Public transport disruptions force travellers to replan under uncertainty, increasing cognitive load and requiring personalised decision support. We frame this as an HCI problem centred on travellers’ needs and decision processes. Using a stated-preference discrete-choice experiment that manipulates weather, disruption certainty, travel direction, nearby amenities, and route attributes, we model route preferences with multinomial and mixed logit models. Results show substantial heterogeneity: travel time and crowdedness deter selection on average, while preferences for waiting, transfers, and mode switching vary widely between people. Context systematically shifts preferences; for example, weather, certainty, and commuting direction alter tolerance for transfers, waiting, and crowding. Predictive performance was modest, underscoring the limits of one-size-fits-all recommendations. Effective disruption support should adapt to context, communicate uncertainty, and leverage environmental affordances to reduce traveller burden.