Beginning a Measurement Program
A Learning Experience With Function Point Analysis
Bonndata delivers software and IT services to a large German insurance company and has a staff of about 500 people, software engineers being the largest group. In 1997 we started a metrics program in our company which aimed at establishing a business view upon our software development activities and at controling our improvement efforts from a financial point of view. We first wanted to benchmark us against other companies in the field of financial services. Our hope was that Function Point Analysis (FPA) would give a common base for performance measurement, cost estimating, and software portfolio analysis. When we tried to implement FPA in a variety of different subcultures we got into several dilemmas from which we could not escape and recover. We had to abandon the endeavour and restart again under totally different premises. The main point of this paper is that FPA may not be a tool suitable for beginning a metrics program, because it focuses upon a special technology of functional size measurement instead of establishing a context for gradually incorporating measurement into the management of software processes. We learned that no single software measure – function points or other – was really important in the beginnings, but that better managing the processes of estimating, planning, and controling was what we needed in the first place. These processes determine which measures are useful or useless. When restarting the measurement project in 1999 we first explored the field of measurement and process improvement in many directions. Right now we are entering the mainstream of process improvement and following the path of the Capability Maturity Model (CMM).