Snowflake cost spikes are usually visible first as a change in daily credits, then traceable to a warehouse, service, query, or storage category.
Use METERING_DAILY_HISTORY to locate the first abnormal day and identify the service type. Use WAREHOUSE_METERING_HISTORY to rank warehouse consumption, QUERY_HISTORY to inspect query volume and runtime, and QUERY_ATTRIBUTION_HISTORY to rank compute assigned to individual queries.
Common causes include a warehouse size change, new concurrent workloads, longer idle periods, a more frequent job schedule, a large scan, a serverless feature, or a replication and transfer event. A cost increase can occur without a SQL change if the same query runs more often or against more data.
After identifying the cause, compare the affected period with a representative baseline. Record the configuration change, the measured credit difference, and any latency or throughput effect. This separates a temporary event from a recurring cost problem.