Your waste is unique. Your detection should be too.

AWS, Azure, GCP, OCI and Kubernetes out of the box, plus Snowflake, Databricks, MongoDB Atlas, Datadog, OpenAI, Anthropic and 40+ other SaaS and AI vendors. New connectors ship every few weeks — if you bill for it, we can showback it.

Most detection engines run on one hardcoded definition of waste across every account, so a resource kept alive on purpose — a DR replica, a database sized for a Q4 spike, a snapshot under legal hold — gets flagged over and over with no way to tell the tool it already knows the reason. Engineers stop trusting the alert and start ignoring the queue entirely. Fixing this means letting each team scope and define waste on its own terms, not tuning sensitivity on one global rule.

Start by defining what waste means for each resource type — idle, rightsizing, extended support, or a fully custom rule — rather than accepting a vendor's default thresholds. Then scope those rules with tags, for example production only or excluding DR, so the same detector engine applies your definitions consistently across cloud, storage, managed services, and AI and GPU spend. Finout's detection agent runs this as one configurable policy engine instead of a separate tool per cloud.

Idle detection looks for resources doing effectively nothing — an unattached volume, a stopped but still billing instance, a bucket nobody reads from. Rightsizing looks at resources that are running and doing real work but provisioned larger than the workload needs, which is a judgment call, not just a metric, since a team may size up on purpose ahead of a known spike. Finout treats both as separate detector types so idle waste can be acted on automatically while rightsizing calls route to the resource owner.

The number that matters is what survives after you back out the waste a team has a real reason to keep, not the raw total a detector first flags. Rank issues by savings that hold up against your own definitions and ownership rules, then track how much of that ranked queue actually gets resolved versus dismissed. That gap — flagged versus resolved — is a better health metric than total dollars flagged.