Tune a Noisy Alert with Noz

SigNoz released a new AI assistant, Noz, that helps users tune noisy alert rules by analyzing firing history and suggesting optimal thresholds. The tool separates actionable fires from transient noise and recommends specific threshold adjustments to reduce alert fatigue while preserving real incident detection.

An alert keeps paging and most of the pages are not actionable. Before you mute it, you ask Noz how often it has fired and what a saner threshold would look like. Prerequisites - A SigNoz Cloud https://signoz.io/teams/ account with Noz https://signoz.io/docs/ai/noz/ available. - An alert rule that has fired enough times to have a history. Step 1: Hand Noz the alert Open Noz from the top-right header, use Add Context → Alerts to attach the noisy alert, then ask: This alert is too noisy. Look at its firing history and tell me how to tune it. Noz reads the alert's condition, threshold, and evaluation window, then pulls how often it has fired over the recent period. Step 2: Separate real fires from noise How many of these fires lined up with actual service degradation? Noz compares each fire against the target's error rate and latency, so you can see what share of pages matched a real problem versus a transient blip. Step 3: Get a recommendation Suggest a threshold and evaluation window that would have skipped the noise but still caught the real incidents. Noz proposes specific values and explains the trade-off, with Suggested Actions to refine the rule. Review the recommendation against your own tolerance before you apply it. Tips Judge against impact, not fire count. Have Noz line fires up with real degradation so you tune on signal, not volume. Confirm before changing. Treat the suggested threshold as a starting point and sanity-check it against a known incident. Under the Hood Under the Hood under-the-hood To answer, Noz works through several agentic steps, visible under Worked through N steps : | Step | What It Did | |---|---| | Read alert rule | Loaded the alert's condition, threshold, and evaluation window | | Ran builder query | Pulled the alert's firing history over the recent period | | Ran builder query | Compared each fire against the target's error rate and latency | | Reasoned | Separated actionable fires from noise and recommended a threshold | Next Steps Connect Infrastructure Issues to App Errors with Noz https://signoz.io/docs/ai/use-cases/noz-correlate-infra-errors/ - Check whether the fires trace back to the platform. Alert Fatigue Audit https://signoz.io/docs/ai/use-cases/alert-fatigue-audit/ - Audit noisy and stale alerts across the workspace from your coding agent. If you need help with the steps in this topic, please reach out to us on SigNoz Community Slack https://signoz.io/slack/ . If you are a SigNoz Cloud user, please use in product chat support located at the bottom right corner of your SigNoz instance or contact us at cloud-support@signoz.io mailto:cloud-support@signoz.io .