Measuring system complexity to support development cost estimates

Systems and System-of-Systems (SoS) are being used more frequently either as a design element of stand alone systems or architectural frameworks. Consequently, a programmatic need has arisen to understand and measure systems complexity in order to estimate more accurately development plans and life-cycle costs. In a prior paper, we introduced the System Readiness Level (SRL) concept as a composite function of both Technology Readiness Levels (TRLs) and Integration Readiness Levels (IRLs) and touched on system complexity. While the SRL approach provides a repeatable, process-driven method to assess the maturity of a system or SoS, it does not capture all aspects of system complexity. In this paper we assess the concept of cyclomatic complexity as a system complexity metric and consider its utility as an approach for estimating the life-cycle costs and cost growth of complex systems. We hypothesize that the greater the number of technologies and integration tasks, the more complex the system and the higher its cost to develop and maintain. We base our analysis on historical data from DoD programs that have experienced significant cost growth, including some that have been cancelled due to unsustainable cost (and schedule) growth. We begin by describing the original implementation of the cyclomatic method, which was developed to estimate the effort to maintain system software. We then describe how the method can be generalized and applied to systems. Next, we show how to estimate the cyclomatic number (CN) and show the statistical significance between a system's CN metric and its cost. We illustrate the method with an example. Last, we discuss opportunities for future research.