Alerts and Applied Intelligence in New Relic: Signal Over Noise

Great alerting is about signal, not volume

An alerting system that pages you for everything trains you to ignore it. New Relic’s alerting plus Applied Intelligence (AIOps) helps you get notified only when something genuinely needs a human.

The Building Blocks

  • Condition — the rule that defines a problem (often NRQL-based)
  • Policy — a group of related conditions
  • Workflow — where and how notifications are delivered
  • Destination — Slack, PagerDuty, email, webhook

NRQL-Based Conditions

The most flexible alerts are written in NRQL:

SELECT percentage(count(*), WHERE error IS true)
FROM Transaction WHERE appName = 'my-api'
-- alert when this exceeds 5% for 5 minutes

Applied Intelligence reduces noise by correlating related incidents

Thresholds That Reflect Reality

Setting Guidance
Threshold Base on real baselines, not round numbers
Duration Require the breach to persist (avoid spikes)
Signal loss Alert when data stops arriving, too

Applied Intelligence

AIOps features cut noise automatically:

  1. Anomaly detection flags deviations from learned baselines without hard thresholds.
  2. Correlation groups related incidents into a single issue so one outage isn’t 40 pages.
  3. Enrichment attaches context (recent deploys, related entities) to speed triage.

Every alert should be actionable. If the answer to a page is "ignore it," delete the condition.

What to Learn Next

  • Mute rules for maintenance windows
  • Incident workflows routing by team and severity
  • SLOs and error budgets as a higher-level alerting strategy

Arivanandhan Chitheshwaran