FinOps for the modern lakehouse
The five levers that actually move data platform cost, ranked by impact — and the three that don't.
Cloud data platform costs have a way of becoming a board-level line item. Here is what actually moves them.
Lever one: query concurrency and warehouse sizing. On both Snowflake and Databricks SQL, the largest wins come from right-sizing the warehouses your BI tools connect to. This is unglamorous but consistently the single biggest lever.
Lever two: storage lifecycle. Cold data belongs in cheaper storage tiers. This is nearly free to implement and pays back immediately.
Lever three: workload isolation. Ad-hoc queries and scheduled pipelines should not share the same compute. Mixing them is the fastest way to unpredictable spend.
Lever four: query optimization. Materialized aggregates, clustering, and caching, in that order. Do not attempt this before the first three levers are in place.
Lever five: reserved capacity. Only after your workloads are stable and predictable. Otherwise you're locking in inefficiency.
What doesn't work: vendor swaps as a cost strategy, aggressive migration to open table formats without a runtime plan, or building a bespoke cost dashboard before you've done any of the above.
Priya leads Dataamps' data platform practice, with 15 years shipping lakehouse and streaming systems inside Fortune 500 enterprises.
More from PriyaRelated articles
Notes from an Iceberg migration
Lessons from moving 12 PB of data to Apache Iceberg without downtime — and what we'd do differently.
Reverse ETL is plumbing, not strategy
Why treating activation as a product decision — not a category to buy — is the difference between value and shelfware.
Data contracts are boring — and that's the point
Why the least glamorous idea in the modern data stack is quietly the highest-leverage one for teams tired of 3AM pages.