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.
EN
What it is for
A performance test run from a data centre measures a data centre. Field data measures your actual visitors, on their actual devices and connections, which is the population search engines use when speed affects ranking. This instrument collects that data from your own pages with a small script, and reports it in a form that keeps no one identifiable.
EN
How to use it
Copy the snippet and add it once, before the closing body tag, on the pages you control.
Use the installation check to confirm the page is serving it, then the collector test to confirm the data arrives.
Wait. Field data is a population, not a measurement: a day of traffic tells you very little and four weeks tells you a great deal.
Read the device segments separately. Mobile and desktop are different products as far as this data is concerned.
EN
Reading the result
Figures are daily aggregates over a rolling twenty-eight day window, reported per device segment with the sample count beside them. Read the sample count first: a segment with very few samples is a hint, not a measurement. A segment with zero samples beside a status of "installation not detected" is not an empty meter, it is the answer.
EN
What it cannot tell you
It measures the pages you installed the snippet on, and only visitors whose browser supports the measurement. It is not the same dataset Google publishes, so the numbers will differ. It cannot attribute a change to a cause: it tells you the experience changed, not which release changed it.
EN
What it keeps
Daily aggregates for ninety days, and nothing that identifies a visitor: no page path, no full URL, no user agent string and no IP address is stored. That is a design constraint, not a setting, which is why the guarantee can be printed here.