This paper provides a preliminary overview of a procedure for monitoring software development with respect to quality. The monitoring process is based on the extraction, analysis and interpretation of metrics relating to software products and processes. In particular, we are concerned with the problem of interpreting software metrics data in terms that are meaningful to project and quality managers. We present a four-stage model of the interpretation of metrics which comprises the identification of abnormal metric values (i.e. exceptions), possible causes of exceptions, additional information required to distinguish between causes, and possible corrective actions. We envisage that the model could be implemented as an advice system for managers. We discuss the general principles of project monitoring throughout the development life-cycle, and identify metrics that may be collected throughout, from requirements specification to integration testing. We suggest some general principles for assessing the cause of abnormal metric values, for distinguishing between causes, and for responding to unfavourable project circumstances.