Skip to main content

Query Costs

The Query Costs page at /finops/query-costs attributes warehouse DBUs to individual SQL queries. It's the answer to "which queries are burning the most money?".

Where the data comes from​

  • system.query.history — per-query metadata (user, query text, start/end, status)
  • Warehouse DBU rate × query duration on the warehouse — produces effective cost
  • Photon multiplier applied where applicable

Top expensive queries​

The default sort is cost DESC over the last 24 hours. Each row shows:

  • Query ID — click to view the full text and execution plan
  • User — who ran it
  • Warehouse — which warehouse executed
  • Duration — wall-clock
  • DBUs
  • Cost
  • Status — success / failed / cancelled
  • First seen — first timestamp this query text appeared (helps spot recurring offenders)

Query text grouping​

Toggle Group by query text to collapse identical queries into one row with aggregates:

  • Total runs
  • Total cost
  • Avg cost / run
  • p95 duration
  • Unique users

This is how you find queries running thousands of times that nobody owns.

Filters​

  • Date range
  • User / team
  • Warehouse
  • Min cost threshold (e.g. $10+)
  • Tags

Optimisation hints​

For each expensive query, the page surfaces hints from Recommendations:

  • Predicate pushdown missing — column not in WHERE would benefit from ZORDER
  • Full scan on partitioned table — partition pruning didn't apply
  • Sub-optimal join order — large table on left
  • Photon would help — query type benefits from Photon

Click View plan to see the execution plan with the suggested fix annotated.

API​

GET /finops/query-costs?from=2026-04-01&to=2026-04-30&min_cost=10
GET /finops/query-costs/{query_id}/plan