The use of function points to find cost analogies

Abstract

Finding effective techniques for the early estimation of project effort remains an important — and frustratingly elusive — research objective for the software development community. We have conducted an empirical study of 21 real time projects for a major software developer. The study collected a range of counts and measures derived from specification documents, including a derivative of Function Points intended for highly constrained systems. Notwithstanding the fact that the projects were drawn from a comparatively stable environment, traditional approaches for building prediction systems, (for example, regression analysis) failed to yield a useful predictive model. By contrast, estimation based upon the automated search for analogous projects produced more accurate estimates. How much this is a characteristic of this particular dataset and how much these findings might be more generally replicated is uncertain. Nevertheless, these results should act as encouragement for follow up research on a much under utilised estimation technique.