Most teams don't have a cost problem — they have a visibility and accountability problem. This playbook documents the exact framework we used to reduce AWS spend from $240K/month to $24K/month (90% reduction) for a Series B e-commerce client in Q2 2026.
The Problem: Invisible Waste at Scale
The client ran 200+ EC2 instances across 3 regions, 40+ RDS databases, and a sprawling S3 footprint with no lifecycle policies. Their monthly bill: $240,000. Engineering viewed cost as "finance's problem"; finance viewed it as "engineering's black box." No one owned the number.
"We knew we were overspending, but every attempt to cut costs broke something in production. We needed a framework, not a one-time cleanup."
Phase 1: Establish Visibility (Week 1-2)
1.1 Cost Allocation Tags — Mandatory, Not Optional
Enforced tagging policy via AWS Config Rules + Service Control Policies:
# Required tags on ALL resources
Environment: prod|staging|dev
Team: platform|payments|search|analytics
Owner: slack-handle
CostCenter: CC-001|CC-002
Project: revenue|internal|experimentNon-compliant resources auto-terminated after 72-hour grace period via Lambda + EventBridge.
1.2 CUR + Athena + QuickSight Dashboard
Enabled Cost and Usage Reports (CUR) to S3, built Athena tables, created QuickSight dashboards per team. Each team now sees their daily, weekly, monthly spend with drill-down to resource level.
1.3 Anomaly Detection
Configured AWS Cost Anomaly Detection with 15% threshold per team. Alerts route to team Slack + PagerDuty for immediate triage.
Phase 2: Right-Sizing & Waste Elimination (Week 3-6)
2.1 Compute Right-Sizing (EC2, RDS, ECS)
- Used Compute Optimizer + CloudWatch metrics (CPU, Memory, Network, Disk) over 14-day windows
- Downsized 142/200 EC2 instances (avg 65% CPU over-provision)
- Migrated 18 RDS instances to smaller classes + Aurora Serverless v2 for variable workloads
- Savings: $78K/month
2.2 Reserved Instances & Savings Plans
- Analyzed 90-day steady-state usage → purchased 3-year All Upfront RIs for baseline
- Applied Compute Savings Plans for flexible workloads (EC2 + Fargate + Lambda)
- Leveraged RI Marketplace to offload unused reservations
- Savings: $62K/month
2.3 Storage Lifecycle & Cleanup
- S3 Intelligent-Tiering for unknown access patterns
- Lifecycle rules: Standard → IA (30d) → Glacier (90d) → Deep Archive (365d)
- Deleted 47TB of unattached EBS snapshots & orphaned volumes
- Savings: $14K/month
2.4 Zombie Resource Hunt
Automated discovery via AWS Resource Groups + custom Lambda:
# Finds: stopped instances > 7d, unattached ENIs, idle ALBs,
# unused NAT Gateways, orphaned EIPs, dead CloudWatch log groupsPhase 3: Architecture-Level Optimization (Week 7-10)
3.1 Graviton Migration (ARM)
Migrated 89 eligible workloads to Graviton2/3 — 20-40% price/performance gain. Used EKS node groups with mixed instances for zero-downtime rollout.
3.2 Data Transfer Optimization
- Cross-AZ → Cross-region traffic reduced via Regional endpoints
- CloudFront for static assets (reduced S3 egress 87%)
- VPC Endpoints (Gateway + Interface) for S3/DynamoDB — eliminated NAT Gateway charges
- Savings: $18K/month
3.3 Serverless Where It Fits
Refactored 12 cron/batch jobs to EventBridge + Lambda (from EC2 + Airflow). Eliminated idle compute entirely.
Phase 4: Governance & Culture (Week 11-12)
4.1 Budget Guardrails per Team
AWS Budgets + SNS → Slack per team. Hard limit at 110% of forecast; soft alert at 90%. Teams own their forecasts.
4.2 FinOps Champions Program
Appointed 1 engineer per team as "FinOps Champion" — 2hrs/week dedicated to cost hygiene. Champions review anomalies, optimize their team's resources, report in weekly standup.
4.3 Monthly Cost Review Ritual
30-min cross-functional meeting: Engineering leads + Finance + Platform. Review top 10 cost drivers, anomalies, optimization wins, upcoming capacity changes.
Results Summary
| Category | Monthly Savings |
|---|---|
| Right-Sizing Compute | $78,000 |
| RI / Savings Plans | $62,000 |
| Storage Lifecycle | $14,000 |
| Zombie Resources | $8,000 |
| Data Transfer / VPC Endpoints | $18,000 |
| Serverless Migration | $12,000 |
| Total | $192,000/month |
Additional one-time savings from RI marketplace sales: ~$24K. Total 90-day impact: $216K.
Reusable Artifacts (Open Source)
We've packaged the automation as AWS Cost Optimization Playbook — available as a digital product:
- Terraform modules for tagging enforcement, CUR pipeline, anomaly detection
- Lambda functions for zombie hunting, right-sizing recommendations
- QuickSight dashboard templates (JSON export)
- FinOps Champions runbook + monthly review agenda
Want This for Your Organization?
We run a $5K Cloud Cost Audit — 2-week engagement, deliverable is a prioritized optimization roadmap with projected savings. Book a free discovery call: