DataPulse: high-performance analytics engine for 90+ data sources

Challenge

Operations, finance, traffic, customer service, and fleet — five business units, five data stacks, ninety-plus operational sources. Every cross-functional report required a manual join across three or four databases. Strategic decisions about pricing, route planning, and infrastructure investment were being made on Excel exports a week out of date. Oracle licensing on the analytical workload was the third-largest IT line item.

Approach

Designed a modern analytics layer combining a streaming ingestion bus (Kafka), a columnar warehouse (ClickHouse), and a semantic layer exposing business-friendly models. Built source connectors for the 90+ origin systems with incremental CDC where possible, batch fallback where not. Migrated heavy analytical workload off Oracle entirely, keeping Oracle only for transactional systems where it remained cost-justified.

Outcomes

• €800K+ saved annually on Oracle licensing
• Cross-functional reports now refresh hourly instead of weekly
• Decision latency on operational issues dropped from days to minutes
• Foundation in place for the operator's first ML use cases