Expert Estimation and Historical Data: An Empirical Study
Gabriela Robiolo, Silvana Andrade dos Santos, Bibiana D. Rossi · International Conference on Software Engineering Advances · 2013
Expert estimation is the estimation strategy which is most frequently applied to software projects; however, this method is not very much reliable as the accuracy of the estimations thus obtained is always influenced by the level of experience of the expert. As part of the experts’ experience is made up by the information they obtain from historical data, we wanted to learn about the value such historical data has for an expert estimator. To do so, we designed an empirical study. We compared the accuracy of the estimations made with several estimation methods based on productivity, size, and analogies which use historical data, to that obtained with expert estimation. We used two similar applications; one was used as the target application and the other one was used to obtain historical data. The results show that the accuracy of expert estimation is affected by the expert’s work experience, the level of experience he/she has in the technologies to be used to develop the applications, and his/her level of experience in the domain of the applications. The use of historical data may improve the intuitive expert estimation method when the work experience, the experience in the technologies to be used to develop the application, and the experience in a given domain is low, as well as when the team velocity is unknown. Keywords—Expert; Expert Estimation; Effort Estimation; Empirical Study; Historical Data.