AutomationLive ProjectCode Showcase

Location Matcher

Intelligent location matching for reinsurance portfolios

Munich Re via NTT Data

Geographic data matching reducing Munich Re's portfolio reconciliation from 40 hours to 6 hours, with 98% matching accuracy across 10,000+ locations.

Technologies

frontend

AngularAG GridRxJS

language

TypeScript

Key Metrics

-85%

Processing Time

98%

Accuracy

Completed: 2024

Overview

The Location Matcher transforms Munich Re's manual location reconciliation process from a 40-hour bottleneck into a 6-hour streamlined workflow. Developed in collaboration with NTT Data, the platform employs a sophisticated multi-factor matching algorithm that combines fuzzy building name matching (Levenshtein distance), geographic coordinate comparison, address normalization, insured value correlation, and occupancy code cross-referencing.

The confidence-based scoring system automatically processes high-confidence matches (90-100%) while flagging uncertain cases (70-89%) for expert review, ensuring both efficiency and accuracy across portfolios with 10,000+ locations.

The Challenge

Munich Re's reinsurance underwriting teams faced a critical bottleneck in portfolio renewal processing. When assessing risk for property portfolios, underwriters needed to match locations between renewal periods — but locations frequently changed names, addresses, or were modified between periods. With 10,000+ locations per portfolio and tight renewal deadlines, the manual reconciliation process consumed 40+ hours per portfolio, created audit trail challenges, and introduced costly errors in risk assessment where millions of euros in exposure could be miscalculated.

The technical challenge required building an intelligent algorithm handling variations in building names, addresses, and geographic coordinates, while supporting multi-factor correlation with appropriate confidence scoring across massive datasets.

The Solution

Developed a four-stage workflow application with Angular frontend integrating with a C# .NET backend. The intelligent matching algorithm employs fuzzy building name matching with 85% similarity threshold using Levenshtein distance, geographic coordinate comparison with proximity calculations, address normalization, insured value correlation within 20% variance tolerance, and occupancy code cross-referencing.

The confidence-based scoring system categorizes matches into automatic (90-100%), suggested (70-89%), and unmatched categories. An interactive review interface provides side-by-side comparison with highlighted differences, bulk accept/reject operations, and direct integration with Munich Re's FPP system for end-to-end workflow automation.

Results

Processing time was reduced by 85% (40 hours to 6 hours per portfolio), and matching accuracy improved from 75% to 98%. The system now handles 10,000+ locations per portfolio with same-day turnaround, freeing 3 FTEs for higher-value analytical work. The four-stage workflow (ingestion, matching, review, export) provides complete audit traceability for regulatory compliance, and direct FPP system integration eliminates manual data transfer errors.