Almost every over-sized AWS bill is over-sized in the same handful of ways. After optimizing hundreds of accounts, we work the wins in a consistent order — biggest, safest returns first. Here is the playbook, with the typical impact of each step.
1. Kill idle and orphaned resources
The fastest money on any bill is the resources nobody is using. Look for:
- Stopped EC2 instances still attached to EBS volumes you are paying for
- Unattached EBS volumes and old snapshots
- Unassociated Elastic IPs (AWS charges for IPs that are not in use)
- Idle load balancers and NAT gateways in dev accounts
Typical return: 5–15% of the bill, with essentially zero risk.
2. Migrate gp2 EBS volumes to gp3
gp3 is both cheaper and faster than gp2 for almost every workload — roughly 20% cheaper per GB with baseline performance included. It is the closest thing to free money in AWS. We wrote a dedicated guide on gp2 vs gp3 with the exact steps.
3. Right-size compute and databases
Most instances and RDS databases are provisioned for a peak that never comes. Pull two weeks of CPU, memory and connection metrics and drop over-provisioned resources down a size (or two). Right-sizing RDS and EC2 commonly returns 10–25% of their spend.
4. Tier and lifecycle your storage
S3 data that has not been touched in 30–90 days rarely needs to sit in Standard. Lifecycle rules to Infrequent Access or Glacier, plus enabling S3 Intelligent-Tiering for unpredictable access patterns, quietly cut storage cost with no application changes.
5. Commit to Savings Plans (only after 1–4)
Savings Plans and Reserved Instances are powerful, but commit to your optimized baseline, not today's inflated usage. Buying commitment before right-sizing just locks in the waste. Once steps 1–4 settle, a Compute Savings Plan on your steady-state usage can save another 20–40% on that portion.
6. Catch the next spike automatically
Optimization is not a one-time project — a single misconfigured job can undo a month of savings overnight. Real-time cost anomaly detection is what keeps the bill flat after you have flattened it.
Do it once, then keep it done
Working this list by hand across multiple accounts is a slog. CloudMonitor runs the whole playbook continuously: it finds the idle resources, ranks every saving by dollars, hands you the exact remediation steps, and alarms you the moment spend spikes again. See the cost optimization feature for how it works.