Date of Award

Winter 1998

Document Type


Degree Name

Doctor of Philosophy (PhD)


Engineering Management

Committee Director

Resit Unal

Committee Member

Derya A. Jacobs

Committee Member

Abel A. Fernandez

Committee Member

C. Michael Overstreet


As software costs become an increasingly higher percentage of total computer system costs, it becomes increasingly more important for software development managers to have the ability to predict development costs with reasonable accuracy early in the software development cycle. Software development cost estimates are based in large measure on software size. Function points are considered by many to be a de facto industry standard as a size metric. The function points technique, unlike lines-of-code, can be applied early in the software development cycle and is language independent. Critics claim that the function point “value adjustment factor,” which purports to capture the effects of software complexity considerations in the final function point count, is inadequate as currently determined. The research described herein develops a new approach, using a less restrictive multiplicative model instead of the existing additive model, for calculating the value adjustment factor. The proposed approach was implemented and evaluated using data from 301 software development projects. However, no improvement in model performance was realized, at least using the limited data available, which were for the most part representative of only one software application domain. Therefore, no conclusion can be drawn from the results of this effort as to whether the proposed multiplicative model is an improvement over the additive model for software development projects in general. The results do, however, focus attention upon two areas which merit further investigation: the performance which would be realized in applying the proposed new model to data more representative of a cross-section of modern software; and the question of whether the function point general system characteristics (upon which the value adjustment factor is based), as currently defined, adequately capture the effects of potential system cost drivers. This research makes the following contributions: an alternative approach for capturing the effects of Function Point general system characteristics on development effort and cost has been demonstrated; there is an indication that the general system characteristics are in need of a thorough review for appropriateness and adequacy; and the need for better dialogue between various elements of the software sizing and cost estimating community is identified.