The megacity of São Paulo, Brazil recently faced a severe three-year drought (2013- 2015) that highlighted the challenges of growing water demands, unpredictability of future supply, and proper communication with the public during crisis. Given the poor outcomes of the recent drought, the question remains as to how the drought can be used to inform and improve future drought preparedness. This study develops a seven-step framework using simulation and a computational search to identify promising drought plans based on the historic record. These promising plans are then tested under alternative states of the world to explore their robustness and identify regions of vulnerability. While the resulting drought plans are dependent on many assumptions regarding system operations, performance preferences, and future conditions, the process can be used by managers with knowledge of the system in collaboration with stakeholders to identify drought plans that will result in better outcomes.