Audit
We map your most expensive queries, your LookML, and your ad-hoc backlog. Read-only. Output is one shared doc.
Week 1
Data & Analytics · Looker + BigQuery
Your BigQuery is expensive and slow — fixable in 6 to 8 weeks.
We cut BigQuery costs 30 to 60% — measured on your own billing, not a slide. Then agents take ad-hoc reporting off your team. We work on your stack. No warehouse migration.
−30 to 60% 6 to 8 weeks measured on your billing
We work on your infrastructure. Your data never leaves it.
Typical reduction — your exact range comes from the audit, measured on your billing.
The tension · 01
Three questions you'd rather not answer in the next board review.
- 1 Finance asked why the cloud bill grew 70% this year. Do you have a clean answer?
- 2 Your dashboards take 8+ seconds to load. Do you know which queries are doing it?
- 3 Half your Explores haven't been opened in 90 days. Are you still paying for them?
If any of these landed, the cost is hiding in the same three places it always does.
The method · 02
01
02
03
04
Four steps. Six to eight weeks. On your stack, not a rebuild.
Refactor
Partitioning and clustering on heavy tables, LookML moved to derived tables with persistence, n+1 queries killed, scheduled jobs with no viewer turned off.
Weeks 2–6Measure
We compare against your baseline billing. The cost reduction goes in the SOW as a committed number — not an estimate.
Weeks 6–8Reporting agent
An agent takes the ad-hoc reporting queue. Your team answers questions in Slack, not in a dashboard request backlog.
HandoffWhat we don't do first: migrate your warehouse. Your CIO doesn't fight us.
The proof · 03
Harness · US mid-market
MadHive · US SaaS
Two stacks. Two numbers. Both measured on the client's billing.
Measurable cost reduction in 6 weeks
BigQuery optimization across a 6-week cycle — query refactor, LookML rework, FinOps.
Figures under NDA, shared on the call.
What we didn't do: no warehouse migration, no LookML rebuild on the client's side.
Agentic reporting live in the pipeline
Agentic ingestion and reporting running inside their pipeline — the reporting agent in production.
Details under NDA.
What we didn't do: no rip-and-replace of the existing pipeline.
Numbers shared under NDA on the next call. We'd rather show you the billing than quote a slide.
The audit · 04
Half your Explores are never opened.
You're paying for dead weight — and nobody gets an alert. Your senior engineers are doing FinOps by hand when they should be building product. A single misconfigured Explore can cost thousands a month in BigQuery, silently.
30–60% of your bill, found in ~15 minutes — read-only
15-minute read-only check. The dead weight, surfaced before finance asks.
| The manual status quo | The Health Check |
|---|---|
| Senior engineers auditing LookML by hand | 15 minutes, read-only, no engineer time |
| One bad query, a costly surprise, zero alert | We find it before finance does |
| Half your Explores billed, never opened | We name the dead weight |
15 minutes. Read-only access. Zero commitment. No findings, no charge.
Read-only, scoped to the minimum. Your data stays in your house — we never copy it out, and we never use it to train models.
The price · 05
Free Looker Health Check
Scoped audit
Fractional data engineering
The check is free. Here's what comes after.
15 min, read-only. We name the recoverable cost. No findings, no charge.
Free5 business days. One shared doc + a 90-minute session. The full map of your most expensive queries and Explores.
By scopeOngoing optimization + a reporting agent on top. Measured monthly against your billing.
Monthly retainer"Why not hire an in-house AI/data team?"
A serious in-house team is a heavy monthly cost and takes 6–9 months to be productive. We get in within 2 weeks; the first result is on your dashboard in 6–8. The infrastructure we leave is yours, documented, transferable. We accelerate your hire — we don't compete with it.
"How long until I see results?"
6 to 8 weeks to the first cut measured on your BigQuery bill. If it's not on your dashboard by week 8, we don't bill the next phase. That goes in the SOW.
"Isn't this just a big consultancy?"
They start in the high six figures and move in quarters. We start small and deliver the first measured result in 6–8 weeks. They send 40 people; we send 3 partners running a documented agent fleet.
"Is it safe to give you access to my stack?"
We operate on your infrastructure, not ours: your data stays where you have it. We ask for the minimum access needed (read-only for the Health Check), each component with the exact permission, and operations are logged. We don't use your data to train models. Everything is encrypted in transit and at rest by default of the cloud you already run. If we part ways, access is revoked. We don't claim SOC 2 or ISO yet — we'd rather tell you our actual architecture than sell you a badge.
Let's talk · 06
Read-only access. Your number in 15 minutes.
No pitch. We point read-only access at your billing and tell you what's recoverable. If there's nothing there, we tell you that too — and you owe us nothing. Your data never leaves your infrastructure.