Capability-Based Software Cost Estimation (CaBSCE): Proposing a New Method to Estimate Software Costs

Abstract

The dynamic nature of software has led to exciting technology improvements as well as challenges in cost estimation. Defining specific software requirements early in the lifecycle is impossible. The Software Acquisition Pathway (SWP), defined in Department of Defense (DoD) Instruction 5000.87 (Office of the Under Secretary of Defense for Acquisition and Sustainment, 2020), requires only a high-level specification of needed capabilities—a capability needs statement (CNS) or software initial capabilities document (SW-ICD)—before entering the execution phase of software-intensive programs (Defense Acquisition University, 2020b). Most cost estimators for the U.S. government rely on size-based estimation methods, primarily source lines of code (SLOC). However, estimating SLOC requires understanding specific implementation details and can only be accurately estimated near program completion. The software cost estimation community needs a capability-based estimation method that aligns with the SWP and meets the government’s need for evidence-based, flexible, and defensible estimates (Defense Acquisition University, 2020a). In response to these competing needs, the Software Engineering Institute (SEI) proposes developing a software capability-based estimation model requiring only high-level information. Using existing software development effort data, the SEI plans to analyze descriptive fields to cluster data points based on similar software functions and complexity. The resulting model would provide evidence-based rough order of magnitude (ROM) estimates for the initial stages of identifying required capabilities.