The Radar-Based Berthing Aid Dataset (R-BAD) is an openly accessible dataset designed to support the development of safe and intelligent ship berthing systems. It was collected onboard an operational Ro-Ro/Passenger ferry and contains over 69 hours of synchronized FMCW radar point clouds and video recordings across 13 ports and multiple operational scenarios, including arrivals, departures, port idle, and cruising maneuvers. The dataset is organized into two parts: Raw Aggregated Frames Data: structured radar detections (CSV) paired with synchronized video (MP4). Labelled Buffers Data: annotated radar detections grouped into buffers, suitable for machine/deep learning applications in clustering, tracking, and classification. R-BAD provides a unique benchmark for research on short-range maritime obstacle detection, radar-based perception, and autonomous berthing systems. By making this dataset freely available, we aim to foster innovation in maritime safety, intelligent transportation, and autonomous navigation.