Skip to content
Refine

Glossary · Cost Visibility

What is a Cost Anomaly?

A statistically significant unexpected spike or drop in AWS cost for a given service / account / region, relative to its recent baseline.

Definition

Cost Anomaly

A cost anomaly is a usage or spend pattern that deviates significantly from the recent historical baseline for that dimension — a specific service, account, region, or resource. Unlike a threshold alert ("spend exceeded $5,000"), an anomaly is detected relative to the resource's own pattern: a sudden 3× spike on a previously flat dimension is an anomaly even if it didn't cross a dollar threshold.

AWS Cost Anomaly Detection (a native AWS feature) and most third-party FinOps platforms — including Refine — use statistical baselines plus seasonality awareness to surface these.

Why it matters for AWS cost

The expensive bugs are the ones nobody is looking for: a runaway Lambda misconfigured to fan out per-event, a forgotten test deployment, a NAT Gateway burning data-transfer cost because a VPC endpoint was missed. Threshold alerts miss these because the absolute dollar amount is small until it isn't.

Refine's anomaly detection pairs statistical detection with AI-generated commentary: not just "your Lambda cost spiked," but "Lambda processOrder ran 4× more invocations than baseline, likely due to a SQS redrive loop."

Find waste like this in your own AWS account — free

A read-only IAM role, connected in minutes. No signup needed for the bill analyzer.

Refine is built and supported by HabileLabs, an AWS Advanced Tier Services Partner.

Back to glossary15 terms in the AWS cost & FinOps glossary