How do I find and shut down unused AWS resources?
A safe, step-by-step method for small D2C brands to find idle AWS resources across regions, confirm nobody needs them, and remove them safely.
Anadoris Digital · · 4 min read

Almost every AWS account that has been running for a while collects leftovers: a server from a feature that was never launched, a database copied for testing, snapshots from a migration, a load balancer for an app that moved. Each looks small on its own, but together they can make up a surprising share of the bill. They also widen your security exposure, because forgotten resources are rarely patched.
Cleaning up is straightforward if you approach it carefully. The main risk is deleting something that turns out to be needed, so the method below puts confirmation and reversibility before deletion. Plan to spend a few focused hours on the first pass, then much less each month after that.
Step one: see everything, in every region
Resources hide in regions you do not normally look at, especially if someone once followed a tutorial that used a default US region. Start with a full inventory. Cost Explorer grouped by region shows where money is being spent. AWS Resource Explorer can search resources across regions once it is turned on. The Billing console's bills page lists charges by service and region.
- Cost Explorer grouped by region and service.
- AWS Resource Explorer with cross-region search turned on.
- Trusted Advisor checks for idle and underused resources (coverage depends on your support plan).
- Compute Optimizer recommendations for servers, volumes and functions.
- The EC2 console in each region, including volumes, snapshots, IP addresses and load balancers.
Step two: spot what looks unused
- Servers with very low CPU and network use over several weeks.
- Databases with few or no connections.
- Storage volumes not attached to any server.
- Snapshots and machine images older than your retention policy.
- Elastic IP addresses not attached to a running resource.
- Load balancers with no healthy targets or no traffic.
- NAT gateways in networks with nothing running.
- S3 buckets that have not been read or written to in a long time.
- Old Lambda function versions and container images piling up in registries.
Look at metrics over a meaningful period. Something idle for a week might be a monthly report job or a sale-only service. Check at least a full month, and ideally across a recent sale.
Step three: confirm before you remove
For each candidate, find its owner. Tags help if they exist; if not, check CloudTrail for who created it, look at its name and settings, and ask the team and any agencies that have worked on the account. Make a simple list with the resource, what you think it is, who you asked and what they said. Anything nobody claims after a reasonable period moves to the next step.
Step four: make it reversible, then remove
Stop before you delete. Stop servers rather than terminating them, and wait a couple of weeks to see if anyone notices. Take a final snapshot of volumes and databases, and a final export of buckets that may hold records. Then delete, starting with the lowest-risk items. Keep the final backups for a defined period, then remove them too, or they become the next round of clutter.
- Take a final backup of anything holding data.
- Stop or disable first, delete later.
- Delete in small batches and watch alerts after each batch.
- Record what was removed, when and by whom.
- Set an expiry date for the final backups you kept.
Step five: stop the clutter returning
Prevention is cheaper than clean-up. Require tags for owner, project and environment on new resources, and use AWS Organizations policies to block regions you do not use. Define infrastructure in code so everything has a known home. Schedule test environments to switch off automatically, and give experiments an expiry date. A short monthly review of new resources keeps the account tidy without a big annual effort.
Cleaning up without breaking sale-day capacity
Be careful with resources that exist only for peaks. A larger database instance, extra servers in an auto scaling group or a reserved capacity setting might look oversized in a quiet month but be essential during a festive sale. Before removing or shrinking anything, check its metrics during your last big event, and ask whoever planned that event.
A better pattern is to make peak capacity temporary by design: scale up a few days before a known sale and scale back afterwards, using scheduled actions or a short checklist. That way, the account stays lean most of the year without risking the days that matter most. Note these planned changes in your runbook so a clean-up never removes them by accident.
Where to start
Open Cost Explorer, group by region, and look for spending in any region you do not intentionally use; that is often the quickest win. Then check for unattached volumes, unused IP addresses and old snapshots in your main region. Confirm owners, take final backups, stop first and delete later. Finish by adding owner tags to everything that remains.
Related service: AWS & Cloud Management · More on this topic: AWS & Cloud Management articles
