Performance Management in Software Engineering

Abstract

Performance measurement in software engineering has to meet a multiplicity of challenges. Oftentimes, traditional metrics focus on sequential development instead of using incremental and iterative development. Output is measured on a pure quantitative (e.g., SLOC), quality-disregarding basis. A project’s input is hard to assign properly using enterprise-unspecific forecasting tools which have to be calibrated at first and which do not account for time preferences. Requirements necessary for behaviourally adjusted project management and control are rarely discussed. Focusing on these shortcomings, this paper proposes an enterprise-specific approach which combines lifecycle and activity based costing techniques for software development following the incremental and iterative Unified Process model. Key advantages are calibration effort can be avoided, project management decisions are supported by a clear managerial accounting emphasis, precise milestone-depending cost objectives can be determined as the basis for personnel management and control of development teams, and cost and time variance analysis can be supported in a sophisticated way.