On measuring programmer team productivity

Abstract

It is difficult to study industrial programmer productivity because of the extreme variance seen among individual programmers and the difficulty of performing controlled experiments. As an alternative to studying individual programmers, we examine the group productivity of programmer teams. We postulate that there is such a thing as average programmer productivity, in a given context. By studying programmer teams, we can eventually obtain measures of the expected performance of an average programmer in a defined context. Differences in project productivity can then be attributed to process characteristics. Existing project data is examined to see how data could be collected to support the idea of a standard programmer.