Identifying and Justifying Metrics for Software Evolution Investigations Using the Goal-Question-Metric Method

A major aspect of the FEAST programme of research into long-term software evolution processes has been the collection of metrics for evolving software products, and the examination of the trends over time in these metrics. Due to difficulties in collecting suitable metrics over the long timescales necessary to perform FEAST analyses, the selection of metrics has generally been pragmatic, relying on raw (or extracted and suitably processed) data available from FEAST industrial collaborators. These exercises have supported the broad conclusions drawn to date. We suggest that the long-term objectives of the FEAST research programme could now usefully be served by identifying more detailed research hypotheses, drawing on current results. These hypotheses could be used to identify the metrics needed to test them, and thus form the basis of future metrics collection programmes. We propose that the goal-question-metric method be used to derive relevant metrics from FEAST goals, and that these metrics be used as the basis for data collection programmes in current and future FEAST collaborative research exercises, particularly in those cases where new metrics collection programmes are to be set up specifically to support FEAST research.