Skip to main content

​
What is a Credits metric?

Credits metrics track compute credit consumption at the database and schema level. They give data and platform teams direct visibility into cloud spend tied to query workloads, enabling cost governance, budget alerting, and cost attribution across teams and databases. Type: Automatic — Credits metrics are enabled on all connected database and schema objects as part of the Cost & Performance profile. No configuration is required to activate them. Credits metrics answer the question: How much compute is being consumed, and is it within expected range? Prizm collects credit consumption data from Snowflake’s ACCOUNT_USAGE.QUERY_ATTRIBUTION_HISTORY view on each performance job run (every 24 hours, with a 45-minute overlap for attribution latency). Credits are aggregated per database and schema, trended over time, and evaluated against a statistical baseline.

​
Metrics

​
How Credits monitoring works

Prizm reads from SNOWFLAKE.ACCOUNT_USAGE.QUERY_ATTRIBUTION_HISTORY on each performance job run, summing credits_attributed_compute for all queries in the last 24-hour window per database and schema. Each metric is independently trended and evaluated against a statistical baseline. Credit spend follows recurring patterns — end-of-month reporting spikes, nightly batch windows, and ad-hoc analysis peaks. Prizm’s statistical model adapts to these patterns and alerts on deviations that exceed the expected range.

​
Example use cases

Detecting a runaway query: A poorly optimized ad-hoc query performs a full table scan on a multi-billion-row table and runs for several hours. Total Credits Per Day spikes to 10× the baseline. Prizm fires a Credits alert within the next collection cycle, enabling the platform team to identify and terminate the offending query. Cost attribution for chargeback: Multiple teams share a Snowflake account with separate databases per department. Total Credits Per Day per database gives finance the data needed to allocate cloud costs to individual cost centers without requiring custom SQL reports. Isolating warehouse cost impact: A new warehouse is provisioned for an analytics team. Credits Used for that warehouse’s queries shows it consuming 3× the expected credits within the first week. Prizm surfaces the trend, prompting a review of the team’s query patterns before the monthly budget is exceeded.
Correlate Total Credits Per Day spikes with Queries metrics. If credits rise alongside Execution Time but Total Queries Per Day stays flat, the cause is likely a few slow, expensive queries — a query optimization problem. If both credits and query count rise together, the workload has genuinely grown.

Configure

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