Skip to main content

​
What is a Usage metric?

Usage metrics track how many queries are executed daily against a database or schema. They provide visibility into workload volume trends, helping teams understand growth patterns and detect sudden spikes or unexpected drops in query activity. Type: Automatic — Usage metrics are enabled on all connected database and schema objects as part of the Cost & Performance profile. No configuration is required to activate them. Usage metrics answer the question: How much query activity is this asset receiving daily? Prizm collects query count data from Snowflake’s ACCOUNT_USAGE.QUERY_HISTORY view on each performance job run (every 24 hours). Results are aggregated per database and schema, then trended over time. Anomaly detection alerts when daily query volume deviates significantly from the established baseline.

​
Where to find Usage metrics

Usage metrics live on every database and schema asset page. To view them:
  1. Navigate to Asset and open any database or schema asset.
  2. Click the Metric tab.
  3. Select the USAGE filter chip.
Usage metric in the USAGE filter on the Asset Metric tab

​
Metrics

​
How Usage monitoring works

Prizm reads from SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY on each performance job run and counts all queries in the last 24-hour window (with a 45-minute overlap for Snowflake latency). The daily total is stored per database and schema, trended over time, and evaluated against a statistical baseline. Query volume is expected to follow weekly patterns — higher on weekdays, lower on weekends, with end-of-month spikes for reporting workloads. Prizm’s statistical model adapts to these recurring patterns and alerts on deviations that fall outside the expected range.

​
Example use cases

Detecting a pipeline failure: A nightly ETL pipeline that runs hundreds of transformation queries stops executing due to an upstream dependency failure. Total Queries Per Day drops to near zero. Prizm fires a Usage alert in the next collection cycle — before any downstream dashboards show stale data. Identifying a runaway process: A misconfigured reporting tool begins polling a production database every 30 seconds instead of every 5 minutes. Total Queries Per Day spikes from 2,000 to 200,000. Prizm flags the volume anomaly, allowing the platform team to identify and throttle the offending connection. Tracking adoption growth: After a new self-service analytics platform is launched, Total Queries Per Day climbs steadily week-over-week as more analysts onboard. Prizm surfaces this trend, providing evidence to justify warehouse capacity increases ahead of projected growth.
A drop in Total Queries Per Day combined with no change in Query Success Rate often means the workload stopped running entirely — check pipeline scheduler logs rather than query error logs. If both drop together, a permission change or connectivity issue may be blocking access.

Configure

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