Reserving Methods8 min read

Granular Claims Data in Reserving: What It Adds and What It Costs

By Jonas Mohamed Osman Abdelghafour

Published 16 September 2026

A realistic assessment of moving from triangles to claim-level data, including data quality, lineage, reproducibility and team capability.

Executive answer

Granular claims data improves reserving where the aggregate triangle hides heterogeneity that matters — changing mix, shifting settlement behaviour, or segments too small for their own triangle. It costs data engineering effort, reproducibility discipline and a different skill mix in the team.

Data quality is the binding constraint

Claim-level analysis exposes issues that aggregation conceals: retrospective field updates, inconsistent coding across systems and eras, missing transaction history after migrations, and reopened claims handled inconsistently. Most granular reserving projects spend the majority of their effort here, and projects that do not are usually building on unexamined data.

Reproducibility requirements

Financial reporting requires the ability to reproduce a prior period's figure exactly. That implies immutable data snapshots, versioned transformation code and recorded model versions. Retrospective updates to source fields make this impossible unless snapshots are taken, which is a design decision to make before the first model is built.

Governance considerations

The reserving committee still needs a comprehensible narrative. Granular models should produce the same explanatory artefacts as aggregate methods — movement analysis, actual versus expected, drivers of change — or the increase in information will be offset by a loss of challenge.

Conclusion

Granularity is a means, not a goal. It pays where heterogeneity is material and the data foundation is sound, and disappoints elsewhere.

Primary sources

granular datadata qualityreserving