Telemetry Spend Ratio
Observability cost as a proportion of the infrastructure it observes, used as a tripwire for a category that grows silently.
Observability commonly reaches 10–30% of infrastructure spend, and occasionally exceeds the cost of the systems it monitors. It grows silently because every team adds telemetry and nobody removes any.
Tracking the ratio makes the growth visible before it becomes an emergency, and gives a defensible threshold for review.
The drivers, in typical order: log volume, metric cardinality, trace volume, and retention duration.
Reductions that preserve diagnostic capability: sample successful-request logs aggressively while retaining errors in full; tail-based sampling for traces, keeping all errors and slow traces with a small success baseline; cardinality allow-lists enforced at the collector with an alert on series growth; tiered retention, since almost all queries look at the last few days; and deleting unused dashboards, alerts and metrics — most backends report per-metric query usage and the result is usually startling.
The trap to avoid: cutting telemetry indiscriminately trades a visible cost for an invisible one, and the invisible one is longer incidents. Cut volume, not coverage — sampling and retention are the levers, not switching off instrumentation.
Governance that prevents recurrence: per-team cost visibility, without which nobody manages it.