Launching a health-insurance platform in two countries

Context

A multinational insurer implemented a new health-insurance core platform for its operations in Costa Rica and Honduras.

Problem

Move policies, clients, insured members, dependents, their contact details (addresses, emails, phones) and several extended and intermediate tables, with their history, from the old systems into the new platform.

Every relationship between them had to stay intact, so each country could operate from go-live without losing or corrupting a record.

Constraints

• Healthcare data.
• Insurance rules that differ by country.
• Oracle core systems.

Architecture

The diagram, step by step

1. Legacy Oracle sources.
2. Extract with Data Pump and flat files.
3. Clean and map to the new model with Python and PySpark.
4. Staging through direct database links and cloud storage.
5. Bulk load with SQL*Loader.
6. New core platform for each country.
7. Validation scripts per jurisdiction.

Key decisions

• Validate per jurisdiction. Custom validation scripts enforced each country’s healthcare and insurance rules before go-live.
• Bulk load for volume. SQL*Loader moved the historical records at speed.
• Tune the hot path. PL/SQL procedures and triggers were optimized, cutting execution time by 30% in a high-concurrency environment.

Outcome

Two countries live, with policies, clients, insured members and dependents migrated with 100% integrity. Of the 55,000 insured members, 35,000 are dependents.

Integrity was proven by testing in several environments, cross-checks with subject matter experts, and formal sign-off before each go-live.

My role

I architected and executed the migration: development, validation, testing in each environment and sign-off with the business experts.

What I can’t share

The insurer’s name, the platform’s name, its internal systems and exact volumes per country.