Our data platform costs doubled between October 2024 and May 2025—despite stable usage. The culprit? Uncontrolled growth in Databricks serverless pipelines and Unity Catalog, compounded by the complexity of tracking mixed costs across storage, compute, DBUs, and serverless resources.
This talk shares how we regained control through a four-pronged approach: implementing granular FinOps reporting, empowering DevOps teams with cost visibility, redesigning architecture for efficiency, and enforcing database-level policies and retention rules.
The key to success wasn’t just technical—it required close collaboration between Azure, Databricks, our DevOps teams, architects, and product managers working together to untangle the cost drivers.
The result? Not only did we reverse the cost trajectory, but we built a sustainable, predictable cost model that enables us to ship new data pipelines with confidence—knowing exactly what they’ll cost before we build them.