Software process improvement is seen by many as a viable strategy for overcoming the software crisis. It is, however, difficult to determine the actual effect of such improvement efforts. This paper reports from an organization that is implementing a software metrics program with the expressed purpose of measuring the effects of their improvement project. The metrics program is intended to measure key indicators of all completed projects and summarize progress information in a quarterly management report.The paper describes important events during the implementation of this metrics program. The process turns out to be long and complex and the software organization finds itself confronted with dilemmas based on contradictory demands and value conflicts. The paper interprets the implementation process from two complementary perspectives. First, it is viewed as a rational, engineering process in which a metrics program is constructed and put into use. Second, it is seen as an evolutionary, cultivation process in which basic values of the software organization are confronted and transformed. This analysis leads to practical advice on how to implement metrics programs as part of software management practices.