A bottom-up approach to software size estimation is described. It first identifies factors affecting software size, thus obtaining size explanation equations, and then seeks suitable predictors based on those explanation factors which can be used for size estimation. The approach, or model, is bottom-up in that it sizes individual software components or modules first, and then obtains subsystem and system sizes by summing component sizes. Since components may have different purposes and characteristics, the model allows for the partitioning of system components into several different types, each component type having different size explanation and estimation equations. The partitioning is not fixed, but depends on the particular software development technology. The model is applied to several different software systems, including both business applications and systems programs.>

The relationship between software models and metrics is examined. The opportunities for CASE application measurement are seen to be much greater than for conventional development. CASE metrics requirements are also seen to be different in a number of important ways from conventional software metrics requirements. CASE application metrics are examined under the headings of process metrics and product metrics. CASE measurement issues and methods are considered together with some potential uses of CASE development measurements.>

Because Function Point Analysis (FPA) has now been in use for a decade, and in spite of its increasing uouularitv has met with some recent criticisms, it is time to &view how appropriate it still is for todav's technoloaies.A critical review of the FPA approach exam&es in part&lar the pioneering and continuing work of Albrecht and more recent work by Symons.Technological dependencies in FPA-type metrics are identified and a general model for deriving a new FPA-type metric for a new software technology is given.A model for the calibration of FPA-type metrics for new technologies in terms of a reference technology is also presented.Such calibration is essential for comparative productivity studies.The role of module estimation in exposing parts of the 'anatomy' of the FPA approach is investigated.The derivation and calibration models are applied to a significant case study in which a new FPA-type metric suited to a particular software development technology is derived, calibrated and compared with other published versions of FPA metrics.