How much of your Snowflake bill doesn't need Snowflake?
In one week I score every query in your account by observed spill, reconciled to your invoice,
then show you exactly what can move to a single DuckDB machine, what has to stay, and what you'd
save. Fixed fee. Inspectable, MIT-licensed SQL. No migration gamble.
If this is your week, the assessment pays for itself.
01Your Snowflake bill climbed again, and the renewal quote is already sitting in your inbox.
02You're fairly sure a lot of it is small dashboard and dbt queries that never needed a warehouse. You just can't prove it on paper.
03Finance wants the number down. You want it down without breaking a single report or carrying a new pager.
That's the whole job: prove what's movable, in writing, before anyone touches production.
How it works
Three steps. One week. A number you can take to your renewal.
01
You run our read-only export
We ship the exact SQL against ACCOUNT_USAGE. You run it. We never see your
credentials, and nothing touches production.
02
ducklens scores every query
MIT-licensed SQL grades each query on whether it truly fits one machine, by observed
spill, not scan size. Totals reconcile to your invoice.
03
You get a one-page map
A per-warehouse move / split / keep split, a saving
range, and a fixed-fee path. Or the honest answer: don't.
For the engineer who gets paged if it breaks
Built to survive a skeptical read.
Inspectable, not a black box
Every verdict comes from scoring.sql you can read line by line. Every threshold
is a named key you can challenge and re-run.
Reconciled to your invoice
The headline anchors to METERING_DAILY_HISTORY. If it doesn't match your
Snowsight bill, it's wrong, and you'll see exactly where.
It tells you when not to
Write-heavy, high-concurrency, regulated, and Snowflake-specific SQL all stay. Dollar-counted,
with the exact reason printed on every line.
Proven, not asserted
A parallel-run harness diffs old against new on a frozen snapshot before a single dashboard
repoints. Tamper one row and it goes red.
It runs on your laptop
Your ACCOUNT_USAGE export and your credentials never leave your machine, so there
is no third party to put through security review.
Snowflake renewal pricing rose roughly 15 to 22% year over year for existing
customers. Walking into that negotiation with a scored exit plan is leverage, even if you stay.
The offer
Start with the assessment. Only climb if the numbers justify it.
The free scan gives you the number. The paid Assessment gives you the
invoice-reconciled report, the go / no-go, and someone accountable for it.
Bill read-out
$1,500 · 90 minutes
You send the export (or we run the scan together), we walk through your fit
report live. Credited in full against the Assessment.
Assessment
from $7,500
fixed fee · ~1 week
The decision, on your real 90-day history, reconciled to your invoice.
One-page MOVE / SPLIT / KEEP split
Gross saving range plus payback
An honest go / no-go
If it doesn't find at least its own fee in savings, or it honestly tells you to stay, you don't pay. You keep the report either way.
Pilot recommended
$8–20k
~2–4 weeks · fixed fee
One real workload migrated to DuckDB and DuckLake, proven against your data.
Parallel-run validation harness
Before and after on $ and p50·p95
The reference stack, yours to keep
Migration
$20–50k
scoped from the pilot
We move what the assessment marked MOVE, and only that. You own the stack.
Every dashboard diffed green first
Runbook plus observability
Residual cleanly kept on Snowflake
The honest part
When not to migrate.
"Replace Snowflake" only holds for working sets under ~10 TB, under ~10 to 20
concurrent queries. If your workload is one of these, the assessment tells you to keep it, in writing.
Write-heavy
Continuous MERGE and COPY, high-frequency ingestion. DuckDB is single-writer, so sustained writes hit a ceiling Snowflake doesn't have.
High-concurrency serving
Dozens of concurrent dashboard users, multi-cluster scale-out. That's a serving workload, not single-machine batch.
Regulated & federated
Strict residency, multi-engine federation, Secure Data Shares. Above ~10 TB the honest answer is often Iceberg, not us.
Migrate the slice that genuinely fits, often less than half, and keep the rest.
A consultant who can't tell you what to keep can't be trusted with what to move.
How this is different
Not another optimizer.
vs savings-share optimizers
Optimizers shave 10 to 27% and bill you a share of it forever. This assessment answers a
different question: does 80% of your bill need a warehouse at all?
vs query-routing proxies
No proxy in your query path, no new SaaS dependency. A one-time assessment and a migration
you own.
vs Adaptive warehouses
Adaptive improves queries per dollar. The per-credit price is unchanged. The assessment shows
which queries cost almost nothing on hardware you own.
vs free cost dashboards
Dashboards show you spend by warehouse and call it a day. This scores every query against a
real engine and hands you a verdict: move it, split it, or keep it, with the dollars attached.
The deliverable
See the actual deliverable.
It is the one-page verdict your finance team gets. Invoice-reconciled, every threshold inspectable, the saving attached. Not a deck. The real thing you pay for.