What a finding looks like
What every CloudIQ finding contains, the families of checks that produce findings, and how a saving is estimated.
A finding is the unit of work in CloudIQ. Everything else exists to produce findings, rank them, and track them to done.
What a finding contains#
| Field | What it holds |
|---|---|
| Resource | Account or project, region, and the resource identifier |
| Condition | A named condition, such as "idle instance" or "unattached volume" |
| Evidence | The metrics and configuration examined, and over what window |
| Current cost | What the resource costs today |
| Estimated saving | The monthly saving and the calculation behind it |
| Risk and confidence | A risk level for the change and a confidence rating for the estimate |
| Fix | The specific change, how it ships (pull request, runbook, SQL rewrite or ticket), and the rollback steps |
| Owner | The team or person responsible, from your tags and labels |
Written out, a finding reads like this:
Resource vol-0a1b2c3d4e5f (us-east-1, account 4471...9082)
Condition Unattached EBS volume
Evidence Detached since 2026-05-02 (103 days). No snapshot since
2026-04-28. 500 GiB, gp3.
Cost $40.00 / month
Saving $40.00 / month (direct: the volume stops existing)
Risk Low. Confidence: high.
Fix Snapshot, then delete. Runbook with rollback from the snapshot.
Owner platform-team (tag: owner)
The evidence line is what turns "the tool says delete this" into a claim the volume's owner can check in under a minute. Checking it is exactly what you want them to do.
Families of checks#
Each check looks for one condition in one service. A rule that tries to detect "waste" in general produces findings nobody trusts. A rule that detects "a load balancer with no healthy targets for 14 days" produces one that is either true or false.
| Family | Examples | Why it matters |
|---|---|---|
| Idle and orphaned | Unattached volumes, load balancers with no targets, idle databases, NAT gateways with no traffic, endpoints with no invocations | Cheapest to act on. There is nothing to preserve |
| Oversized | Instance and database rightsizing, volume type and IOPS, cache node size, cluster scaling floors | Needs a utilization window long enough to include the peak |
| Structural | Storage in the wrong class, data crossing a boundary it need not cross, backups kept warm, previous-generation families | Does not look like waste on a utilization chart |
| Rate and commitment | Uncovered steady usage, expiring commitments, the wrong instrument, support plan tier, Marketplace renewals | Correct in volume, priced worse than it could be |
| Data platform | Query patterns that scan more than they return, unpartitioned tables, idle slot capacity, pricing model fit | Query-level evidence, not just the invoice |
| Governance | Untagged spend, budget overruns, daily anomalies with a probable cause, quota headroom | Exposure and allocation, reported as such, never as a saving |
How a saving is estimated#
Every estimate carries the method that produced it, because the methods differ in confidence.
- Direct: the resource stops existing, so the saving is its run rate. An unattached volume is the clearest case.
- Differential: the resource changes shape, so the saving is the gap between two prices. For example, gp2 against gp3 at the same size, after your commitment coverage.
- Modeled: the saving depends on behavior that has to be projected: commitment coverage, scheduling a workload off out of hours, a reservation baseline. These are shown as ranges.
Where a check is advisory, or produces no dollar figure by design, the page says so rather than presenting it as a saving. Quota headroom, for example, is reported as exposure, and reserved GPU capacity as committed spend, never as a saving.
Ranking#
Findings are ranked by estimated saving across every service, so the list starts with the largest opportunity. Risk and effort are shown beside each one, so you can choose. A $200 change that ships today can beat a $400 change nobody will approve.
Many teams start with idle and orphaned resources because they are low risk, then work down the ranked list.
Next#
- How CloudIQ works
- Cloud Cost Optimization: the coverage, service by service
- From finding to fix
Try this against your own spend
The free Cloud Cost Assessment switches on your billing exports with you and takes a read-only snapshot. You get more than 30 insights and your top savings ranked by dollars a month.