Parallel DWH Development During Core System Modernization for a German Insurer

Challenge

A German insurance company with over 100 years of history was modernizing its central insurance administration system, the backbone of policy management, claims processing, and customer data. The modernization introduced a bi-temporal information model in the new source system, tracking both the time an event occurred in the real world (business time) and the time it was recorded in the system (system time). This added a layer of complexity that the existing data warehouse was not designed to handle.

The compound challenge was that the DWH had to be developed in parallel with the ongoing source system rollout, before the source system was even fully operational. Traditional approaches would have required the source system to be completed first, delaying analytics and reporting capabilities by months. The business could not afford that delay: migrated customer and policy data needed to be available for reporting immediately to validate the migration and maintain operational continuity. Without timely reporting, the migration team would have lacked the means to verify that data had been transferred correctly, and business operations would have continued to rely on outdated figures.

Approach

Alligator Company designed the data warehouse around a compact, concrete object model built directly from the insurance domain using ELM (Ensemble Logical Modeling). Rather than introducing generalization or specialization layers that would have added abstraction without value, the team kept the model close to the actual insurance business objects: policies, claims, customers, and their relationships. This domain-driven approach made the model understandable for both technical and business stakeholders and reduced the communication overhead between the DWH team and the insurance domain experts.

To handle the bi-temporal source system, the team applied a specialized Data Vault pattern: hubs with and without versioning, replacing the need for multi-active satellites (MAS). Versioned hubs track changes to business keys over time, while non-versioned hubs represent stable identifiers. This preserved both system time and business time without introducing excessive complexity into the Data Vault structures.

Model-driven automation (MDA) through Datavault Builder then accelerated the implementation. Because the source system was still evolving, the data structures changed frequently. MDA allowed the team to regenerate affected DWH components from the model rather than manually adjusting each one, keeping development cycles short and consistency high across iterations.

New QlikView reports were built on database views, providing the business with immediate access to migrated data. Where query performance required it, the team applied selective materialization through Business Vault persistence, ensuring that frequently accessed or computation-heavy reports ran within acceptable response times without materializing the entire model.

Outcome

The DWH went live while the source system was still being completed, decoupling the two timelines and eliminating the traditional sequential dependency. This parallel delivery approach saved the project several months compared to a sequential rollout.

Migrated data from the central insurance system was directly accessible and verifiable in the reporting layer, supporting migration validation in real time. The business used DWH reports to confirm that customer and policy data had been transferred correctly, catching discrepancies early rather than after the full system cutover.

• DWH delivered in parallel with the source system, saving months of sequential delay
• Migrated insurance data immediately available for reporting and migration validation
• Model-driven automation kept development cycles short and quality consistent across frequent source system changes
• Selective Business Vault materialization ensured report performance without over-engineering the persistence layer