A hybrid model of the software development process

Software projects are often significantly late or over budget, and many fail to deliver the performance they promise. Software management has followed other areas such as manufacturing in focusing on process as a way to improve both the project and the final product. Models of the software process have either modeled the project environment using system dynamics, or have modeled the activities in the software process using discrete event simulation. Discrete event models may be based an state transitions or a sequence of activities. Each paradigm gives valuable insights, but neither is capable of modeling the entire process within the context of the project environment.; This dissertation presents a new simulation model of the software development process that combines the two major paradigms. By creating a hybrid model, we may examine the impact of changes to process within the context of different assumptions about the project environment.; For example, hiring delays and attrition rate affect the number of experienced staff. A project with inexperienced staff may benefit from a Unit Test process since inexperienced staff may be more likely to create errors. A more experienced staff may be able to skip the Unit Test step. A systems dynamics model can evaluate the difference in experience, but would not model the unit test activity. A discrete event model can evaluate the impact of the Unit Test activity, but may not model the dynamic behavior of the experience level correctly. In order to ask how changes to the hiring delay affect our decision to perform the Unit Test activity, we must model both the process steps and the dynamic staffing levels.; This dissertation presents a hybrid model in detail, demonstrates its validity and then describes the application of the model to a major software development process.