How to reduce your AWS bill — answered.

Focus on four levers, biggest first: rightsize over-provisioned EC2/RDS (measured on p95/p99, not averages), remove or schedule idle resources, buy Reserved Instances / Savings Plans for steady usage, and add S3 lifecycle rules for cold data. The AWS Billing Optimizer scans your account and lists each of these as an itemized, dollar-quantified fix so you tackle the highest-impact items first.

When a large upload to S3 is interrupted, the parts already sent are left behind. They're billed as storage but never show in the S3 console object list, so they pile up silently — often for years. Add an AbortIncompleteMultipartUpload lifecycle rule (e.g. 7 days) to auto-clean them. This is exactly the kind of invisible cost a scan surfaces.

Savings Plans are more flexible: a Compute Savings Plan applies across instance families, regions, and even Fargate/Lambda. Reserved Instances can be slightly cheaper for one specific type, and still apply to RDS/ElastiCache. For most steady compute, a 1-year no-upfront Compute Savings Plan is the simplest big win. Just as important: check that commitments you already own are fully utilized — unused commitment is pure waste.

Dropping an over-provisioned instance one size typically cuts its cost ~50%. The trick is using p95/p99 CPU so you don't downsize something that spikes at peak. Non-production instances save even more by being stopped nights and weekends (~65% off).

Only with read-only access. The AWS Billing Optimizer connects via a read-only IAM role or read-only keys, never changes your infrastructure, encrypts stored credentials at rest, and isolates each user's data. You can revoke access at any time from your own IAM. See the setup docs.

Usual suspects: idle-but-running resources, data transfer & NAT gateway charges, public IPv4 addresses (now billed hourly), never-expiring CloudWatch logs, orphaned snapshots and upload parts, and commitments you bought but under-use. A scan attributes each increase to a specific service and resource so you know exactly what changed.

Cost Explorer shows what you spent and offers rightsizing + commitment recommendations. It doesn't surface hidden waste like orphaned multipart uploads, never-expiring logs, or per-resource idle detection across every service. Think of the AWS Billing Optimizer as the layer that turns your bill into an itemized, fixable to-do list.

Monthly, plus after any big deploy or migration. Costs drift as workloads change; a monthly scan catches new idle resources, anomalies, and commitment gaps before they compound — and tracks the savings you've actually realized over time.