Context
SWICA is one of Switzerland's largest health insurers and has been investing heavily in becoming a data-driven organisation. For years, the business relied on a central Data Warehouse for analytics and reporting. As demand grew, an extended Data Warehouse (eDWH) was built on top to serve additional use cases — more flexibility, more data sources, faster turnaround.
But the central model had limits. Every new report or analysis had to go through the eDWH team, creating bottlenecks. Technical debt accumulated. Lead times grew. It became clear that a fundamentally different approach was needed.
The answer was Dragonfly, a decentralised data platform based on Data Mesh principles, built in-house. The idea: empower business domains to own and manage their own data, supported by a shared platform layer on Microsoft Azure (Azure Synapse, Azure App Services).
My role
I led the eDWH migration end to end — from planning and stakeholder alignment to architecture decisions and coordination with ipt as our implementation partner. With a team of 5–10 people, we were responsible for moving the full eDWH workload onto Dragonfly and ultimately shutting down the old system.
What made it hard
The platform wasn't ready. Dragonfly had been productive for small use cases since late 2023, but the eDWH migration was its first real stress test. We weren't migrating onto a finished product — we were migrating onto a platform that was still being built. That meant fixing platform issues in parallel with running the migration itself. It was uncomfortable, but waiting for perfection wasn't an option.
The timeline was tight. One year to migrate 70 data pipelines and a set of complex reports and data marts that the business depended on daily. There was no room for extended discovery phases or slow onboarding. We had to move fast while keeping production workloads stable.
Change management was real work. Moving from a central eDWH to a decentralised model doesn't just change technology — it changes ownership, responsibilities, and ways of working. Getting teams to accept that they now owned their data pipelines, rather than handing requests to a central team, required persistent communication and hands-on support.
What we delivered
By the end of 2024, all 70 integration data pipelines and the critical data marts and reports had been migrated to Dragonfly. The old eDWH was fully decommissioned in February 2025 — not mothballed, not running in parallel, fully shut down. That's a milestone that many migration projects never reach.
What I took away
Don't wait for the platform to be perfect. If we had waited until Dragonfly was "ready" by every definition, we would still be waiting. Migrating into an evolving platform forced us to surface issues early, prioritise ruthlessly, and build the platform through real usage rather than theoretical requirements. It's messy, but it works.
The eDWH was the smaller sibling of SWICA's main Data Warehouse. The playbook we built — technically and organisationally — will inform the bigger migration ahead.
