On this page
Where performance data comes from
Where: Monitoring › Performance.
The numbers come from real page loads in visitors' browsers, not lab benchmarks. When a visitor leaves the first screen, the script sends the timings once. It's on by default; data-vitals="false" on the script switches it off.
Five metrics and their thresholds
| Metric | Good (at most) | Poor (above) | What it means |
|---|---|---|---|
| Largest Contentful Paint (LCP) | 2.5 s | 4 s | How soon the main content appears |
| Interaction to Next Paint (INP) | 200 ms | 500 ms | How soon the page responds to a tap or click |
| Cumulative Layout Shift (CLS) | 0.1 | 0.25 | How much the page jumps while loading |
| First Contentful Paint (FCP) | 1.8 s | 3 s | How soon the first pixels appear |
| Time to First Byte (TTFB) | 800 ms | 1.8 s | How soon the server starts responding |
Anything in between is "needs improvement".
P75 or the median?
Cards switch between P95, P75, median and average.
- Median — half of visitors were faster, half slower.
- P75 — three quarters were faster. It's closer to what your slower visitors experience, and it's the usual yardstick for passing.
Read P75 day to day. The average is easily pulled up by a handful of very slow loads.
Finding what's slow
- Look at each metric's P75 and its share of "good".
- Group by page to find the worst paths. There's also a "slowest pages" list sorted by P75 LCP.
- Group by device, country and browser to see whether it's concentrated in one kind of visitor. Slowness in one region only is often a CDN issue.
- The load breakdown splits a load into server response, network transfer and front-end rendering, which tells you whether to talk to backend or front-end engineers.
Percentiles on low-traffic pages jump around; don't draw conclusions from a day or two.
Switching on error monitoring
Where: Monitoring › Errors.
Error monitoring is off by default. Add data-errors="true" and the script loads a small error collector:
<script defer src="https://static.pathclue.com/client.js" data-site="YOUR_SITE_KEY" data-errors="true"></script>You then see error count, error rate (errors ÷ pageviews), affected visitors and sessions. Errors are grouped by message; mark a group as resolved or ignored, or expand it to see each occurrence's time, page, region and device and open that session directly.
Frustrated clicks
With autocapture on, the errors page also has a frustrated clicks table:
- Rage clicks — the same element clicked three or more times within 500 ms, usually a button that didn't respond.
- Error clicks — a script error within one second of the click.
A page needs at least 20 pageviews to appear in this table. It works best next to a heatmap, where you can see exactly where people keep clicking.
Related
- Monitors and alerts — get notified when the error rate jumps.
- Web Vitals feature overview