Skip to main content

​
What is a Queries metric?

Queries metrics measure the volume, efficiency, and reliability of query workloads running against a database or schema. They surface query performance degradation, execution bottlenecks, and reliability issues before they escalate into user-facing failures. Type: Automatic — Queries metrics are enabled on all connected database and schema objects as part of the Cost & Performance profile. No configuration is required to activate them. Queries metrics answer the question: Are queries running successfully, at expected volume, and within acceptable execution times? Prizm collects query execution data from Snowflake’s ACCOUNT_USAGE.QUERY_HISTORY view on each performance job run (every 24 hours, with a 45-minute overlap for latency). Results are aggregated per database and schema, trended over time, and evaluated against a statistical baseline.

​
Metrics

​
How Queries monitoring works

Prizm reads from SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY on each performance job run, aggregating each metric across all queries in the last 24-hour window per database and schema. Each metric is independently baselined and evaluated against a statistical model that accounts for recurring patterns — higher weekday volumes, end-of-month reporting spikes, and nightly batch windows.

​
Example use cases

Detecting schema breakage impact: After a schema migration, downstream queries begin referencing a column that no longer exists. Query Success Rate drops from 99% to 68% within the next performance collection cycle. Prizm alerts, giving the data team a quantified measure of blast radius before users report errors. Catching execution regressions: A dbt model refactor introduces an expensive cross-join. Average Execution Time rises from 200ms to over 4 seconds for queries on the affected schema. Prizm surfaces the anomaly within the next collection window, before it causes SLA misses. Spotting pipeline outages: Total Queries Per Day drops to near zero on a schema that normally sees several hundred queries per day. Prizm fires an alert, prompting the team to investigate an upstream pipeline failure that silently stopped feeding the schema.
Use Total Queries Per Day alongside Average Execution Time and Query Success Rate together to triage issues quickly. A drop in query count with a rising success rate suggests a pipeline stopped sending queries. A drop in success rate with stable query count points to schema or permission breakage. Rising execution time with stable count and success rate indicates a query efficiency regression.

Configure

Set thresholds, view the metric detail page, and understand Queries alerts.