Average Function Points Per Person-Month
The average rate of productivity in the United States is five function points per person-month. Many organizations developing systems with COBOL, database tools, and the traditional IS life cycle average about eight function points per person-month. Some have used productivity aids with otherwise manual COBOL development to push productivity to 12 function points per person-month.
Banking Example of Productivity Increase
This example shows the increase in productivity when a bank moved from a third generation development environment to a fourth-generation environment with CASE tools.
With this change, COBOL development reached a plateau of about 12 function points per person-month, a fairly typical number for well-managed use of a CASE tool and a separate code generator.
Productivity with CASE
Highly trained teams using integrated CASE tools and rapid development techniques can achieve productivity figures much higher than 12 function points per person-month. Small teams of developers at DuPont Fibers, using "timebox" development and CASE tools, achieve an average life-cycle productivity of about two hours per function point, or 80 function points per person-month. This is more than four times the productivity DuPont achieved when using a traditional life cycle and fourth-generation languages in conjunction with COBOL and PL/l. (DuPont's productivity using a traditional development life cycle was higher than that of most other enterprises.)
The average development productivity achieved with the Timebox methodology compared with the productivity with traditional methodologies (including the use of fourth-generation languages) in the IT community of DuPont, Fibers Division.
Wide Variations in Productivity
The move from third-generation languages to CASE tools can greatly increase productivity. Yet there are wide variations in productivity even among the organizations using the most up-to-date tools. These variations can be attributed to management differences.
Dr. A. Picardi made a detailed study of 30 applications developed at 22 sites using the Cortex toolset. He measured productivity for each development effort, and compared the results with productivity in a traditional COBOL life cycle (using estimates derived with a common estimating technique). The productivity of the 30 development efforts ranged from twice that of COBOL to 45 times that of COBOL. The average productivity of the 30 efforts was 13.2 times better than COBOL productivity. For both small and large systems, productivity varied over a wide range.
These figures indicate that the toolset is useful, but that different teams achieve widely different levels of productivity with it. The same variation has been found for most other productivity tools. Dr. Picardi found eight projects that achieved productivity higher than 100 function points per person-month, but these high productivity projects did not seem to have any special defining characteristics other than that they were well-managed. Maximum productivity, then, requires good management, which includes applying an appropriate RAD methodology and carefully selecting and motivating the people involved.
Today, there is an astonishing gap between the best applications development efforts and the work being done in most IS departments. Relatively few IT communities have acquired the ability to build high-quality applications fast, and the best rapid development teams achieve productivity figures that are 10 times those of most organizations.
Previous : Metrics
Section Overview : Management
Overview : Table of Contents