From fragmented data to business intelligence: a data-centric CRISP-DM framework for delegated insurance

The increasing complexity and heterogeneity of data in Delegated Broker Business insurance models challenge insurers’ ability to ensure data quality, traceability, and operational reliability. While the CRISP-DM is widely adopted, its limited support for data governance and multi-source integration constrains its use in modern, data-intensive environments. This paper proposes an extended, data-centric framework that adapts CRISP-DM to the realities of delegated broker business ecosystems. The framework integrates principles from Agile, Lean, and DataOps, embedding quality assurance and governance across six iterative phases: Business Understanding, Data Understanding, Data Preparation, Dashboard Development, Validation, and Sustainability. It prioritises the standardisation and transformation of heterogeneous data into reliable, analysis-ready assets. A real-world implementation within an insurer operating in the French market demonstrates the framework’s efficacy, transforming a manual, fragmented process into an automated business intelligence pipeline. This transformation yielded significant improvements in data consistency, a substantial reduction in data dimensionality, and enhanced operational alignment. Governance artefacts, including a data cleaning rulebook and residual risk matrix, enables continuous monitoring and refinement. The study confirms the framework successfully bridges data engineering, governance, and business intelligence, offering a scalable methodology to build trustworthy foundations for decision-making and future AI adoption within the insurance sector and other data-intensive industries.