Key details

  1. Announced October 15, 2026; auto-apply generally available.
  2. New cost-first mode prevents net growth in projected active-series count per cycle.
  3. Existing coverage-first mode still restores full queried detail without cost cap.
  4. Modes can be configured separately by telemetry segment.
  5. Requires Grafana Cloud Adaptive Metrics access and Admin permissions.
  6. Rule limits and metric exemptions can affect the automation outcome.

What builders should take away

  1. Use cost-first for exploratory or agent-generated metric workloads where uncontrolled cardinality is risky.
  2. Keep coverage-first for high-priority monitoring that must restore full query detail.
  3. Set exemptions for metrics essential to alerts and incident response.
  4. Review total billing and telemetry usage separately from active-series count.
  5. Test mode changes on a non-critical segment before automating across production.

What changed

On October 15, 2026, Grafana announced general availability of Adaptive Metrics auto-apply with a new cost-first mode alongside the existing coverage-first behaviour. Adaptive Metrics recommends aggregation rules based on how dashboards, alerts and queries use time-series detail. Cost-first automatically accepts series-reducing recommendations, then accepts series-increasing recommendations only while the projected total remains at or below the series count at the start of the cycle. Coverage-first applies both types without that constraint. Modes can be set separately by telemetry segment.

Why it matters

Automated dashboards, AI agents and exploratory queries can inadvertently cause metric-cardinality growth and surprise bills. Cost-first makes a bounded default available to teams that prioritise predictable active-series counts, while critical services can retain coverage-first behaviour. The tradeoff is explicit: cost-first may withhold detail that a newly issued query requests until an operator changes policy.

Cost-first changes what automation is allowed to do

The new mode accepts all recommended aggregations that reduce active series, ranks series-increasing recommendations from smallest to largest, and applies them only until the original series count would be exceeded. It prevents auto-apply itself from raising the projected series count in that cycle.

Coverage-first remains available

Coverage-first restores raw metric detail when queries demand it, even if that increases the total series count. Grafana recommends it for critical datasets where full coverage matters more than spend predictability. The two modes can coexist in separate segments.

Why AI-driven queries change the risk

A coding or observability agent may issue broad exploratory queries without understanding how many metric dimensions they touch. Automatically restoring every dimension can expand series counts and cost. Cost-first provides a mechanical cap, but does not replace alert and data-retention reviews.

Configuration and limits

Existing Grafana Cloud Adaptive Metrics users can enable auto-apply per segment and choose the mode; an Admin role is required. The documentation warns that segments near their aggregation-rule limits may skip some recommendations. Teams should create exemptions for metrics that must never lose necessary detail.

The guarantee is narrower than a flat invoice

Grafana's guarantee concerns projected active-series counts during an auto-apply cycle, not all usage, storage, retention, query, log or trace charges. Savings figures are vendor-reported; independent operational benchmarks are not provided in the launch.

What to watch next

  • Independent evidence of realised savings and alert fidelity.
  • Whether cost-first gains explicit monetary budgets or forecasting.
  • Handling of query spikes and aggregation-rule limits.
  • Availability across more Grafana plans and self-hosted deployments.

Still unclear

  • The series-count cap does not guarantee a fixed total invoice.
  • Vendor performance and savings claims are not independently audited here.
  • Actual impact depends on metric cardinality, usage and segment policies.

Sources

Direct reading behind this dossier.

2 sources
Choose a mode
Grafana Documentation primary product documentation

Operational configuration, mode guarantees and rule-limit caveats.

Discussion

Discussion is reader-contributed. Comments are not part of the BTN dossier or its editorial evidence.

0 visible comments

Join the discussion

Keep comments useful and relevant. Reader contributions may be moderated and are not BTN editorial evidence.

Sign in to comment