We present a method to develop socially-constructed metrics for ascertaining agile software development quality. Canonical action research (CAR) is our mode of inquiry, conducted in a key European player of healthcare information systems. The result is a set of meaningful metrics that are built according to three interrelated dimensions: (1) evidence from practice; (2) stakeholders expectations; and (3) stakeholders evaluation. Our contribution suggests simple artefacts to create socially-constructed metrics and the main guidelines to use them. Agile teams struggle with quality measurement, often supported by a plethora of metrics that do not adhere to rapidly changing project environments. We argue that socially-constructed metrics can address this problem, offering a contextualized perspective of quality that can improve tacit knowledge transfer; critical reflection about quality; and effective support in daily meetings, retrospectives, and audits. Moreover, it suggests a participative approach for continuous improvement in agile software development.