I find the 20–40% you’re burning in Snowflake.
Two weeks. Fixed fee. Read-only access to metadata, never your data. If I don’t find at least three times my fee in annual savings, you pay nothing.
I’ll tell you in a paragraph whether there’s anything worth chasing. Free, no call required.
or book 20 minutes →The Audit
How it works
Ten working days, from access to readout. Your time: one 45-minute kickoff, one 60-minute readout. Nothing in between.
What I look at
- Warehouses: sizing, auto-suspend, idle time, and compute running out of hours
- The queries and jobs that cost the most, full scans, failing and retried jobs, and result-cache misses
- Storage: tables nobody reads, time-travel retention, and clustering that costs more than it saves
- Serverless features — Trust Center, tasks, compute pools — which rarely sit in anyone’s budget
- Power BI and Streamlit apps: what they cost on Snowflake, and whether they keep warehouses awake
- Where the money goes: the bill split by warehouse, team and environment
What you get
A savings ledger — every finding priced in euros, with effort in days, risk, and how confident I am. Built to be forwarded to finance.
A findings report — one page per material finding: root cause, the fix as ready-to-run SQL or config, and what could go wrong doing it. Ordered into this week, this quarter, and needs a decision, so you can bank the easy savings the same week.
A recorded readout, and a free check-in after 30 days.
What I need
Read-only access to platform metadata. No customer, employee or business data at any point.
- Snowflake: a role with IMPORTED PRIVILEGES on SNOWFLAKE.ACCOUNT_USAGE
- Optional: dbt manifest.json, your BI tool’s usage log, and twelve invoices for the trend line
Pricing
| Monthly platform spend | Fee | Note |
|---|---|---|
| under €8,000 | — | Not worth it. I’ll tell you up front. |
| €8,000 – 25,000 | €6,500 | Standard scope |
| €25,000 – 80,000 | €9,500 | Standard scope |
| over €80,000 | €14,000 | After a short scoping call |
If the annualised savings I identify come to less than three times the fee, you get the fee back in full.
What I’ve built
design2code — turns design screenshots into working HTML, CSS and JavaScript. 677 stars, 103 forks, and a hosted product at design2code.dev.
auto-pilot-computer — GPT-4 Vision driving a computer, in Rust.
Day to day I build production data platforms — pipelines, warehouses, dbt modelling — and the interfaces that sit on top of them. The combination is unusual, and it’s why the fixes in my reports are ones your team can actually ship.
Also
Databricks or BigQuery? Not yet — the Audit is Snowflake only for now. Email me and I’ll let you know when that changes.
Data interfaces — embedded analytics, internal tools, and governed chat over a warehouse. Where your data team can model it and your product team can design it, but nobody owns the layer in between. Ask me.
Contact
Email is best
mostafa@mostafasadeghi.com
Or put 20 minutes in my calendar. I reply within a day.