Trusted Advisor — Cost Optimization Checks

TL;DR: Most AWS accounts pay for things nobody uses — a load balancer for a decommissioned service, an EBS volume unattached since a one-off experiment, an Elastic IP for a project that never shipped, an instance at 3% CPU since 2024. Trusted Advisor is the automated scanner for exactly this, each finding tagged with estimated monthly savings. The catch: the full Cost Optimization suite needs Business+ Support ($100/mo or 10% of spend, whichever is greater); on Basic/Developer, Compute Optimizer and Cost Optimization Hub cover ~60–70% of it free.

The numbers

• The support-tier wall: Basic/Developer get a limited subset (mostly Service Limits + core Security); Business+ unlocks all cost checks, the Support API, weekly digests, and (Enterprise) the Organizational View.
• Core cost checks: low-utilization EC2 (<10% CPU + <5 MB net on 4+ of 14 days), idle load balancers (<100 req/wk), underutilized/unattached EBS, unassociated EIPs (~$3.60/mo each), idle RDS (no connections 7+ days), Redshift, Lambda timeouts/errors, RI/SP utilization, RI expiry.
• Break-even: an unaudited account over ~$1,000/mo typically recovers 3–10% of monthly spend in a first-pass cleanup — often paying for a year of Business Support in the first session.
• Field examples: ~200 leaked unassociated EIPs = $720/mo ($8,600/yr) flagged in one shot; ~30 orphaned load balancers post-migration = $580/mo (and surfaced two still quietly in use).

Do this

1. Confirm your Support tier, then open Trusted Advisor → Cost Optimization and sort by estimated savings descending.
2. Work the top 5–10 findings first — unassociated EIPs and idle load balancers are the classic no-architecture-change wins ("still needed? no? delete").
3. Verify before deleting — an "idle" RDS might be a quarterly reporting or compliance DB; the finding is a signal to investigate, not to delete.
4. Refresh each check after cleanup (per-check manual refresh, or the 24h auto-refresh) and watch the savings number drop; iterate, since deletions expose new findings (a killed instance leaves an unattached volume).
5. Enable the weekly email digest for ongoing rhythm, and use the Organizational View (Enterprise) for a cross-account team scorecard.

Gotchas

• "No cost checks" = Basic/Developer tier — upgrade, or use Compute Optimizer + Cost Optimization Hub free.
• Estimates use on-demand list price — RI/SP-covered resources save less than shown.
• No native "ignore this resource" on most checks — track deliberate exceptions externally (a ta-ignore=true tag or a spreadsheet) and skip them.
• Findings are recent snapshots (1–4 weeks) — a resource busy last month but idle now still gets flagged; context is on you.
• It's a checklist, not a strategy — no architectural or cross-service recommendations (it won't say "use Intelligent-Tiering" or "co-locate these services").

Skip this if

• You spend under ~$1,000/mo and can't justify Business Support — the free Compute Optimizer + Cost Optimization Hub pairing covers most of the ground.
• You want per-check depth — see the individual check guides: Idle Load Balancers, Low Utilization EC2, Unassociated Elastic IPs, Underutilized EBS Volumes, and Unutilized NAT Gateways. Verify before/after with AWS Cost Explorer.