Insights:Core Web Vitals Tools & Commands: Ready to Use
Tool Roles
| Tool | What to Check | When to Use |
|---|---|---|
| Search Console | Real-user CWV, queries, and page groupings | Finding issue patterns |
| PageSpeed Insights | Field data + lab summary | Quick single-page review |
| CrUX / BigQuery | Real-user distribution | Trends and groupings |
| Lighthouse CLI | Lab metrics, diagnostic items | Local regression |
| WebPageTest | Multiple locations, devices, and video recording | Reproduce slow loading |
| Chrome DevTools | Main thread, network, layout shifts | Identify root causes |
| curl | TTFB, status codes, cache headers | Check server and CDN |
1. Search Console: Find templates first
Open the CWV report and analyze it grouped by mobile/desktop and by page. Don't start with the homepage, as it is often optimized separately.
Focus on:
- Which template has the most URLs marked as "Needs Improvement".
- Whether the issue is LCP, INP, or CLS.
- Whether changes appear within the 28-day window after deployment.
2. PageSpeed Insights: Single-page snapshot
Enter a URL and check for field data first; if no field data is available, look at the lab section.
Note: PSI's lab scores are from simulated environments and don't represent your real users. Don't treat a single score as acceptance.
3. Lighthouse CLI: Local & Repeatable
After installing Node, you can run it directly:
npx lighthouse https://example.com/page \
--only-categories=performance \
--form-factor=mobile \
--throttling-method=simulate \
--output=json \
--output-path=./lh-mobile.json
Key points:
- Fix
form-factorandthrottling-method, or results won't be comparable over time. - Run the same page after each change and save the JSON for comparison.
- Check LCP elements, render-blocking resources, and unused CSS/JS—don't just look at the overall score.
4. WebPageTest: Multi-Location Reproduction
Ideal for investigating "why the site is slow in a specific region." You can select location, device, and connection type to generate video recordings and waterfall charts.
Focus on:
- Where TTFB increases.
- When the first-screen images start downloading.
- Whether third-party requests are blocking.
5. DevTools: Diagnose, Don't Just Score
Capture one page load in the Performance panel:
- Look for long tasks on the main thread.
- See when the LCP element renders.
- Identify which frame triggers layout shifts.
- Check network request priorities.
The Layout panel reveals CLS sources. Don't rely on eye-balling and guessing "that area seemed to jump."
6. curl: Check TTFB and Cache Headers
curl -o /dev/null -s -w \
"dns=%{time_namelookup} connect=%{time_connect} tls=%{time_appconnect} ttfb=%{time_starttransfer} total=%{time_total}\n" \
https://example.com/page
Check three things:
- Whether TTFB is significantly above baseline.
- Whether CDN cache headers are present.
- Whether incorrect cache status is returned.
A Reusable Checklist
- Use Search Console to identify problem patterns.
- Confirm field data with PSI or CrUX.
- Reproduce locally with Lighthouse CLI.
- Locate specific elements and long tasks in DevTools.
- Change one variable at a time and rerun the same commands.
- After going live, wait for field data to be collected.
What Not to Do
- Don't compare scores taken under different networks or cache states.
- Don't run tests on the homepage only.
- Don't strip useful content just to achieve a perfect Lighthouse score.
- Don't treat lab scores as real-user results.
Tools are magnifying glasses, not answers. Define the question first, then pick the right tool.
Related reading: