Socially-Informed Jupyter Notebook Quality: A Role- and Lifecycle-Aware Metrics Framework

Jupyter Notebooks are widely used in data science, education, and research for their ability to combine executable code with narrative and visualization. However, their interactive and exploratory nature often leads to quality concerns—such as inconsistent execution order, hidden dependencies, and low modularity—that negatively affect key software quality characteristics like maintainability, reliability, and analyzability. Existing quality models, including those based on ISO/IEC 25010, are not well-suited to the specific structure and workflows of notebooks. In this paper, we propose a role- and lifecycle-aware quality assessment framework tailored to Jupyter Notebooks. Drawing from ISO/IEC 25010 and the EOSC TF–SQ3 catalogue, we identify 25 measurable quality metrics relevant to common notebook practices. We organize these metrics using Kruchten’s 4+1 View Model and map them across the stages of the data science lifecycle. Drawing on practitioner insights and developer social-network patterns, we combined expert interviews with Rank-Biased Precision (RBP) and statistical methods (PCA and clustering) to prioritize our 25 metrics. The resulting framework organizes them into universal, viewpoint-specific, and context-dependent categories—offering Tier 1 and Tier 2 users a lightweight roadmap to boost notebook quality. Expert feedback confirms its clarity, relevance, and practical utility, paving the way for tool integration and future refinements.