Best Practices

Performance Monitoring

Best practices for getting the most out of xcelera's performance auditing.

Choosing an audit frequency

The right frequency depends on how often your site changes and how quickly you need to catch regressions:

  • Daily — good default for most production pages. Catches regressions within a day.
  • Weekly — suitable for low-traffic or infrequently updated pages.
  • Hourly — best for high-traffic pages on active projects where you deploy frequently. Uses more credits.

Start with daily audits and increase frequency for pages that matter most. You can always adjust later.

Understanding scores and metrics

Lighthouse produces scores (0–100) across four categories, but the individual metrics are often more actionable:

  • LCP (Largest Contentful Paint) — how long until the main content is visible. Target: under 2.5 seconds.
  • TBT (Total Blocking Time) — how long the main thread is blocked. Target: under 200ms.
  • CLS (Cumulative Layout Shift) — how much the page layout shifts during loading. Target: under 0.1.

These three metrics are the core web vitals that most directly impact user experience.

Tracking regressions

xcelera automatically detects when a page's performance score drops by 5 or more points. To make the most of regression detection:

  1. Set up notifications — configure notification levels to alert you on score regressions
  2. Review the audit detail — click into a regressed audit to see which specific metrics changed
  3. Check the git context — if the audit was triggered from CI, the commit context helps identify what caused the regression

Reducing score variance

Lighthouse scores have natural variance between runs. xcelera reduces this by running Lighthouse multiple times per audit and selecting a representative result. To further reduce noise:

  • Audit the same URL consistently (avoid redirects or A/B tests that change content between runs)
  • Ensure the page is reliably reachable — intermittent network issues inflate variance
  • Look at trends across multiple audits rather than reacting to single data points

Setting up alerts

Combine notifications with webhooks to build an alerting pipeline:

  • Use Warning level notifications to catch failures and score regressions
  • Route webhooks to Slack for team visibility
  • Use Critical level for PagerDuty or on-call systems
Help