Data Lineage
The documented path data takes from source through transformation, model training, and final use. Essential for AI explainability and regulatory review.
Data lineage is the record of where data came from, how it was transformed, and where it is used. For insurance AI, lineage matters because a model’s output can only be explained if someone can trace the inputs back to their sources and through each processing step.
Regulators increasingly ask for lineage as part of model documentation. The NAIC evaluation tool expects carriers to describe training data, external data sources, and how data is prepared for model use. Vendor-hosted models often make this harder, because the data may be processed outside the carrier’s systems under terms that limit visibility.
A defensible program documents data sources, cleaning steps, feature engineering, version control, and where the data ends up in production. The exam context for these records is under the NAIC AI Evaluation Tool entry; the vendor side is in AI vendor risk assessment.