Transit operations control, the task of implementing the operations plan in daily operations on a metro line, plays a key role in service delivery because it determines the quality of the service experienced by passengers. Yet, it is one of the most poorly understood aspects of rail transit operations. Faced with a disruption or infeasibility, dispatchers typically choose between several response strategies. However, to date, it has been very difficult to evaluate the positive and negative effects of individual control strategies with respect to operations and passenger travel times under real-world conditions. This paper proposes a framework for the study of rail operations control decisions that integrates automatically collected service and passenger demand data, which are increasingly available and accessible to transit agencies. The framework supports a multiperspective analysis methodology that can inform operational policies and plans, and help operations control decision-makers choose the most appropriate strategies to manage service. By using automatically collected data, taking into consideration the operations control decision environment, and acknowledging that the reliability of the resulting service depends on many factors endogenous to it, this paper takes a distinctly different approach from previous studies, which have relied heavily on modeling, assumed simple operating contexts, and did not consider the full range of available data. Two real-world applications ofThe research presented in this paper was carried out when the corresponding author was a student at the Massachusetts Institute of Technology. the framework, where control decisions are evaluated in terms of their operational and passenger impacts, are presented. The methodology is found to be versatile and valuable in providing insights that could not have been gained otherwise. Although the framework is applied to the London Underground, its logic, structure, and procedures are applicable and transferable to other metro systems recognizing that certain specifics would need to be tailored to the available data.