Estimating the Scope of Software Projects Using Statistics

Most companies follow a well-defined System Development Life Cycle, and would prefer to estimate the size of projects as early as possible. For software projects, this is difficult because of the intangible nature of software.

There have arisen two popular methods for estimating the size of a software project. One is counting Function Points, which is a weighted count of the functions identified during the Design phase of the project. Another approach, born out of the object oriented design movement, is Use Case points which is a weighted count of the tasks identified during the Requirements phase of the project.

Both of those approaches require metrics which are not known until a specific phase of a software project. This paper presents the hypothesis that metrics gathered during even earlier phases of a software project could also be used to estimate the size of a software project.

A survey was conducted to gather a key metric at each stage of a software project, as well as the resulting Source Lines Of Code (SLOC) count that were written to implement that software project. A least squares linear regression is performed to derive a linear equation (of the form SLOC = Ax + B) that can be used to predict the resulting SLOC. A coefficient of correlation is calculated for each equation, indicating how well those equations actually do predict SLOC.