Evidence · document 25

Real-User History

In the workspace: /crux-history/

Availability

Included on Advanced

This document is open to every reader on every plan. Availability tells you what a plan includes; it does not decide what you may read.

What it is for

Your own beacon can only know what happened after you fitted it, and a page speed test measures a data centre. This reads what Chrome already measured on real visits to your origin, week by week, for up to forty weeks - including the months before this platform ever saw the site. For a site taken over from somebody else that is the difference between waiting a quarter to have an opinion and having one today.

How to use it

Enable the Chrome UX Report API on your own Google Cloud project from Google setup, then restrict an API key to it. An unrestricted key is refused by Google outright, so this is not optional hardening.
Select a verified project and a device class. Chrome publishes phone and desktop separately, so those are two measurements rather than one split in half.
Select Read from Chrome. It reads only: nothing about your site is changed, and nothing is sent to Google except the origin.
Read the oldest weeks first if you have taken this site over. What the trend did before you arrived is the part nobody else can tell you.

Reading the result

Every figure is the 75th percentile exactly as Google published it, which is the same statistic the Core Web Vitals assessment itself uses, and the good share beside it is the proportion of visits in Chrome's own good bin. Nothing here is averaged, differenced or scored by this platform. A cell reading "not published" means Chrome published no value for that metric in that week - INP is frequently absent on a page with little interaction - and it is never rendered as zero, because an absent measurement is not a fast one.

What it cannot tell you

Chrome only publishes an origin with enough visits to keep individual users anonymous. Below that threshold the answer is that the origin is not in the dataset, which is a statement about the dataset and not about the speed of your site: a fast site with few visitors is indistinguishable from a slow one here, and that is why the page says so in words instead of showing zeros. It is Chrome traffic only, so it is a subset of your visitors rather than all of them, and it is aggregated weekly, so it can never explain a single bad afternoon. Use your own beacon for that, and read the two separately: a figure blended from both would be a measurement neither method produced.

What it keeps

The weekly series is cached per project and device class for a day and can be re-read at any time, because it is a copy of a public dataset rather than a record of your own. Your API key is stored encrypted and is never displayed again after saving. Nothing is sent to Google except the origin: no page path, no visitor, no account detail.