RADAR: A Risk-driven decision framework for architecture view and artifact selection

Architecture views and artifacts are widely used to support design and review decisions in software systems [1,2]. However, in industrial practice, architectural documentation is often driven by fixed checklists or informal convention, leading either to excessive documentation or insufficient visibility into dominant architectural risks [5,8,10]. While existing viewpoint taxonomies clarify what architectural viewpoints exist and what concerns they address, they provide limited guidance on which viewpoints and artifacts should be produced for a given project [1,2,4]. In practice, architecture views are often realized through a small number of shared artifacts, each supporting multiple analytical concerns, rather than through one-to-one view–artifact mappings [2,3].

This paper presents RADAR, a Risk-Aware Decision framework for Architecture Representation, which operationalizes architecture view and artifact selection as an explicit decision activity [5–7]. RADAR operates on top of a stable viewpoint taxonomy and a project-driven architecture meta-model, linking project characteristics to dominant risk categories and mapping risks to viewpoint relevance [6,8]. Selection outcomes are expressed as decision results, classifying views as Mandatory, Recommended, or Optional, rather than prescribing fixed documentation sets [1,6,7].

The framework emphasizes minimal sufficiency, stakeholder alignment, and traceable justification, enabling architects to focus documentation effort on areas of highest uncertainty and risk exposure [2,3]. Its application across multiple industrial case studies demonstrates reduced documentation overhead, improved review focus, and earlier surfacing of cross-domain and non-functional risks [7,9]. RADAR does not replace architectural judgment; instead, it structures and amplifies it, providing a practical decision-support mechanism for risk-aware architecture representation in heterogeneous projects [3,5].