The cost of knowledge
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
High cardinality in observability metrics is often treated as a problem to eliminate, but dropping labels means losing the ability to diagnose production issues. Every label represents a dimension of knowledge — merchant IDs, pod IDs, content IDs — and removing them can corrupt data (e.g., breaking Prometheus rate() calculations) or hide which customers are affected during incidents. Modern infrastructure with Kubernetes, autoscaling, and ephemeral pods makes high cardinality inevitable through identity churn. The real issue is whether observability platforms force a trade-off between budget and visibility, taxing teams on cardinality rather than ingestion volume.
Table of contents
The two faces of the billWhy high cardinality is inevitable: The churn factorScenario 1: A payments platformScenario 2: A streaming platformThe label you cannot dropCardinality is the shadow of realityScale without the blind spotsSort: