Use Web Vitals alongside traffic analytics
Performance reports are easier to investigate when you know which pages and visitors supplied the samples. Combine real-user Web Vitals percentiles with traffic trends, route filters, and sample counts.
Enable collection intentionally
The browser SDK accepts an optional web-vitals module. Without it, performance code is not loaded. Enable it in a browser initialization path and verify a hard navigation on a non-sensitive test page.
The current SDK records hard page-load measurements. Client-side route transitions do not automatically create independent Web Vitals samples for every virtual pageview. Keep that boundary in mind when grouping by page.
Install web-vitals with npm install web-vitals after installing the SDK from its local package
import { createAnalytics } from '@untitled-analytics/web';
import * as webVitals from 'web-vitals';
const analytics = createAnalytics({
writeKey: 'YOUR_SDK_WRITE_KEY',
endpoint: 'https://api.your-domain.com',
webVitals,
});Read percentiles with the sample count
LCP describes loading of the largest relevant visible content; INP describes interaction responsiveness; CLS describes layout instability. FCP and TTFB provide additional timing context. These are different measurements with different units.
The reports expose p50, p75, p90, and p99. Missing samples are excluded rather than inserted as zero. A single sample can make several percentiles equal; that is not evidence of a stable population. INP can be missing if no qualifying interaction was observed.
Investigate one change at a time
Choose a date range, route, and device segment. Compare the performance trend with pageviews so you can see whether the sample population changed. A campaign bringing slower devices can shift a percentile even without a code release.
Use a lab performance run to reproduce a suspected issue, then field data to observe it in visitors’ browsers. A lab score and a real-user percentile answer different questions.
Build a recurring performance dashboard
Add Web Vitals and top-page cards beside traffic trends. Make the selected period consistent across charts and note sample volume in reviews. Track a release date as context rather than declaring every movement caused by the release.
Plans are based on customer ARR with monthly visitor allowances. Performance measurements still need bounded ingestion and data quality reviews.
Frequently asked questions
Why is INP blank?
No valid interaction sample may have been collected in the selected population. A missing measurement is not zero.
Are SPA routes measured independently?
Not by the current SDK’s hard-navigation collection. Do not interpret every virtual pageview as a new performance sample.
Should I use lab or field data?
Use lab runs for controlled diagnosis and field measurements for observed visitor experience.