How fast pages load for real visitors

The Performance tab shows how fast pages load and respond for real visitors. It uses Core Web Vitals from the visitor’s browser, not a lab test. This page matches what you see at Performance in the sidebar (Traffic), including the Web vitals settings behind the gear.

Trfyx Performance metrics: Core Web Vitals tracking, experience score rating, and per-page breakdowns

Turn measurement on

Nothing is collected until you enable it. The tracker already on your site does the rest. You do not add another snippet.

Open Performance

In the dashboard sidebar, under Traffic, click Performance.

Turn on web vitals

If the page says web vitals are not turned on, click Turn on web vitals. Or open the gear, titled Web vitals, and switch on Measure web vitals. That adds one small request per page load.

Wait for real visits

Visitors’ browsers start measuring within a few minutes. The experience score stays on Collecting data until the selected range has 50 or more measured loads and all five metrics have a value.

You need permission to configure the site to change the toggle. The date range, granularity, and filters in the header work the same way as on other reports.

Skip measurement on one page by adding data-disable-web-vitals="true" to the tracking script on that page. Pages without fetch or PerformanceObserver are skipped automatically.

How a load is measured

After web vitals are on, the existing tracker asks /api/vitals/{websiteId}/config. If enabled is true, it loads /vitals.js from the same host as script.js (your managed proxy hostname if that is Active). The collector records five numbers only. It never sends page content.

When	What happens
A visitor loads a page	The script prepares LCP, INP, CLS, FCP, and TTFB
The tab is hidden or the page unloads	One beacon is sent to `/api/vitals/{websiteId}`
The visitor clicks around a single-page app	That load is not measured again
There was no interaction	INP can be empty. The other metrics can still store

Bots are ignored. Exclusions are ignored. The Trfyx dashboard itself is never measured. The beacon attaches to the same visit as the pageview. If that pageview is still in flight, the collector retries once. Readings older than 400 days are dropped.

The five metrics

The top band shows the five timings at the percentile you pick (default p75, which is how Core Web Vitals are judged). Click a metric to drive the chart and the table’s share and change columns.

Shown as	Full name	Role	Good	Poor
LCP	Largest Contentful Paint	Core Web Vital. When the largest thing in view appears.	2.5s or faster	slower than 4s
INP	Interaction to Next Paint	Core Web Vital. Time from a click, tap, or key to the next paint.	200ms or faster	slower than 500ms
CLS	Cumulative Layout Shift	Core Web Vital. How much visible content jumps while loading.	0.1 or less	more than 0.25
FCP	First Contentful Paint	Supporting. When the first content appears.	1.8s or faster	slower than 3s
TTFB	Time to First Byte	Supporting. Time until the first byte of the response.	800ms or faster	slower than 1.8s

Each cell is rated Good, Needs improvement, or Poor with those limits. A circle, square, or triangle sits next to the words so the rating still reads without colour. The line under the value is the share of loads that were good. Loads measured is how many beacons landed in the range, and what share of pageviews that is.

Experience score

The card under the band folds the five p75 timings into one number from 0 to 100.

90 or more: Great experience
50 to 89: Needs improvement
Below 50: Poor experience

Weights: LCP 30%, INP 30%, FCP 15%, CLS 15%, TTFB 10%. A metric at its Good limit scores 90. At its Poor limit it scores 50. A value of 0 scores 100. Twice the Poor limit scores 0. Until there are 50 measured loads in the range, or any metric is still empty, the ring shows a dash and the title stays Collecting data. Try the same math the dashboard uses:

Chart and comparison

Compare is Previous period or No comparison. Previous period uses the matching range before the one in the header. Timing deltas treat up as worse (red). Loads treat up as better (green). A change under 0.05% is hidden. The percentile tabs (p50, p75, p90, p99) change the band, the table, and the map. The chart can plot any mix of those four. p50, p75, and p90 start on. Dashed lines mark Good and Poor for the selected metric. A time bucket with no loads is a gap, not a zero.

Breakdowns

The table under the chart has tabs: Pages, Countries, Regions, Map, Devices, Browsers, Operating systems.

Each row shows all five metrics at the selected percentile, how many loads were good for the selected metric (the tick on the bar is 75%), optional change versus the previous period, and load count.
Click a row name to filter the whole Performance view. Click again to remove the filter. A page filter applies to the measured path. Other filters apply to the visit.
On Pages, the external-link icon opens https://{your-domain}{path}.
The icons at the end of a row open Sessions or Replay with the same filter and date range.
The table paginates 10 rows at a time. Sort by a metric or by loads.

Map colours countries by the selected metric and percentile: green good, amber needs improvement, red poor. Click a country to filter by it.

If the page is empty

You see	What to check
Web vitals aren’t turned on	Use Turn on web vitals or the gear toggle
Collecting data	Wait for 50 measured loads in this range, with all five metrics present
No performance data	Widen the date range or remove a filter. Confirm the tracker is on the pages you care about
Script still not measuring	Confirm the snippet is installed. Tracking is off on localhost unless the script has `data-allow-localhost="true"`. A Content Security Policy must allow `/vitals.js` and `/api/vitals/` on the script’s host

Next steps

Install tracking if the main script is not on the site yet.
Managed proxy if you serve the script from your own subdomain. Vitals follow that host.
Exclusions if your own visits should not be measured.
Session replay to watch a slow page.