Living with function points

The number of people who develop software at AT&T Bell Laboratories has grown steadily since the 1960´s. Major productivity improvements were made, but efforts to reliably measure them were unsuccessful. With more than half of its technical people working on software projects and one in every 12 employees in AT&T now involved in one way or another in the creation of software, a major effort to produce world-class software is underway. The key to this new program is a shift from counting lines of code to using function points to size software projects. Improved predictability, simplified designs, contained feature creep, and reduced redundancy are the direct results of deploying formal function point analysis (FPA). Estimates of the effort required to do a job are twice as accurate with FPA as with conventional methods. This presentation describes: the software initiative at AT&T, the history of usage of FPA in AT&T, source lines of code (SLOC) vs. function points and the best current practices, automation of the function point count and estimation process, changes in AT&T´s metrics policy and experiences in measurements based on FPA. The bumpy road of software measurement, the problems in estimating software project size and effort, and some positive results of using a measurement program are also discussed