Observability

A performance dashboard is only as good as the data behind it. This one measures two different things and refuses to mix them: field telemetry from real visitors, and lab measurements taken from a production build.

Field: what real visitors experience

Core Web Vitals at the 75th percentile, from visitors at the analytics consent tier. Scored against Google's published bands, so the verdict matches what Search Console reports.

CLS Good

0.001 13 samples · provisional

FCP Poor

3187 ms 14 samples · provisional

INP Good

44 ms 11 samples · provisional

LCP Needs work

3412 ms 13 samples · provisional

TTFB Poor

2959 ms 14 samples · provisional

…and how much of it is real

Every number above reads the production lane only. This is the split it was filtered from — shown rather than hidden, because an aggregate that quietly drops rows looks identical to one with nothing to drop.

Production 65 Unclassified 17

Production traffic is 79.3% of collected samples. If that share moves sharply, the filter is the first thing to check — not the site.

Lab: what the last build weighs

Bundle size never shows up in real-user data until it has already shipped. These come from a production build, committed to the repository so growth is visible in review.

route_js_kb Poor

483.8 KB warn 150 · fail 250

baseline_js_kb Good

79.6 KB warn 80 · fail 120

doc_html_kb Good

2.8 KB warn 50 · fail 100

Measured 2026-09-18 · NODE_ENV=production