Early FP Estimation and the Analytic Hierarchy Process

Several methods exist in order to estimate the size of a software project, in a phase when detailed Function Point counts cannot be performed. While the Derived Estimation (or Algorithmic Model) Methods result in more accurate, objective, and repeatable estimates, these need always to be calibrated through a set of historical (statistical) data; this also means that estimation for atypical cases or new software typologies cannot be provided because of this "calibration from the past". The paper shows how the combination of the Early FP estimation method with the quantification technique of subjective judgements, known as Analytic Hierarchy Process, can result in a enhanced, multi-approach estimation method. The EFP method provides a breakdown, hierarchical structure of the software functional items, while the AHP technique, based on pair-wise comparisons of all the items (and sub-items) of the hierarchy, provides the determination of a ratio scale of relative values between the items, through a mathematical normalisation. Consequently, it is not necessary either to evaluate the numerical value of each item, or to use statistical calibration values, since the true values of only one or few components are propagated in the ratio scale of relative values, providing the consistent values for the rest of the hierarchy. This methodological combination can be even improved by using as reference a set of few well-known functions, structured into Catalogues of Functionalities, if present. This new combined estimation method is robust to errors in the pair-wise comparisons, and selfconsistent because of the redundancy and the normalisation process of the comparisons. A natural extension to this approach is the link with the general estimation of other variables of the software project, as effort or quality, by means of the same reasoning.