|Authors||M. Jørgensen and S. Grimstad|
|Editors||A. Tveito, A. M. Bruaset and O. Lysne|
|Title||Software Development Effort Estimation: Demystifying and Improving Expert Estimation|
|Afilliation||Software Engineering, Software Engineering|
|Publication Type||Book Chapter|
|Year of Publication||2009|
|Book Title||Simula Research Laboratory - by thinking constantly about it|
Research background: Inaccurate estimation of software development effort is one of the most important reasons of IT-project failures. While too low effort estimates may lead to project management problems, delayed deliveries, budget overruns and low software quality, too high effort estimates may lead to lost business opportunities and inefficient use of resources. These IT-project problems motivated the BEST (Better Estimation of Software Tasks) project at Simula Research Laboratory to conduct research with the goal of improving effort estimation methods. Scientific challenges: The project's main focus is to improve judgment-based effort estimation (expert estimation), which is the estimation approach most frequently used by the software industry. We argue that a better understanding of the mental steps involved in expert estimation is necessary to achieve robust improvement of software development estimation processes. A great challenge when studying expert judgment is to understand the unconscious steps involved. To study these steps, we make use of multidisciplinary competencies, especially psychology and software engineering, and financial resources enabling studies in realistic software development effort estimation contexts. Obtained and expected results: In this section, we present the main results of the project from its beginnings in 2001 until today and describe how the project has benefited from being an integrated part of Simula Research Laboratory, especially through the opportunity to conduct large-scale, controlled experiments in field settings. The section includes practical results on how to improve estimation methods, scientific results leading to a better understanding of the mental processes involved in judgment-based effort estimation, and our innovative research methods in the field of empirical software engineering. Results of the BEST project are currently in use by several software companies, as well as by researchers in forecasting and psychology, and they are also included in project management and software engineering textbooks. The BEST project plans to continue its multidisciplinary effort with the goal of constructing and evaluating better models of the mental steps involved in judgment-based effort estimation. An improved model of these steps will enhance our ability to accurately estimate software development effort and to predict when different types of estimation methods can be expected to deliver accurate effort estimates. This will in turn lead to a reduction of the number of IT-project failures and make better use of scarce financial and human resources.