Two Properties of Function Point Analysis

Abstract

Function Point Analysis (FPA) has been revised and modified several times since its first publication by Albrecht. These revisions and modifications present different views on functional size. In this paper, we discuss three variants of FPA that are in use today: the IFPUG standard, Mark II FPA and the Full Function Points approach.

The measurement process of FPA is characterized as a two step abstraction, consisting of an abstraction of software functionality and the actual measurement. We present a generalized representation for the abstraction of software functionality that applies to the three FPA variants. This representation is then formalized in the generalized Function Point Structure.

The Function Point measures are formulated as functions in this structure. Based on the generalized Function Point Structure, empirical properties of FPA can be studied. We present two conditions that characterize empirical assumptions made by Function Point Analysis: dominance and monotonicity. These conditions have implications for the prediction of other variables, e.g., effort, with Function Point measures. The three FPA variants are examined with the two conditions, and significant differences in their empirical assumptions are demonstrated.