Software documentation is an important impact factor to achieve high software maintainability, especially for those large-scale complex changing legacy systems. The question: “how much documentation is enough”, is concerned by organizations who are turning their software process to agile development that claims “just enough” documentation. It is therefore important to be able to understand the cost of documentation activities, and what are the underlying cost-drivers, in order to monitor, control and improve documentation practice. However, there is a general lack of such studies dedicated for software documentation cost and cost-drivers.
To address this need, a systematic methodology is proposed to analyze cost and cost-drivers of technical software documentation. The methodology primarily consists of the definition of documentation cost and cost-driver metrics, mining software repositories with tool support for automatic measurements, and cost-driver analysis. The main contributions of this thesis are to provide a practical way to understand documentation cost from the perspectives of single document, one documentation type and each author, and to identify underlying cost-drivers towards documentation process improvement. Results from an initial validation from an industrial case study at NovAtel, a leading provider for a comprehensive line of Global Navigation Satellite System (GNSS) products, are reported.