This paper presents a three-layer model that captures the fundamentals of software metrics within a unifying framework. The model readily lends itself for use in both instructional and practitioner environments.

The first (lowest) layer of the model consists of the three primitive software engineering metrics: person-months (PM), function-points (FP), and lines of code (LOC). They are presented as "primary" metrics from which other metrics are computed. Time is also included as a fundamental (not necessarily software) primary metric. The second layer consists of general-purpose metrics such as productivity measures, which are computed from the primary metrics, and the third layer consists of special-purpose metrics such as reliability and quality measures. This third layer is inherently extensible.