Estimating Software Productivity and Quality on Large Systems
Estimating productivity (e.g., lines of source code developed per hour) and quality (e.g., code defect rates) are difficult on large software projects that involve several companies or sites, emphasize reuse of Commercial-Off-The-Shelf (COTS) components or adaptation of legacy code, and require open architectures. Using actual metrics from such software development projects, this paper illustrates problems encountered and lessons learned when measuring productivity and quality. These include: how to count different types of code; effects of lengthy development times on productivity/quality; variability
between estimates obtained from different models; and tracking and reporting metrics on productivity/quality for projects based on incremental or evolutionary development.