A composite case study from 3 real Snowflake audits — 5 waste patterns, 30 days of fixes, and exactly which SQL statements moved the needle.
Lakshmi Kiranmai Guduru
Founder, CARTIEAI
A note on this story: The dollar figures and timeline below are based on a composite of 3 real Snowflake audits we've run. We've combined patterns and numbers to protect customer privacy. Every fix described is a real fix we've seen work.
A 200-employee SaaS company. Series B. $2M ARR. Snowflake spend: $87K/month ($1M/year). Growing 15%/quarter.
Their Head of Data, "Maya", reached out:
"Our Snowflake bill is killing us. I've tried following best practices but I have no idea where the waste actually is. Can we talk?"
We jumped on a 15-minute call. By the end, we'd identified 5 specific waste patterns worth a combined $40K/month. By day 30 of implementation, they'd cut their bill from $87K to $46K — a $487K annualized savings.
Here's exactly what we did.
Maya granted us a read-only Snowflake role (AUDITOR_RO). It took 5 minutes. We ran 8 diagnostic SQL queries against SNOWFLAKE.ACCOUNT_USAGE. The results were…instructive.
RT_ANALYTICS_WH (size: X-Large, multi-cluster: 1–4)We tested RT_ANALYTICS_WH on Medium for 1 week. After 7 days:
ALTER SCHEMA DEV_DB.PUBLIC SET DATA_RETENTION_TIME_IN_DAYS = 1;
ALTER DATABASE TEST_DB SET DATA_RETENTION_TIME_IN_DAYS = 0;
Result: $5K/month in storage savings. Took 5 minutes.
ALTER WAREHOUSE ML_RESEARCH_WH SET AUTO_SUSPEND = 60;
Then we audited the Tableau auto-refresh — 4 of 7 dashboards refreshed every 30 min "just because." Changed them to 4-hour refresh. Warehouse went from 24/7 active to ~6 hours/day. $8K/month saved.
8 dev/staging warehouses → 2. Sized down to Small with 60s auto-suspend. $5K/month saved.
CREATE OR REPLACE RESOURCE MONITOR ALL_WAREHOUSES_MONTHLY
WITH CREDIT_QUOTA = 12000
FREQUENCY = MONTHLY
TRIGGERS
ON 80 PERCENT DO NOTIFY
ON 95 PERCENT DO SUSPEND
ON 100 PERCENT DO SUSPEND_IMMEDIATE;
Every Monday, Slack bot posts top 5 warehouses by previous-week credits, anomalies, and owner per warehouse. This single change drove 10–15% additional savings just from awareness.
Top 5 cost-burning queries → sat with the analysts who wrote them. 3 of 5 had simple optimizations.
SELECT table_catalog, table_schema, table_name,
ROUND(active_bytes/1024/1024/1024, 2) AS gb, last_altered
FROM snowflake.account_usage.tables
WHERE table_type = 'BASE TABLE'
AND last_altered < DATEADD(day, -90, CURRENT_TIMESTAMP())
AND active_bytes > 1073741824
ORDER BY active_bytes DESC;
We found ~3TB of "leftover" tables. $2K/month saved.
| Phase | Days | Monthly Savings |
|---|---|---|
| Phase 1 (Quick wins) | 1–7 | $20K/month |
| Phase 2 (Mid-risk wins) | 8–15 | $14K/month |
| Phase 3 (Process wins) | 16–30 | $7K/month |
| Total | 30 | $41K/month |
Annualized: $487K/year in savings.
Maya's team had been focused on query optimization. The actual biggest waste was warehouse over-provisioning — a 5-minute fix.
Maya's team knew most of these fixes existed. But nobody had authority to change warehouse config.
The single highest-ROI fix wasn't a SQL statement. It was the Monday morning Slack post. Awareness compounds.
If your Snowflake bill is over $20K/month, you almost certainly have $5K–$30K/month of waste sitting in your account right now.
Take 60 seconds to find out which patterns apply to you: Run our free Snowflake Cost Health Score. No signup, no credentials, no email required.
Snowflake bills 10x more than people expect because the pricing is workload-based and most teams have never tuned a warehouse. This guide gives you the 9 patter…
THE FINOPS BRIEF
Built for finance & engineering teams who are tired of paying for cloud they don't use. No fluff. Just what works.
Unsubscribe anytime. We never sell your data.
ABOUT THE AUTHOR
Founder, CARTIEAI · Building in public
I'm building CARTIE AI to fix the cloud-cost problem I saw drain millions at companies I worked for — where engineering and finance kept talking past each other. If you liked this post, here's where I share unfiltered notes on building this in public: