⚡ Quick Wins (Week 1)
🔍 Visibility & Discovery
- Enable cost allocation tags on ALL resources (Project, Environment, Owner, CostCenter)
- Set up Azure Cost Management / AWS Cost Explorer / GCP Billing dashboards
- Identify untagged resources and assign ownership
- Set up daily/weekly cost anomaly alerts (>10% variance)
🗑️ Eliminate Waste
- Delete unattached disks and snapshots older than 30 days
- Remove unused Elastic IPs / Public IPs
- Terminate stopped VMs that haven't run in 30+ days
- Delete orphaned load balancers with no backends
- Clean up old container images in registries
📉 Right-Size Immediately
- Identify VMs with <10% average CPU utilization
- Downsize over-provisioned databases (check DTU/vCore usage)
- Review and reduce over-allocated storage
💰 Short-Term Optimizations (Month 1)
Commitment Discounts
- Analyze 12-month usage patterns for stable workloads
- Purchase Reserved Instances for always-on production VMs (save 40-72%)
- Evaluate Azure Savings Plans / AWS Savings Plans for flexibility
- Set up RI utilization monitoring and alerts
Architecture Quick Wins
- Enable auto-scaling for variable workloads
- Implement scheduled scaling (scale down nights/weekends for non-prod)
- Move dev/test to spot instances / preemptible VMs (save 60-90%)
- Enable storage tiering (hot → cool → archive)
📊 Potential Savings by Service
| Service | Optimization | Savings |
|---|---|---|
| Compute | Right-sizing | 20-40% |
| Compute | Reserved Instances | 40-72% |
| Compute | Spot/Preemptible | 60-90% |
| Storage | Tiering (cool/archive) | 50-80% |
| Database | Reserved capacity | 30-65% |
| Network | CDN for static | 40-60% |
🚨 Red Flags Checklist
If you answer "yes" to any of these, you're likely overspending:
- No one knows what the cloud bill will be next month
- Developers can provision resources without approval
- You don't know which team/project owns which resources
- Production and dev environments are sized the same
- You haven't reviewed Reserved Instance coverage in 6+ months
- You're running 24/7 workloads on on-demand pricing