Skip to main content

Snowflake FinOps

Snowflake FinOps is the operating model for making Snowflake cost visible, owned, and reducible. The hard part is not only finding expensive warehouses; it is proving which team, report, model, or application caused the spend and which optimization actually reduced it.

chukei adds wire-level cost attribution and signed savings evidence to Snowflake cost management.

Why attribution breaks down

Common attribution patterns all have tradeoffs:

PatternProblem
Warehouse per teamMultiplies idle cost and hides shared workloads
Manual QUERY_TAG disciplineDecays as tools, scripts, and users bypass conventions
ACCOUNT_USAGE onlyUseful after the fact, but weak for app/team ownership
BI-tool reportsGood within one tool, incomplete across the platform

chukei sees every query at the proxy, so it can attach team, user, app, dbt, and workload metadata at the point of execution.

What to attribute

Track these dimensions first:

  • Snowflake user and role;
  • warehouse;
  • client application or BI tool;
  • dbt model or job when available;
  • query fingerprint;
  • cache decision, rewrite decision, and suspend decision;
  • estimated and realized avoided credits.

Signed savings evidence

Savings claims need to survive finance review. chukei records conservative avoided-credit estimates and can export Ed25519-signed evidence bundles.

Use signed evidence when you need to show:

  • which queries were served from verified cache;
  • which warehouse suspend decisions avoided idle compute;
  • which rewrite rules changed query cost;
  • which team or tool created the savings;
  • whether the evidence bundle was modified after export.

Snowflake FinOps workflow

  1. Export or connect QUERY_HISTORY.
  2. Run replay to estimate savings before changing traffic.
  3. Pilot chukei with suggest-only suspend mode.
  4. Review cache hits, misses, attribution, and signed evidence.
  5. Move one workload at a time into enforced optimization.

Next steps