Jonathan Kini

Your warehouse bill keeps growing.
Nobody can explain why.

I trace BigQuery and Snowflake spend back to the queries, pipelines, models and dashboards driving it—so you know where the money goes and what to optimize first.

Book a discovery call →See what you'll receive

30 minutes. No commitment. No sales pitch.

Jonathan Kini

Jonathan Kini

9+ years building and operating production data platforms. Snowflake, BigQuery, dbt, Airflow, Data Vault — owned on call, not audited from the outside.

I've run the systems this investigation reads: billing exports, query history, INFORMATION_SCHEMA, dbt artifacts, orchestration logs, BI usage. I know what each source can prove, what it cannot, and where two of them disagree.

One person can run this because the scope is narrow on purpose: two warehouses, metadata only, a fixed set of questions, seven days. Depth on a small surface, not coverage on a large one.

Data Vault taught me one habit: every record carries where it came from. Cost works the same way. If you can't trace a euro back to a query, an owner and a decision, you can't govern it.

Read my story →

WHERE I'VE WORKED ON DATA PLATFORMS

  • BKW
  • Adiktev
  • CCF Group
  • Servier
  • Solocal
  • Les Echos
  • La Poste
  • Caisse des Dépôts
  • La Redoute

The problem

Native tools show costs. They don't explain them.

BigQuery and Snowflake have cost explorers. They tell you how much you spent by project, by day, by slot. But they don't answer the questions that matter.

Without attribution, a cost number is just accounting.

  • Which dashboard is costing €3,000/month and gets opened twice a week?
  • Which dbt models run hourly but nobody needs that freshness?
  • Which team's ad-hoc queries are driving most of the compute?

  • Attribution, not aggregation

    Instead of cost per project, cost per decision.

  • Cross-layer analysis

    Follow a euro from query to dashboard.

  • Prioritized

    Know what deserves engineering time.

  • Built for governance

    Every recommendation identifies the decision required.

The product

Cost Discovery

Stop optimizing blindly.

Understand where your data platform costs come from before deciding what to optimize.

  • Expose ownership gaps.
  • Understand business context.
  • Know what to optimize first.

The executive investigation for BigQuery and Snowflake

Before you approve another optimization project, know where the money actually goes.

Engagement fee€3,500Excl. VAT

What it is

A focused investigation that traces data platform spend back to the queries, pipelines, models, dashboards and usage patterns driving it.

Every material driver is followed through the same four layers — usage, flow, design, ownership — until the cause is established from platform evidence or the gap is recorded as unknown.

Included

  • Executive Report
  • Executive Walkthrough
  • 30-Day Action Plan

Delivery

Fixed scope

7-day delivery

2 × 20-minute team interviews

Traceability guarantee

Every material finding cites its evidence: the source it was read from, the period it covers, and the workload or usage behind it. Billing exports, query history, INFORMATION_SCHEMA, dbt artifacts, BI usage — named finding by finding, so your team can re-run the check without me.

Nothing is estimated from a benchmark or inferred from what usually happens. What the platform cannot prove is written down as unknown rather than filled in.

If I can't establish the evidence behind the findings, you don't pay.

Book your Cost Discovery →

Starts with a 30-minute call. No commitment.

Stack covered

  • BigQuery
  • Snowflake
  • dbt
  • Airflow
  • Looker
  • Fivetran

Deliverables

Here's everything you'll receive.

Three deliverables. One executive report. One working session. One plan your team can execute.

Illustrative finding

A €18,400 increase. 73% traced to two dbt models switched to full refresh.

Owned by one team. Triggered by one deployment. Visible in the bill, unexplained by native cost tools.

How that conclusion was reached

  1. 01

    Problem. The monthly bill rose against the previous month. The native cost explorer attributed the rise to a single project holding dozens of pipelines, with no line item large enough to explain it.

  2. 02

    Investigation. Query history for both months, grouped by job label and service account, ranked by bytes scanned. Two dbt models carried most of the delta. Their run history was cross-checked against the dbt manifest and the deployment log to date the change.

  3. 03

    Root cause. Both models had moved from incremental to full refresh in one deployment, while keeping an hourly schedule. Every run rescanned the entire history instead of the new partitions.

  4. 04

    Recommendation. Restore incremental materialization on both models, and reset the schedule to the cadence the downstream dashboards are actually read at. One owning team, one deployment, reversible in a day.

  5. 05

    Expected outcome. Bytes scanned by those two models return to their pre-deployment baseline. The plan names the figure to re-measure, the exact query that produces it, and the date it gets re-measured.

  6. 06

    What stayed unknown. Nothing in the platform explained why the materialization was changed. That question went to the owning team and was recorded as open in the report, rather than answered by inference.

Illustrative, not a client case. What repeats from one engagement to the next is the sequence: no step is skipped, every step names the evidence it rests on, and anything the platform cannot prove leaves the report as an open question.

Four layers. Cost drivers hide differently in each.

ANALYSIS DEPTH →

  • Usage

    Who runs what, how often

  • Flow

    How data moves through pipelines

  • Design

    Model structure, refresh logic

  • Ownership

    Whether accountability exists for each cost

Deliverable 01

Executive Report

The document you'll forward to your Head of Data or CFO after the engagement.

Instead of dozens of disconnected dashboards, a single executive document that explains where costs come from, what they support, where accountability exists or is missing, and what should happen next.

Every material finding traces back to platform evidence, a time period and the workload or usage behind it. Where ownership cannot be established, the report makes that gap explicit. Nothing comes from a generic benchmark.

  • 7 pages
  • One page per chapter
  • PDF, yours to keep
Chapter 01: Executive Summary, where the spend goes and what to do first

Most of the bill comes from a few workloads. Several have no owner. Some feed dashboards nobody opens.

One page. Six numbers. One priority.

Cost Discovery Report · Confidential · IllustrativePage 01 of 07

Deliverable 02

Executive Walkthrough

A working session where we go through the findings together and agree on what deserves attention first.

Most audits end with a document. This one ends with a decision log.

Chapter 06: Executive Walkthrough, the decisions are made together

60 minutes. You leave with priorities agreed, decisions identified and accountability gaps made explicit—not a document to read later.

Findings, priorities, decisions, next actions.

Cost Discovery Report · Confidential · IllustrativePage 06 of 07

Deliverable 03

30-Day Action Plan

A practical roadmap your team can start executing the next morning.

  1. 01Accountability clarified for every action
  2. 02A sequence, not a backlog
  3. 03A measurement date, and the exact figure to re-check
  4. 04No implementation from me unless you ask

Thirty days later you re-measure the same figure. Either it moved, or you know why.

Chapter 07: 30-Day Action Plan, the page you hand your team on Monday

Accountability clarified. Actions sequenced. Measurement dates set. A plan your team can execute, and that you can hold them to.

Four windows across 30 days.

Cost Discovery Report · Confidential · IllustrativePage 07 of 07

How it works

A 7-day process from connection to clarity.

No workshops every day. No endless consulting engagement. One week. One report. One executive walkthrough.

Before we start

Your team

Create a read-only IAM role and share the connection details. About 15 minutes of your team's time, once.

My side

I analyze the metadata, map costs to workloads, dashboards and usage patterns, and prepare the investigation before day one.

The 7 days

  1. Day 1

    Connection

    15 min from your team

    You create a read-only IAM role. I collect the metadata.

  2. Day 2

    Analysis

    Nothing from your team

    I reconcile billing exports with query history and rank workloads by what they scan.

  3. Day 3

    Investigation

    2 × 20 min from your team

    I investigate the main cost drivers and validate their purpose, usage and business context with the people closest to them.

  4. Day 4

    Recommendations

    Nothing from your team

    I rank cost drivers by impact, effort, risk and confidence. Each one carries the evidence it rests on.

  5. Day 5

    Report

    Nothing from your team

    I compile the executive report (7 pages).

  6. Day 6

    Preparation

    Nothing from your team

    I prepare the walkthrough with your team.

  7. Day 7

    Walkthrough

    60 minutes from your team

    I present findings and align on next steps.

  • Privacy by design

    I only access metadata and query logs. Your actual data is never viewed or touched. Read-only access to INFORMATION_SCHEMA is enough. The access is yours to revoke, and everything I run against your platform lands in your own query history, so the investigation can be audited as it happens.

  • You decide what to implement

    Every recommendation carries an effort estimate. You choose what fits your priorities. I do not implement anything myself.

Fit

Is Cost Discovery the right fit?

Cost Discovery is probably the right fit if you recognize yourself in at least three of these situations.

A great fit

  • Your bill keeps growing and nobody can explain why.
  • Your cost dashboards don't say what deserves attention first.
  • Your engineers investigate costs more than they act on them.
  • You need clarity before investing in optimization work.

Not the right fit

  • Your platform is still small and costs aren't a concern yet.
  • You want hands-on implementation, not an assessment.
  • You already know exactly what needs to be optimized.
  • Your priority is cloud FinOps outside the data platform.

Not sure? Score your platform in 12 questions →

Common findings

Every platform looked different. The same problems didn't.

Not because the numbers are missing. Because the connection between spend, usage and accountability is.

  • Unowned workloads

    Still running for a team that no longer exists.

    Traced from job labels and service accounts, checked against who is still there.

  • Wrong refresh cadence

    Hourly pipelines feeding a weekly decision.

    Traced from run frequency in the orchestration logs against last-access times downstream.

  • Hidden downstream usage

    One table, forty dashboards, no map.

    Traced from query-history lineage and BI usage logs, not from the documentation.

FAQs

Questions?

Everything you need to know before booking your Cost Discovery.

Book a discovery call →

  • Cost attribution depends on metadata that is specific to each platform: query history, credit or slot accounting, information schema layout. I know these two well enough to go deep in 2 days. dbt, Airflow, Looker and Fivetran are read as context around them, not audited on their own. On a third warehouse I would be learning on your budget.

From the blog

Written while doing the work

Before you optimize your platform, understand it.

Book a discovery call →

30 minutes. No commitment. No sales pitch.