The Modeling Approaches To Establishing Process Performance Models For Small Software Projects

Process Performance Models (PPMs) become the essential ingredient for a high maturity organization as described by the Capability Maturity Model Integration (CMMI). However, many organizations encountered difficulties in establishing successful PPMs for software projects. On the one hand, modeling techniques are in a distant area from those knowledge and skills of software practitioners. Oftentimes software projects struggle for a useful PPM with frustrations due to their lack of sound understanding of statistical thinking. On the other hand, the nature of software projects seems to make them inherently difficult to apply statistical methods upon. The software industry is considered to be much more dynamic than the traditional manufacture industry, in terms of people, technology, and even process. Managers feel that they are hardly able to stabilize a process before it changes, let alone model the process for performance prediction. That is particularly true for small software projects considering the amount of data available for analysis is even less.Our research aims to explore a practical approach to the establishment of PPMs for small software projects. ANOVA, regression analysis, as well as the Monte Carlo simulation tool are applied to a carefully selected small project with desirable results. This paper presents how those analytical methods are used jointly to create a successful performance model for the review sub-process. How the model provides a quantitative foundation for the project to make management decisions are also explained. In the end, constraints of using the model are discussed and experiences and lessons are shared.