If the outcomes of the recent COAG meeting are implemented, Australia will have a new set of benchmarks for its health system within a few months. This is a non-trivial task. Choice of benchmarks will, explicitly or implicitly, reflect a framework about how the health system works, what is important or to be valued and how the benchmarks are to be used. This article argues that the health system is dynamic and so benchmarks need to measure flows and interfaces rather than simply cross-sectional or static performance. And also argue that benchmarks need to be developed taking into account three perspectives: patient, clinician and funder. Each of these perspectives is critical and good performance from one perspective or on one dimension doesn't imply good performance on either (or both) of the others. The three perspectives (called here the dimensions patient assessed value, performance on clinical interventions and efficiency) can each be decomposed into a number of elements. For example, patient assessed value is influenced by timeliness, cost to the patient, the extent to which their expectations are met, the way they are treated and the extent to which there is continuity of care. The authors also argue that the way information is presented is important: cross sectional, dated measures provide much less information and are much less useful than approaches based on statistical process control. The latter also focuses attention on improvement and trends, encouraging action rather than simply blame of poorer performers.