Improving software quality using comparative metrics

Data-driven decision is a core practice in the software engineering discipline. A long list of metrics exists in the IT industry. Majority of these metrics (e.g. defect removal efficiency), when applied well, have proven to provide useful historical perspective of the software quality. The proper interpretation of these metrics is typically time dependant, requiring meeting of a certain milestone. This creates a constraining effect on the software engineer’s ability to take actions on a cost-effective and ongoing basis on a project. In this paper, we introduce a set of time-independent dimensions to a well-established quality metric. We refer to the metrics, analyzed along the time-independent dimensions, as comparative metrics. Upon slicing and dicing the data along the time-independent dimensions, software engineers can draw logical conclusions and therefore take corrective actions on an ongoing basis. The net result is improved software quality.