The Bill Nobody OwnedWarehouse · medium · ~12 minLeadership wants team-level chargeback for the Snowflake bill by next quarter. A platform engineer proposes using the vendor's new native query-attribution view: "it's built in, zero maintenance."As a sanity check, you ran a controlled experiment first: one X-Small warehouse (1 credit/hour), 600-second auto-suspend, serving a BI tool that fires short sub-second queries. One full metered hour of activity.The numbers came back looking wrong — and how you explain them decides whether chargeback launches on a foundation teams will trust.Evidence

The experiment hour

Native view documentation notes

Chargeback context

Checkpoints

The hour metered 0.88 credits; queries ran for ~40 seconds; the native view attributed 0.43 credits (~49%) to those queries. What actually happened?

In that metered hour, roughly what percentage of the credits was idle time rather than query execution? (40s of queries in a fully-metered hour; ignore per-query overheads.)

For the chargeback design: who should pay for a warehouse's idle credits?

Write the attribution design you'd ship: the algorithm, the reconciliation invariant, and how you'd validate it before teams see a single number.