5 Snowflake Cost Optimizations You Can Make This Week

Snowflake's consumption-based pricing model means your bill is directly tied to how efficiently you use the platform. The good news: most organizations can cut their Snowflake spend by 20-40% with five changes that take less than a week to implement.

1. Right-Size Your Warehouses

This is the highest-impact optimization. Snowflake warehouses come in sizes from X-Small to 6X-Large, with each size doubling the compute power and credit consumption of the previous one. A Medium warehouse costs 4x what an X-Small costs per hour.

The common mistake: setting all warehouses to Medium or Large "just to be safe." The reality: most interactive queries and light ETL jobs run just fine on X-Small or Small warehouses. The query might take 8 seconds instead of 4, but at one-quarter the cost.

Action this week: Run SHOW WAREHOUSES and check the avg_running time and query volume for each warehouse. Any warehouse that consistently runs queries in under 30 seconds on its current size can likely be downsized without noticeable impact. Start with development and staging warehouses -- there is zero business risk.

2. Set Aggressive Auto-Suspend

Snowflake warehouses continue to consume credits when they are running but idle. The default auto-suspend timeout is 10 minutes, but many teams increase it to avoid cold-start latency.

For most workloads, 1-2 minutes of auto-suspend is plenty. The cold-start penalty for resuming a suspended warehouse is typically 1-5 seconds. Unless your users are running interactive queries every 30 seconds, the idle cost of a 10-minute timeout far exceeds the occasional resume delay.

Action this week: Set auto-suspend to 60 seconds for all non-production warehouses and 120 seconds for production. Monitor for any user complaints about resume latency. Most teams find there are none.

3. Separate Workloads by Warehouse

Running ETL, reporting, and ad-hoc queries on the same warehouse means your interactive users compete with batch jobs for compute. It also means you cannot right-size: the warehouse needs to be large enough for your heaviest ETL job, even though 90% of queries are lightweight.

Create separate warehouses for each workload type:

• ETL/loading: Medium or Large, auto-suspend at 60 seconds, scheduled to match your pipeline cadence
• Reporting: Small, auto-suspend at 120 seconds, used by dashboards and scheduled reports
• Ad-hoc: X-Small or Small, auto-suspend at 60 seconds, used by analysts and data scientists

This separation lets you right-size each warehouse independently and provides clearer cost attribution.

4. Optimize Expensive Queries

A small number of queries often account for a disproportionate share of credit consumption. The 80/20 rule applies: 20% of your queries likely consume 80% of your credits.

Action this week: Query Snowflake's QUERY_HISTORY view to identify the top 10 most expensive queries by credit consumption over the past 30 days. Common culprits:

• SELECT * queries scanning entire tables when only a few columns are needed
• Queries missing partition filters, causing full table scans on large tables
• Materialized views or CTAS (CREATE TABLE AS SELECT) jobs that run more frequently than necessary
• Queries joining very large tables without appropriate clustering

Fix the top 5 and you will likely see a 10-20% reduction in total credit consumption.

5. Monitor and Alert on Credit Budgets

Snowflake provides built-in resource monitors that let you set credit budgets with automated alerts and actions. Most organizations do not use them until after a surprise bill.

Action this week: Create resource monitors for each warehouse with:

• A monthly credit budget based on the last 3 months of usage plus 20% buffer
• Email alerts at 75% and 90% of budget
• Auto-suspend action at 100% for non-critical warehouses

This provides a safety net against runaway queries, misconfigured pipelines, and the inevitable developer who accidentally runs a full table scan on your largest dataset.

The Compound Effect

Each of these optimizations delivers 5-15% savings individually. Combined, they typically reduce Snowflake spend by 20-40%. More importantly, they create a foundation for ongoing cost discipline. Once you have right-sized warehouses, separated workloads, and established credit monitoring, you have the visibility to catch future cost issues before they hit your invoice.