How to Reduce Snowflake Costs:
7 Proven Strategies

Snowflake's consumption-based pricing model is its greatest strength and its most dangerous quality. Pay only for what you use — but usage can spike 300% overnight and nobody knows why. This guide walks through the 7 highest-impact strategies data teams use to systematically reduce Snowflake spend.

Strategy 1: Warehouse Right-Sizing & Auto-Suspend

Snowflake warehouses are billed per second with a 60-second minimum. An XL warehouse costs 8× more than an XS, so running the wrong size is the single largest source of preventable credit waste.

The goal is to match warehouse size to the actual workload pattern — not to provision headroom for the busiest possible query.

Right-sizing steps

1. Query SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY for the P95 execution time and credits-per-query for each warehouse.

2. If a warehouse spends more than 70% of its time idle, it's a right-sizing candidate.

3. Drop one size (XL → L) and benchmark against the same workload for 3 business days.

4. Automate via Anavsan's anomaly detection to catch regression before it costs you.

Auto-suspend policy

Set AUTO_SUSPEND = 60 for interactive warehouses and AUTO_SUSPEND = 30 for ETL-style batch jobs. The default 10-minute suspend means you pay 10× longer than needed for short queries.

ALTER WAREHOUSE ETL_WH SET AUTO_SUSPEND = 30 AUTO_RESUME = TRUE;

---

Strategy 2: Query Result Cache Optimization

Snowflake caches query results for 24 hours. If the same query runs again with the same parameters and the underlying data hasn't changed, it returns the result for 0 credits.

The catch: the result cache is disabled if the query includes CURRENT_TIMESTAMP(), CURRENT_DATE(), or user-defined functions. Many dashboards unknowingly break caching with date-based filters.

Audit your cache hit rate

SELECT
  QUERY_TYPE,
  COUNT(*) AS query_count,
  SUM(CASE WHEN EXECUTION_STATUS = 'SUCCESS'
    AND QUERY_TAG LIKE '%RESULT_REUSE%' THEN 1 ELSE 0 END) AS cache_hits
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY
WHERE START_TIME >= DATEADD('day', -7, CURRENT_TIMESTAMP())
GROUP BY 1
ORDER BY query_count DESC;

A well-tuned environment typically sees 40–60% of dashboard queries served from cache. If yours is below 20%, audit your queries for timestamp injection.

---

Strategy 3: Clustering Keys & Partition Pruning

Snowflake automatically micro-partitions data by insertion order. When your most common filter (e.g., WHERE event_date BETWEEN...) doesn't align with the natural partition order, every query scans every partition — multiplying your credit usage.

Clustering keys tell Snowflake to reorganize partitions around your most-queried dimensions.

When to use clustering

• Table is larger than 1TB

• Queries consistently filter on the same 1–2 columns

• Partition pruning ratio is below 30% (check via Query Profile)

Strategy 4: Query-Level Optimization

Before any infrastructure change, audit your most expensive queries. The top 10 queries by credit consumption often represent 60–80% of your total compute cost.

Top patterns that silently drain credits

• Full table scans: SELECT * without filters on large tables.
• Repeated uncached queries: Dashboards running identical queries every 5 minutes.
• Cartesian joins: Missing JOIN conditions creating rows × rows explosions.
• DISTINCT on large result sets: Forces a sort and deduplication pass.
• Multiple CTEs scanning the same base table: Each CTE re-scans instead of materializing.

-- Find your top 10 credit consumers last 30 days
SELECT
  QUERY_TEXT,
  USER_NAME,
  WAREHOUSE_NAME,
  TOTAL_ELAPSED_TIME / 1000 AS elapsed_seconds,
  CREDITS_USED_CLOUD_SERVICES AS cloud_credits,
  PARTITIONS_SCANNED,
  PARTITIONS_TOTAL
FROM SNOWFLAKE.ACCOUNT_USAGE.QUERY_HISTORY
WHERE START_TIME >= DATEADD('day', -30, CURRENT_TIMESTAMP())
  AND EXECUTION_STATUS = 'SUCCESS'
ORDER BY CREDITS_USED_CLOUD_SERVICES DESC
LIMIT 10;

---

Strategy 5: Storage Lifecycle Management

Snowflake storage costs have three hidden contributors: active storage, Time Travel retention data, and Fail-Safe data (7 days, non-configurable). For organizations with large historical tables, storage can represent 20–30% of their total bill.

Time Travel optimization

The default DATA_RETENTION_TIME_IN_DAYS = 1 is appropriate for most tables. Large fact tables with high churn don't need 7-day or 90-day retention:

ALTER TABLE raw.events SET DATA_RETENTION_TIME_IN_DAYS = 1;
ALTER TABLE staging.temp_loads SET DATA_RETENTION_TIME_IN_DAYS = 0;

Audit which tables have retention set higher than needed using INFORMATION_SCHEMA.TABLES filtered by RETENTION_TIME.

---

Strategy 6: AI & Cortex Cost Attribution

Snowflake Cortex AI functions (AI_COMPLETE, AI_EMBED, AI_CLASSIFY) use AI Credits — separate from compute credits and billed at $2.00/AI Credit on-demand. A single AI_COMPLETE call with Claude Sonnet 4.6 at 8,000 tokens costs approximately $0.03 — small per call, but 500,000 calls/month is $18,000.

Most teams don't know their AI credit consumption by function, model, or owning team. The Cortex Cost Simulator gives you a directional estimate before you deploy.

Governance approach

1. Tag all Cortex workloads with a QUERY_TAG identifying the pipeline and team.
2. Monitor SNOWFLAKE.ACCOUNT_USAGE.METERING_DAILY_HISTORY filtered by SERVICE_TYPE = 'AI_SERVICES'.

3. Set per-team budgets using Resource Monitors scoped to AI credit consumption.

---

Strategy 7: Close the Accountability Loop with APEX

Strategies 1–6 are one-time fixes. Without an accountability loop, the same expensive patterns return — new ETL jobs, new dashboards, new engineers who don't know the right sizing policy.

The 4-stage governance loop that Anavsan APEX provides:

• Detect: Automated anomaly detection across warehouses, queries, storage, and Cortex workloads — alerts in minutes, not at month-end.
• Route: Every anomaly is assigned to the engineer accountable, with full query context, blast radius, and prior fix history.
• Simulate: Proposed optimizations are validated against historical workload patterns before production deployment — projected savings, not guesses.
• Prove: Savings are documented with before/after credit evidence " attribution, timestamp, owner " ready for leadership review and audit.

Summary: 7 Strategies at a Glance