Cost Attribution & Chargeback Models
When multiple teams share a Snowflake account, the inevitable question is: "who's spending what?" Without clear cost attribution, you end up with finger-pointing, budget surprises, and no accountability. In this lesson, we'll build a cost attribution framework from tagging through to chargeback reporting.
Why Cost Attribution Matters
In a single-team setup, the team owns the bill. In multi-team organizations, you need to know which team is spending what and which project is driving a cost spike. Without attribution, the platform team becomes the sole cost center for all analytics. That's neither fair nor sustainable.
Object Tagging for Cost Tracking
Snowflake's tagging system lets you attach metadata to any object. For cost attribution, we tag warehouses, databases, and schemas with ownership information.
CREATE TAG IF NOT EXISTS cost_center
ALLOWED_VALUES 'engineering', 'marketing', 'finance', 'data-science', 'platform';
CREATE TAG IF NOT EXISTS project
COMMENT = 'Project or initiative name for cost tracking';
Apply these tags to your objects:
ALTER WAREHOUSE analytics_wh SET TAG
cost_center = 'marketing', project = 'customer-360';
ALTER WAREHOUSE etl_wh SET TAG
cost_center = 'engineering', project = 'data-platform';
ALTER DATABASE marketing_db SET TAG cost_center = 'marketing';
Tags propagate through TAG_REFERENCES, making it possible to join cost data with ownership metadata.
QUERY_TAG for Workload Identification
While object tags cover warehouses and databases, QUERY_TAG lets you attribute costs at the individual query level — useful when multiple teams share a warehouse.
ALTER SESSION SET QUERY_TAG = '{"team":"marketing","pipeline":"campaign_metrics"}';
You can then aggregate QUERY_HISTORY by the tag:
SELECT query_tag, COUNT(*) AS query_count,
SUM(total_elapsed_time) / 1000 AS total_seconds
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY
WHERE start_time >= DATEADD('day', -30, CURRENT_TIMESTAMP())
AND query_tag IS NOT NULL AND query_tag != ''
GROUP BY query_tag
ORDER BY total_seconds DESC;
A structured JSON format (as shown above) is much more useful than freeform strings — it lets you parse and aggregate by team, pipeline, or any other dimension.
Credit Tracking with Metering Views
-- Warehouse compute costs by day
SELECT warehouse_name, DATE_TRUNC('day', start_time) AS usage_date,
SUM(credits_used) AS total_credits
FROM SNOWFLAKE.ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY
WHERE start_time >= DATEADD('month', -3, CURRENT_TIMESTAMP())
GROUP BY 1, 2 ORDER BY 1, 2;
For all metering including serverless features, use SNOWFLAKE.ACCOUNT_USAGE.METERING_HISTORY with its service_type column.
Serverless Feature Costs
These features consume credits without a user-managed warehouse, making them easy to overlook:
| Serverless Feature | Cost Driver | Monitoring View |
|---|---|---|
| Snowpipe | File volume and frequency | PIPE_USAGE_HISTORY |
| Auto-clustering | Table size and churn rate | AUTOMATIC_CLUSTERING_HISTORY |
| Materialized views | Maintenance on base table changes | MATERIALIZED_VIEW_REFRESH_HISTORY |
| Search optimization | Index maintenance | SEARCH_OPTIMIZATION_HISTORY |
| Replication | Data volume replicated | REPLICATION_USAGE_HISTORY |
| Serverless tasks | Task execution compute | SERVERLESS_TASK_HISTORY |
SELECT 'Snowpipe' AS feature, SUM(credits_used) AS credits
FROM SNOWFLAKE.ACCOUNT_USAGE.PIPE_USAGE_HISTORY
WHERE start_time >= DATEADD('month', -1, CURRENT_TIMESTAMP())
UNION ALL
SELECT 'Auto-clustering', SUM(credits_used)
FROM SNOWFLAKE.ACCOUNT_USAGE.AUTOMATIC_CLUSTERING_HISTORY
WHERE start_time >= DATEADD('month', -1, CURRENT_TIMESTAMP())
UNION ALL
SELECT 'Materialized Views', SUM(credits_used)
FROM SNOWFLAKE.ACCOUNT_USAGE.MATERIALIZED_VIEW_REFRESH_HISTORY
WHERE start_time >= DATEADD('month', -1, CURRENT_TIMESTAMP())
ORDER BY credits DESC;
Storage Cost Monitoring
Storage is billed separately from compute. Track it with:
SELECT table_catalog AS database_name, table_schema, table_name,
ROUND(active_bytes / POWER(1024, 4), 4) AS active_tb,
ROUND(time_travel_bytes / POWER(1024, 4), 4) AS time_travel_tb,
ROUND(failsafe_bytes / POWER(1024, 4), 4) AS failsafe_tb
FROM SNOWFLAKE.ACCOUNT_USAGE.TABLE_STORAGE_METRICS
WHERE active_bytes > 0
ORDER BY (active_bytes + time_travel_bytes + failsafe_bytes) DESC
LIMIT 25;
Time Travel and Fail-safe storage add up fast. A 1 TB table with 90 days of Time Travel and 7 days of Fail-safe can use 3-4x its active storage in total billed bytes.
Data Transfer Costs
Data transfer costs apply when data moves between Snowflake regions or to external cloud storage:
SELECT source_cloud, source_region, target_cloud, target_region,
SUM(bytes_transferred) / POWER(1024, 4) AS tb_transferred
FROM SNOWFLAKE.ACCOUNT_USAGE.DATA_TRANSFER_HISTORY
WHERE start_time >= DATEADD('month', -3, CURRENT_TIMESTAMP())
GROUP BY 1, 2, 3, 4
ORDER BY tb_transferred DESC;
Cross-region and cross-cloud transfers are the most expensive.
Building a Chargeback Report
This query joins warehouse metering data with tags to produce a cost-by-team breakdown:
WITH warehouse_tags AS (
SELECT object_name AS warehouse_name, tag_value AS cost_center
FROM SNOWFLAKE.ACCOUNT_USAGE.TAG_REFERENCES
WHERE domain = 'WAREHOUSE' AND tag_name = 'COST_CENTER'
),
monthly_credits AS (
SELECT warehouse_name, DATE_TRUNC('month', start_time) AS usage_month,
SUM(credits_used) AS total_credits
FROM SNOWFLAKE.ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY
WHERE start_time >= DATEADD('month', -6, CURRENT_TIMESTAMP())
GROUP BY 1, 2
)
SELECT wt.cost_center, mc.usage_month,
SUM(mc.total_credits) AS credits_consumed,
SUM(mc.total_credits) * 3.00 AS estimated_cost_usd -- adjust per contract
FROM monthly_credits mc
JOIN warehouse_tags wt ON mc.warehouse_name = wt.warehouse_name
GROUP BY 1, 2 ORDER BY 2, 4 DESC;
For a complete view, combine this with serverless and storage costs attributed via the same tagging hierarchy.
Dashboard Patterns for Finance Teams
Structure your chargeback output to answer:
- Monthly total by cost center — headline number for each team's budget review
- Trend over last 6 months — is spending growing, stable, or declining?
- Top 10 most expensive warehouses — where are the big numbers?
- Serverless vs. warehouse split — are hidden serverless costs growing?
A scheduled task refreshing a summary table daily works well as backing data for dashboards.
Best Practices for Org-Wide Cost Governance
Establish tagging standards early. Define your tag taxonomy before teams start creating warehouses.
Make cost visibility a habit. Share chargeback reports monthly with team leads.
Automate untagged object detection:
SELECT w.warehouse_name
FROM (SELECT warehouse_name FROM SNOWFLAKE.ACCOUNT_USAGE.WAREHOUSES
WHERE deleted_on IS NULL) w
LEFT JOIN (SELECT object_name FROM SNOWFLAKE.ACCOUNT_USAGE.TAG_REFERENCES
WHERE domain = 'WAREHOUSE' AND tag_name = 'COST_CENTER') t
ON w.warehouse_name = t.object_name
WHERE t.object_name IS NULL;
Don't optimize prematurely. Track costs for at least one billing cycle before making cuts.
Key Takeaways
- Tag every warehouse and database with cost center and project metadata from day one.
- Use structured
QUERY_TAGvalues to attribute shared warehouse costs to specific teams. - Monitor serverless feature costs separately — they're easy to miss and can grow quickly.
- Build automated chargeback reports that join metering history with tag references.
- Cost governance is a cultural practice, not just a technical one. Visibility drives accountability.