A Large-scale Empirical Evaluation of Function Points Estimation Methods

Functional size measures of software—especially Function Points—are widely used, because they provide an objective quantification of software size in the early stages of development, as soon as functional requirements have been analyzed and documented. Unfortunately, in some conditions, performing the standard Function Point Analysis process may be too long and expensive. Moreover, functional measures could be needed before functional requirements have been elicited completely and at the required detail level. To solve this problem, many methods have been invented and are being used to estimate the functional size based on incomplete or not fully detailed requirements. Using these methods involves a trade-off between ease and timeliness on one side and accuracy on the other side. In fact, estimates are always affected by some error; knowing the magnitude of estimation errors that characterize the estimates provided by a given method is of great importance to practitioners. This paper reports the results of an empirical study devoted to evaluate the accuracy of estimates provided by Function Points estimation methods. The results of the study show that some of the evaluated methods—including the Early & Quick Function Points, the ISBSG average and the NESMA estimated method)—provide estimates that are accurate enough for practical usage, while some other methods appear quite inaccurate.