Glossary · Monitoring

Golden Signals

The four golden signals, latency, traffic, errors, and saturation, are the minimum set of metrics Google's SRE book recommends monitoring for any service.

Updated October 21, 2026

The four golden signals are latency, traffic, errors, and saturation. The idea comes from Google's Site Reliability Engineering book, which suggests that if you can measure only four things about a user-facing system, these give the broadest view of its health.

Why it matters

Teams often drown in hundreds of graphs and still miss the obvious. Golden signals give a consistent starting point for every service and a quick triage order during an incident. They also map neatly onto alerts that matter to users.

Example

For a web API, the four signals could be:

  • Latency: 95th and 99th percentile response time, tracked separately for successful and failed requests, since fast errors can hide slow successes.
  • Traffic: requests per second.
  • Errors: share of responses with 5xx status, plus wrong answers returned with a 200.
  • Saturation: CPU, memory, queue depth, connection pool use.

A page might fire when the 5xx ratio exceeds 2 percent for five minutes, or when p99 latency doubles against a weekly baseline.

Common mistakes

  • Averaging latency, which hides the slowest users. Use percentiles.
  • Counting only HTTP 5xx and missing errors that return 200 with an error body.
  • Treating the four signals as a full observability strategy. They tell you that something is wrong, not why.
  • Confusing them with the RED method (rate, errors, duration) and USE method (utilization, saturation, errors), which are related but have different scopes.

Get notified when this term updates

Plus occasional practical tips on uptime and monitoring, nothing else. Unsubscribe anytime.

Keep reading

Glossary · MonitoringAlert (Monitoring)A monitoring alert is an automatic notification sent when a check or metric crosses a defined condition, such as a site failing to respond.Glossary · MonitoringAlert DeduplicationAlert deduplication merges repeated alerts about the same underlying problem into one incident, so responders are notified once rather than hundreds of times.Glossary · MonitoringAvailabilityAvailability is the share of time a service is working and reachable when users need it, usually expressed as a percentage over a defined period.Guide · MonitoringAlert Fatigue and On-Call Basics for Small TeamsSet severity levels, escalation paths, alert thresholds and runbooks so a small team gets paged for real problems only and still sleeps through the night.8 min readGuide · MonitoringCron Job Heartbeat Monitoring to Catch Silent Failures EarlyLearn how heartbeat monitoring catches silent cron failures, with copy-ready examples for crontab, systemd, GitHub Actions, Docker, WP-Cron, Laravel and more.9 min readGuide · MonitoringA Blameless Incident Postmortem Template With a Worked ExampleWrite a blameless incident postmortem that drives real fixes: timeline, root cause vs contributing factors, 5 whys limits, a copy-paste template and an example.10 min read

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.