Why We Don’t Use Prometheus or Grafana for Cost Monitoring (And You Shouldn’t Either)

Prometheus and Grafana are great for SLOs, not for cost. Here’s why we keep cost separate.

• Cardinality pain: cost by namespace/service/label explodes series count.
• Retention mismatch: cost data needs weeks/months; Prometheus hot storage is pricey.
• Ownership: price sheets and budgets belong near manifests, not bolted onto dashboards.
• Simplicity: a single agent with tiny storage is easier to run than a full observability stack.
• Security posture: shipping price sheets and cost data to centralized observability can break data residency rules.

We export cost metrics to Grafana when teams want a view, but primary computation lives in a lightweight agent designed for the job.***