01 — Cost
Where the money goes
Every workspace, job cluster, and SQL warehouse priced from system.billing.usage — attributed by SKU, workload, and owner, not guessed at from a monthly total.
A ranked, read-only sweep of your own Databricks account — cost, waste, governance, storage, jobs and serving — delivered as findings you can verify line by line.
Plain SELECTs against your system.* tables. No vendor installed, no agent running, no live connection to your data — a ranked list of findings across six domains, each one tracing back to a query you can run yourself. The queries are the receipts.
Every workspace, job cluster, and SQL warehouse priced from system.billing.usage — attributed by SKU, workload, and owner, not guessed at from a monthly total.
All-purpose compute left running, clusters sized for a peak that never comes, and auto-termination gaps — the spend that buys nothing.
Unity Catalog grants, external locations, and audit-log coverage — who can reach what, and where the trail goes quiet.
The tables nobody reads but everybody pays to keep — sizes, un-vacuumed history, and cold data still billed at hot-path rates.
Workflows ranked by real cost: retries, long tails, and the pipelines that quietly doubled without anyone opening the bill.
Model-serving endpoints left warm with no traffic, and the scale-to-zero settings that would have caught them.
Every finding traces back to a query you can run yourself. Here is the published evidence — the same methodology, worked in public.
A paragraph about your platform, one number that looks wrong, or just a hello — send an email. Replies within two working days.