Achievement of targets and indicators for Sustainable Development Goals (SDGs) are of key interest to policy-makers worldwide. Sustainable development requires a holistic perspective of several dimensions of human society and addressing them together. With the increasing proliferation of Big Data, Machine Learning, and Artificial Intelligence, there is increasing interest in designing Policy Support Systems (PSS) for supporting policy formulation and decision-making. This paper formulates an architecture for a PSS, based on combining data from several sources. These datasets are subject to cleaning and semantic resolution and are then used as inputs to support the building of semantic models based on Bayesian Networks. A set of models built for different SDG sub-goals and indicators is used to create a “Policy Enunciator” in the form of data stories. Policy formulation is supported by modeling interventions and counter-factual reasoning on the models and assessing their impact on the data stories. Two kinds of impacts are observed: (a) downstream impacts that track expected outcomes from a given intervention, and (b) lateral impacts, that provide insights into possible side-effects of any given policy intervention.