SOFTWARE PROCESS IMPROVEMENT AND PRACTICE Softw. Process Improve. Pract. 2008; 13: 5–17 Published online 24 January 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/spip.362 Probabilistic Analysis of Process Economics ‡ Research Section Florian Deissenboeck* ,† and Markus Pizka Institut f ¨ ur Informatik, Technische Universit¨ at M ¨ unchen Boltzmannstr. 3, D-85748 Garching b. M ¨ unchen, Germany The economic benefit of a certain development process or particular activity is usually unknown and indeed hard to predict. However, the cost-effectiveness of process improvements is of paramount importance and the question how profitable certain activities are needs to be answered. Within a large-scale commercial organization, we were challenged with the task to quantify the economic benefit of isolated test and development environments. To answer this question, we defined a generic process model based on absorbing Markov chains that allows the analyzation of the economic benefit of software process variations. This model exposes conflicts between process steps and reiterations of development activities and thereby provides a highly flexible tool for the investigation of the effects of changes to a development process on its overall performance. This model was used to predict the impact of isolated testing on the overall effort and duration of projects. The results obtained correspond well with the perception of experienced developers and gives a detailed explanation for the effects. Besides this, it can be used to analyze various other economic aspects of software development processes and yields an interesting alternative for cost estimation. Copyright  2008 John Wiley & Sons, Ltd. KEY WORDS: software process economics; process simulation; industrial application; absorbing Markov chains 1. SOFTWARE PROCESS ECONOMICS How does one determine the economic impact of selecting a certain process model? Is XP cheaper than RUP? What are the economic risks of the waterfall model? Does the spiral model actually yield faster time-to-system? All of these questions are of practical relevance, highly important but very difficult to answer. Surely, in a specific situation we assume that alternative A is faster (cheaper, better) than B based on our own experience or those of others but the benefit can neither be quantified nor guaranteed. ∗ Correspondence to: Florian Deissenboeck, Institut f ¨ ur Infor- matik, Technische Universit¨ at M¨ unchen Boltzmannstr. 3, D-85748 Garching b. M ¨ unchen, Germany † E-mail: deissenb@in.tum.de ‡ Part of this work was sponsored by the BMW Group Copyright  2008 John Wiley & Sons, Ltd. In contrast to this, the costs of activities are usually clear and precisely documented in invoices. For example, while the costs are clear, it is unclear how much money can be saved in a given project setting by spending one additional dollar on model-based development techniques. The same applies for well- established activities like documentation as well as for more specialized methods like requirements engineering with formal methods. 1.1. The Value of Isolated Development and Testing Within a large-scale industrial organization, we were challenged with the task to determine how much time and effort is saved by using isolated test and development environments for IBM mainframe (i.e. PL/I, COBOL) based commercial software projects.