Before you buy a cost platform or hire anyone (including us), you should know that AWS ships eight cost tools that range from free to nearly free — and that they are not interchangeable. Each one sees a specific slice of your waste and is structurally blind to the rest. Vendors rarely explain this, because "the native tools cover more than you think" is a bad pitch. It happens to be true, so here it is, ranked by what each tool catches in practice.
Key takeaways
- The CUR is the only source of truth; everything else is a lossy view of it. Enable it on day one — it has no memory of the past.
- Anomaly Detection catches spikes, never drift — the slow 8%/month climb that actually ruins budgets sails straight past it.
- Compute Optimizer knows utilization, not intent: it will recommend shrinking the headroom you deliberately keep for peak days.
- No native tool remediates anything. The gap between "flagged" and "fixed" is where cloud waste actually lives.
The ranking, in one table
| Rank | Tool | What it actually catches | What it structurally misses |
|---|---|---|---|
| 1 | Cost & Usage Report | Everything, at line-item and resource-ID grain — the ledger every other tool summarizes | Nothing, but it answers only the questions you engineer against it |
| 2 | Cost Anomaly Detection | Spikes against learned baselines: the forgotten GPU box, the runaway Lambda loop | Slow drift, anything that was already expensive on day one of its learning window |
| 3 | AWS Budgets | Threshold breaches, forecast overruns; the only native tool with actions (it can cut IAM permissions) | Anything under its thresholds; tells you that, not why |
| 4 | Cost Optimization Hub | Aggregates idle, rightsizing and commitment recommendations org-wide, deduplicated and dollar-quantified | Only sees what its feeder tools see; no architecture-level waste |
| 5 | Compute Optimizer | Over-provisioned EC2/ASG/EBS/Lambda against observed utilization, incl. Graviton candidates | Business context; memory (without agent metrics); your month-end peak |
| 6 | Cost Explorer | Trends and what-changed questions a filter can answer; RI/SP purchase recommendations | Resource-level truth; anything needing more than 13 months of history |
| 7 | S3 Storage Lens | Bucket-level waste: incomplete multipart uploads, stale non-current versions, tiering candidates | Everything outside S3; advanced metrics are the paid tier |
| 8 | Trusted Advisor | Classic idle-resource checks: unattached EBS, idle load balancers, low-util instances | Full checks need Business/Enterprise support; shallow next to the tools above |
Why the ledger outranks every dashboard
The Cost & Usage Report wins because every other tool is a pre-baked summary of it. The question that actually finds money is never "what did EC2 cost?" — it's "which untagged resources in the staging account cost more than $500 last month, and whose IAM principal created them?" That's a five-line Athena query against the CUR and unanswerable anywhere else. The catch everyone learns too late: the report only accumulates from the day you enable it. Enabling it costs pennies in S3 storage; not having enabled it costs you the history every future analysis wants. It's the same reason we run this site's own analytics as Athena over raw logs — summaries expire, ledgers don't.
Where the free tools quietly end
Line the eight up and their shared blind spots draw a neat border around what "native" can do. Nothing remediates: every tool produces a finding; a human (or automation you build) produces a fix, and unactioned findings age into wallpaper — the Flexera data showing waste rising in 2026 is largely a monument to flagged-but-unfixed recommendations. Nothing sees architecture: no native tool will tell you your NAT gateway design is taxing every S3 call, that your chatty microservices are billing you cross-AZ, or that the whole batch tier belongs on Spot — the levers that made up most of the savings in our fintech cost case study. Nothing owns the process: tags decay, budgets go stale and anomaly monitors drift out of scope unless someone runs cost as a monthly practice — which is the actual subject of FinOps, not any tool.
The honest hierarchy: native tools for visibility and alarms, engineering for architecture-level fixes, practice for keeping both alive. If someone is selling you a platform, ask which of those three it replaces; the answer is usually "the first, more prettily." Our full lever ordering — commitments, rightsizing, traffic, storage, architecture — lives in AWS cost optimization, ranked by what pays, and the cloud waste calculator will give you a first estimate of what's recoverable without creating an account anywhere.
The enablement order for a new team
Day one: activate cost-allocation tags, set two or three Budgets with alerts wired to a channel humans read, and turn on Anomaly Detection with per-service monitors (coarse monitors average real spikes into silence). Same week: enable the CUR to S3 — you don't need Athena queries yet, you need the history to start existing. First month: read Compute Optimizer and the Cost Optimization Hub with your load calendar open, and act on the top five items only. That 90-minute setup catches the embarrassing failure modes for roughly $0, which is exactly why we make clients do it before any paid engagement starts — consultants who skip the free layer are billing you for alarm installation, not engineering.
Related: AWS cost optimization guide · Cloud billing & FinOps consulting · DevOps & cloud cost statistics 2026 · RI & Savings Plan calculator · Cloud bill audit checklist