Insurance Data Unification

An insurance group needed to unify policyholder data across 4 acquired companies, each with different policy administration systems, data models, and reporting standards.

Each acquisition had brought its own technology stack — one ran a modern cloud-based PAS, another used an AS/400-based system from the 1990s, and the remaining two used different mid-market commercial platforms. Policyholder data was fragmented across these systems with no common identifiers, inconsistent naming conventions, and duplicate records.

The lack of unified data meant the group couldn't identify customers who held policies across multiple subsidiaries, missing significant cross-sell opportunities. Regulatory reporting required manual consolidation from all four systems, consuming an entire team's capacity for 2 weeks each quarter. Data quality was poor — duplicate detection estimated that 28% of records had quality issues including incomplete addresses, mismatched policy-to-customer linkages, and inconsistent product codes.

The board had mandated a unified view of all policyholders within 12 months to support both regulatory requirements and a planned group-wide digital customer experience initiative.

We designed a master data management (MDM) architecture with three core components:

First, a Canonical Data Model defining a common schema for policyholder data across all product lines and subsidiaries. The model accommodated the unique attributes of each subsidiary's products while enforcing consistency for core customer, policy, and claims data. A data dictionary documented every field mapping from each source system.

Second, a CDC-Based Integration Layer using Change Data Capture to stream updates from all four source systems into the MDM hub in near real-time. Each source system connector handles the translation from its native format to the canonical model. Advanced entity resolution algorithms match and merge records across systems using probabilistic matching on name, address, DOB, and policy attributes.

Third, a Unified Data Warehouse optimized for analytics and regulatory reporting. Pre-built report templates for state insurance filings, actuarial analysis, and executive dashboards eliminate manual consolidation. The warehouse supports both standard reporting and ad-hoc analysis with self-service BI tools.

The project was executed in four phases over 11 months:

Phase 1 — Data Discovery and Modeling (Months 1-3): Comprehensive profiling of all four source systems, identification of 2.1 million unique policyholder records, and development of the canonical data model through workshops with business stakeholders from each subsidiary.

Phase 2 — Integration Build (Months 4-6): CDC connector development for each source system, entity resolution algorithm tuning, and initial data load with quality validation. Manual review of 15,000+ potential duplicate pairs to train the matching algorithms.

Phase 3 — Warehouse and Reporting (Months 7-9): Data warehouse deployment, regulatory report template development, and self-service BI tool configuration. Parallel reporting runs comparing MDM-generated reports against manual processes to validate accuracy.

Phase 4 — Go-Live and Optimization (Months 10-11): Full cutover to MDM-sourced reporting, source system connector optimization for near real-time latency, and deployment of cross-sell analytics dashboards. Data stewardship processes established for ongoing quality management.

The data unification project delivered significant business value across multiple dimensions:

A single source of truth was established for 2M+ policyholder records with 97% data quality score (up from 72%). Entity resolution identified 180,000 customers who held policies across multiple subsidiaries — a population that had been completely invisible to the organization.

Cross-sell identification increased revenue by 8% by enabling targeted campaigns to customers holding policies with one subsidiary but not others. The unified customer view also improved retention by enabling service teams to see the full relationship when handling inquiries.

Regulatory reporting preparation time was reduced by 70%, from 2 weeks per quarter to 2 days. Automated report generation from the unified warehouse eliminated manual data consolidation and reduced error rates to near zero.

Data quality improved from 72% to 97% through automated validation rules, duplicate detection, and a data stewardship program. The MDM platform now serves as the authoritative source for all customer-facing and regulatory data initiatives.