Cost Optimization~8-10 hours

Fix a $300K Cloud Bill

FinScale Analytics got a $300K Snowflake invoice and leadership wants answers. Detect the waste, cut compute, compact storage, and build the governance system that keeps costs down.

~8-10 hrsIntermediate-Advanced4 Parts

SnowflakeApache IcebergApache SparkdbtSQLPythonTerraformGrafanaFinOps FrameworkResource Monitors

Cost Reduction Roadmap

Before$300K/yr

Part 1$240K/yr

Part 2$180K/yr

Part 3-4$120K/yr

Total Savings$180K/yr (60%)

What You'll Build

Query Forensics

Mine ACCOUNT_USAGE to identify the top 20 costliest queries, profile execution plans, and map credit burn to teams.

Warehouse Right-Sizing

Analyze utilization patterns, implement auto-suspend/resume, and configure compute isolation per workload type.

Spark Repartition

Eliminate shuffle spills by repartitioning Spark jobs, tuning parallelism, and optimizing data distribution.

Iceberg Compaction

Compact small files into optimal 256MB targets, reducing file count by 85% and slashing storage scan costs.

Cost Dashboard

Real-time Grafana dashboard powered by dbt models tracking credit usage, storage growth, and cost anomalies.

FinOps Governance

Team-level chargeback model, budget alerts, resource monitors, and a governance playbook for sustained savings.

Progressive Build Path

Detect → Optimize → Prevent → Systemize. Each part targets a specific cost layer. Watch the bill drop from $300K to $120K as you work through all 4 parts. Complete in ~8–10 hours.

Detect — Query Forensics & Cost Baseline

2 hrs

Finance is asking questions and no one has answers. Mine Snowflake ACCOUNT_USAGE to find the top 20 most expensive queries, map credit burn to teams, and hand leadership a hit list with a dollar figure next to every line.

✅ You know exactly where the money is going

Cost baseline report + ranked hit list ($300K → $240K)

ACCOUNT_USAGEQUERY_HISTORYCost AttributionQuery ProfilingSnowflake Credits

Optimize — Warehouse Tuning & Compute Right-Sizing

2-3 hrs

Your hit list shows compute is the biggest cost driver — start there. Right-size warehouses, configure auto-suspend/resume, isolate workloads, and repartition Spark jobs to eliminate shuffle spills.

✅ 40% compute cost reduction verified

Optimized warehouse config + compute isolation policy ($240K → $180K)

Warehouse SizingAuto-SuspendResource MonitorsSpark RepartitionCompute Isolation

Prevent — Storage Compaction & Data Lifecycle

2-3 hrs

Even after fixing compute, your bill is still too high. Compact small Iceberg files, implement lifecycle policies, prune stale tables, and eliminate zombie clones quietly burning storage credits.

✅ 50% storage cost reduction verified

Compacted tables + lifecycle policies deployed ($180K → $120K)

Iceberg CompactionFile SizingStorage TieringData RetentionClone Cleanup

Systemize — Cost Dashboard, Chargeback & Playbook

2 hrs

Costs always creep back. Build a real-time FinOps dashboard with team-level chargeback so every team sees their bill. Add budget alerts and export a portfolio-ready FinOps playbook.

✅ Costs stay down — system governs itself

Live cost dashboard + FinOps playbook (portfolio export)

dbt MetricsCost DashboardChargebackBudget AlertsFinOps Governance

Total: ~8-10 hours across 4 parts

Built-in Data Sources

No external datasets needed. This project uses Snowflake's built-in ACCOUNT_USAGE schema — the same views every production Snowflake deployment already has.

SNOWFLAKE.ACCOUNT_USAGE

QUERY_HISTORY

Every query executed — credits, duration, warehouse, user

WAREHOUSE_METERING_HISTORY

Credit consumption per warehouse per hour

TABLE_STORAGE_METRICS

Active bytes, time travel, fail-safe per table

AUTOMATIC_CLUSTERING_HISTORY

Credits spent on automatic reclustering

STORAGE_USAGE

Daily storage bytes across stages and tables

ACCOUNT_USAGE views have a 45-minute latency. The project teaches you how to work with this constraint and supplement with INFORMATION_SCHEMA for real-time checks.

Tech Stack

Snowflake

Apache Iceberg

Apache Spark

dbt

SQL

Python

Terraform

Grafana

FinOps Framework

Resource Monitors

Why Cost Optimization?

#1 Demanded Skill

FinOps roles grew 35% year-over-year. Every company with a cloud bill needs engineers who can cut costs without cutting capabilities.

Immediate ROI

Unlike feature work that takes months to measure, cost optimization delivers quantifiable savings from day one — $180K saved is $180K saved.

Career Leverage

"I reduced our Snowflake bill from $300K to $120K" is the kind of bullet point that gets you promoted and recruited. Cost engineers have outsized career leverage.

Resume Ready

Complete this project and walk away with 4 quantified resume bullets that hiring managers love — real numbers, real impact, real tools.

Reduced Snowflake bill from $300K to $120K annually (60% savings) through query optimization, warehouse right-sizing, and storage compaction

Built automated FinOps dashboard with team-level chargeback tracking, processing 500K+ query events daily using dbt and Snowflake ACCOUNT_USAGE

Implemented Iceberg table compaction pipeline reducing small file count by 85% and cutting storage costs by 50% across 40TB data lake

Designed compute isolation strategy with auto-suspend policies, eliminating $8K/month in idle warehouse costs across 12 workload clusters

Prerequisites

SQL & Snowflake BasicsRequired
dbt FundamentalsRequired
Apache Spark BasicsHelpful
Apache Iceberg ConceptsHelpful

Related Learning Path

This project pairs perfectly with the Cost Optimization skill toolkit. Complete the modules first for foundational FinOps knowledge, or dive straight in if you have prior Snowflake administration experience.

View Skill Toolkits

→

Cost Optimization Learning Path

/learn/cost-optimization

→

What is Data Cost Optimization?

/guide/what-is-data-cost-optimization

Start Fixing the $300K Cloud Bill

FinScale Analytics is counting on you. Run the audit, cut the bill, ship the dashboard. Complete in ~8–10 hours.