Safety is a key requirement in physical human- robot interaction. In particular, the influence of robot dynamic properties on human injury probability has to be understood well. Then, potential harm can be minimized already at an early stage of the robot design process. In this work, we propose the safety map concept, a map that captures the robot inherent safety properties and human injury occurrence in a unified way from both a global, and a task-dependent perspective. This makes it a novel, powerful, and convenient tool to quantitatively analyze the safety performance of a certain robot design. We elaborate the concept, describe how to process human and robot data towards the safety map representation and show how it can be integrated into the safety assessment and (automatic) robot design workflow.