Adaptive Structural Productivity Modeling and Analytics for Software Projects

Abstract

Productivity evaluation in industrial software projects is still predominantly based on aggregate Function Point-to-effort ratios. Such an approach overlooks the internal structural and functional heterogeneity of software projects, potentially leading to misleading comparisons between systems characterized by similar functional sizes but markedly different structural complexities. This work investigates the hypothesis that projects with similar functional structures, within the same application domain, tend to exhibit comparable productivity behavior. For this purpose, an adaptive and structurally informed productivity assessment approach is proposed, based on a Structural Productivity Index that integrates domain-specific productivity information with project functional composition. The proposed model is validated through real-world and synthetic project datasets, combined with consistency indicators and clustering-based analyses to examine its behavior under different conditions and its capability to distinguish regular productivity patterns from structurally anomalous projects. Experimental results suggest that the proposed adaptive productivity modeling approach provides a more coherent interpretation of productivity estimates compared with traditional productivity metrics. The proposed structural representation provides an interpretable basis for future AI-based productivity analytics.