ThisCalculator

Everyday

Cut Your Cloud Bill with a FinOps Checklist

Updated · Standard formula, shown below

This guide walks you through a practical FinOps checklist that helps you find and eliminate wasted spending on AWS, Azure, or Google Cloud. You’ll go through a series of audits—from right-sizing and scheduling to storage and reserved capacity—and end with concrete actions that reduce your monthly invoice. No prior FinOps certification needed, just access to your cloud console and a willingness to question every line item.

What You’ll Need

  • Read-only or Billing access to your cloud account (AWS, Azure, or GCP).
  • A spreadsheet or notes app to track findings.
  • Approval from a manager or budget owner before making changes that affect production.
  • Optional: a tag or naming convention already in place—if not, you’ll start one.

Step-by-Step

  1. Pull your cost report and group by service. In AWS, open Cost Explorer; in Azure, use Cost Management; in GCP, go to Billing > Reports. Filter to the last 30 days. Sort by cost descending. Note the top 5 services—they usually account for 80% of spend.
  2. Identify idle or orphaned resources. Look for running instances or VMs with near-zero CPU or network activity. In AWS, use Compute Optimizer or check CloudWatch metrics; in Azure, use Advisor; in GCP, use Recommender. Stop or delete these. Also check for unattached Elastic IPs, unused load balancers, and orphaned disks.
  3. Right-size over-provisioned instances. For each running VM, compare its average CPU, memory, and network utilization over the last 14 days against its instance family. If CPU is consistently below 10% and memory below 20%, downgrade to a smaller size. Use the cloud provider’s sizing recommendations, but verify manually for your workload.
  4. Schedule non-production workloads to shut down during off-hours. For dev, test, and staging environments, create a schedule that stops them at 7 PM and starts at 7 AM on weekdays. Use AWS Instance Scheduler, Azure Automation, or GCP Cloud Scheduler with a startup/shutdown script. This can cut those costs by up to 65%.
  5. Review storage tiers and delete stale data. Check your object storage (S3, Blob, GCS) for objects older than 90 days that aren’t accessed frequently. Move them to infrequent access or archive tiers. Also list all snapshots and backups—delete any that are more than 30 days old unless required for compliance.
  6. Leverage reserved capacity or savings plans. For predictable, always-on workloads, purchase Compute Savings Plans (AWS), Reserved Instances (Azure), or Committed Use Discounts (GCP). Start with a 1-year term for a small portion (e.g., 20%) of your baseline usage. This yields immediate 20-40% savings.
  7. Check data transfer and egress costs. Look for large data transfers between regions or out to the internet. Move data within the same region when possible. Use CloudFront or a CDN for repeated downloads. If you have a hybrid setup, consider a direct connect or VPN to reduce egress charges.
  8. Enable budget alerts and cost anomaly detection. Set up a monthly budget at 80% of your expected spend. Configure alerts at 50%, 80%, and 100% of that budget. In AWS, use Budgets; in Azure, use Cost Alerts; in GCP, use Budgets and Alerts. This prevents bill shock.
  9. Tag all resources with cost centers. Create tags like project, owner, environment. Apply them to every resource. Then use the tag in your cost report to allocate spend. This makes it easier to spot waste per team.
  10. Review your monthly bill line by line. Once you have tags, go through each line item. Question anything that doesn’t have a clear owner or purpose. If a resource isn’t tagged, it’s a candidate for deletion.
  11. Automate cleanup of temporary resources. Use a script or serverless function to automatically stop instances that have been running for more than 7 days with low utilization, or delete snapshots older than a threshold. This prevents recurrence.
  12. Repeat monthly. Cloud usage changes. Set a recurring calendar reminder to run this checklist every month. The first pass will find the biggest wins; subsequent passes catch new waste.

Common Mistakes

  • Stopping production instances accidentally. Always check the environment tag before scheduling or stopping. Use separate accounts or VPCs for production.
  • Right-sizing based on peak only. If you look at the max CPU, you’ll never downsize. Use average and p95 metrics.
  • Deleting backups that are needed for compliance. Always verify retention policies before deleting snapshots. When in doubt, keep the most recent 30 days.
  • Ignoring data egress costs. Moving data between regions or to the internet can cost more than compute. Always check egress before migrating workloads.
  • Buying reserved capacity for variable workloads. If your usage fluctuates wildly, a savings plan with a low commitment is safer than a full reserved instance.

Tips

  • Start with the top 3 cost centers—fixing those gives the fastest ROI.
  • Use the cloud provider’s free cost management tools before buying third-party software.
  • Set up a weekly report of instance count and cost to catch runaway resources early.
  • When in doubt, stop an instance for 24 hours and see if anyone complains—if not, delete it.
  • For storage, use lifecycle policies to automatically move objects to colder tiers after 30, 60, or 90 days.

Troubleshooting

What if I can’t see cost data? Make sure you have the required IAM role (e.g., ce:GetReservationUtilization in AWS, Cost Management Reader in Azure, or Billing Viewer in GCP). If you still can’t see, ask your admin for read-only billing access.

What if I stopped a VM and a service went down? Restart it immediately. Then check the dependency—maybe it was needed for a scheduled job. Add a tag like critical=true to prevent future stops.

What if my savings plan doesn’t cover the right instance type? Savings plans automatically apply to any instance in the same family and region. If you bought a specific reserved instance, you may need to exchange it (AWS allows some exchanges) or wait for expiration.

What if the cost report shows a spike but I can’t find the resource? Use the filter by tag or resource ID. If it’s a data transfer charge, look at the source and destination IPs. Sometimes it’s a misconfigured load balancer or a forgotten NAT gateway.

What if my budget alert doesn’t fire? Check that the alert is set to actual spend, not forecasted. Also verify the email address is correct and not in spam.

Frequently Asked Questions

How much can I realistically save with a FinOps checklist?

Most organizations find 20-30% waste in their first audit. Idle resources, oversized VMs, and unused storage are the biggest contributors. After the first pass, ongoing savings are smaller but still meaningful.

Is it safe to stop dev/test instances overnight?

Yes, as long as they aren’t part of a production pipeline. Use a schedule and always tag them as environment=dev to avoid accidental stops. You’ll lose any in-memory state, so ensure data is persisted.

What’s the difference between reserved instances and savings plans?

Reserved instances lock you into a specific instance type and region. Savings plans are more flexible—they apply to any instance in a family and automatically adjust to size changes. Both offer similar discounts, but savings plans are easier to manage.

Should I use spot instances to save money?

Spot instances are great for fault-tolerant, stateless workloads like batch processing or CI/CD. They can be 60-90% cheaper, but they can be terminated at any time. Never run a database or critical service on spot.

How often should I run this checklist?

Monthly is ideal. Cloud usage changes with new projects, seasonal spikes, and team changes. A monthly review catches new waste before it accumulates.

Can I automate the entire checklist?

You can automate parts—like stopping idle VMs and deleting old snapshots—using cloud-native tools or scripts. But you’ll still want a human to review the cost report and make judgment calls on what to keep.

Related

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.