cases

While there is growing interest in supply chain resilience, conceptualization of the main constructs is discussed at a high level in the literature. We aim to learn more about what supply chain resilience means through modelling, specifically using Dynamic Bayesian Networks (DBNs). DBNs are directed acyclic graphs for reasoning under uncertainty through time. A DBN is capable of using partial knowledge about one variable to update the uncertainty about other variables in the model. In principle, DBNs present a possible model class for analysing resilience because they can capture the dynamic uncertain behaviour of a supply chain due to the effects of potential hazards. Between 2013-15 we have conducted multiple cases for four distinct manufacturing and retail supply chains with focal companies based in the UK, Canada and Malaysia. We present the characteristics of each chain and the insights gained through application of our DBN modelling protocol. Our analysis has generated better understanding of the relative importance of resilience enablers reported in the literature and the propagation of uncertainties on predictions of resilience under different risk scenarios. Reflecting on our modelling methodology has generated insights into the benefits and limitations of DBNs for modelling supply chain resilience.