A unified framework for rich routing problems with stochastic demands

•Proposes a modeling approach capturing dynamic probabilistic information in the objective function that is unaffected by the curse of dimensionality.•Models explicitly undesirable events such as stock-outs and route failures, their probabilities, costs, and associated recourse actions.•Integrates real-world demand forecasting techniques relaxing typical assumptions from the literature.•Maintains computational tractability through the ability to preprocess the probabilistic information for a general inventory policy.•Demonstrates the performance of the approach on rich instances derived from real data for two conceptually different routing problems.