Every startup eng team I've talked to has the same story: cloud costs crept up, someone finally looked at the bill, and 20-30% of it turned out to be resources nobody remembers spinning up.

A few patterns show up over and over:

• The EC2 instance from a hackathon project. Someone spun it up 8 months ago, the project shipped or died, the instance never got torn down.

• Unattached EBS volumes. You terminate an instance but the disk survives, quietly billing you forever with nothing attached to it.

• S3 buckets from a migration that finished a year ago. Data nobody reads, storage nobody audits, cost that never goes away on its own.

The reason this never gets fixed isn't laziness — it's that nobody owns it. Your best engineers are optimizing product, not auditing infra for waste. FinOps tools exist, but most are built for enterprises with a dedicated team to run them, not a 3-person startup eng team.

The fix that actually works at small scale: passive detection, not another dashboard. A tool that just tells you "this specific resource is idle, it's costing you $X/month, click here to kill it" removes the audit step entirely — you make one decision instead of running a discovery process.

If you haven't checked in the last quarter, it's worth 10 minutes to look for the resources that outlived their purpose. It's usually more than people expect.