CASE STUDY: FarmMutualRe

Farm Mutual Re

How we migrated a legacy reinsurance data environment into a cloud-native Lakehouse — cutting reporting time from weeks to minutes and unlocking real-time executive analytics.

The Challenge

FarmMutualRe's data infrastructure was spread across multiple siloed on-premises SQL databases with no centralized governance. Analysts spent 60–70% of their time manually extracting, reconciling, and reformatting data before any reporting could begin. Critical business metrics — loss ratios, reserve adequacy, ceded premium tracking — were often weeks stale by the time they reached decision-makers. As regulatory reporting requirements grew more complex and the volume of policy data scaled, these manual pipelines became a serious operational bottleneck prone to errors and inconsistencies.

END-TO-END DATA PIPELINE — AZURE MEDALLION ARCHITECTUREON-PREMISESSQL ServerPolicy DataSQL ServerClaims DataSQL ServerReserves DataFile SharesExcel / CSVAzure DataFactorySelf-hosted IRSecure Hybrid LinkBRONZERaw IngestionParquet FilesADLS Gen2Raw SchemaNo transformationsFull audit trailSILVERValidated & CleansedPySpark JobsDeduplicationSchema EnforceIncremental mergeDatabricks computeGOLDBusiness-ReadyFact: ClaimsFact: PremiumsDim: PolicyDim: ReservesStar schemaDirectQuery readyUnity Catalog — Access Control, Data Lineage, Discovery

The Method: Azure Medallion Architecture

We designed a three-tier Lakehouse built on Microsoft Azure, following the Medallion Architecture pattern (Bronze → Silver → Gold). This approach gave FarmMutualRe a clear data lineage trail from raw ingestion to business-ready datasets, enabling both auditability for regulatory compliance and flexibility for future analytical workloads. Azure Data Factory orchestrates scheduled and event-driven ingestion, while Databricks provides the compute layer for transformation logic — all governed by Unity Catalog for access control and data discovery.

The Solution

We established a secure hybrid connection between FarmMutualRe's on-premises network and Azure using a self-hosted integration runtime, streaming raw relational data into the Bronze layer as Parquet files in ADLS Gen2. PySpark jobs on Databricks handle schema validation, deduplication, and incremental merge logic in the Silver layer. The Gold layer contains purpose-built star-schema models — fact tables for claims, premiums, and reserves joined against conformed dimension tables — specifically optimized for DirectQuery consumption by PowerBI, eliminating the need for data extracts entirely.

POWERBI CONSUMPTION LAYER — REAL-TIME EXECUTIVE ANALYTICSGold LayerLakehouseDirectQueryNo extractsFARMMUTUALRE — EXECUTIVE DASHBOARDLOSS RATIO62.4%▼ 8.2% YoYRESERVES$48.2MAdequateCEDED PREMIUM$124M▲ 12% QoQREFRESH<5mwas 2-3 weeksMONTHLY CLAIMS BY LINE OF BUSINESSPREMIUM TREND — 12 MONTHSKEY OUTCOMES1,000+ analyst hours reclaimed/yearReporting: weeks → minutesPredictive modeling now possible

End State & Outcomes

FarmMutualRe now operates a fully automated, end-to-end data pipeline that refreshes executive dashboards in near real-time. Reporting cycles that previously took 2–3 weeks are now completed in under 5 minutes. Analysts were freed from manual data wrangling, redirecting over 1,000 hours annually toward higher-value actuarial analysis and strategic modeling. The unified Gold layer also enabled FarmMutualRe to stand up new predictive models for reserve forecasting — something that was previously impossible without a trusted, centralized data foundation.