Short definition

Field data — performance measurements collected during real website visits, rather than in a controlled lab test.

Also known as: real-user performance data, real user monitoring data, RUM.

You can measure a website in two ways: in a lab or in the field. Both produce numbers, but they answer different questions. Field data shows how a site performs during actual visits, on the devices and networks people use.

What field data is

Field data is performance data from real website visits. One public source is the Chrome UX Report (CrUX), which collects data from an eligible subset of Chrome users, not all visitors. The CrUX API aggregates measurements over 28-day windows and reports distributions and percentiles, not individual browsing records. A metric’s p75 is the value at or below which 75% of the measured page loads fall. It does not describe 75% of unique visitors.

Lab data comes from tests with controlled device and network settings, such as Lighthouse in PageSpeed Insights. Those tests help with diagnostics and comparisons, but they cannot represent every visit. In our September 2026 study of 19 web development agencies, lab LCP was up to 16.7x the CrUX LCP (p75) among the seven sites with current field data. The measurements show a gap; they do not establish its cause. Cache state, devices, networks, and user behavior can all contribute.

Where to find field data

PageSpeed Insights shows available CrUX data above the lab results. Check whether it represents the tested URL or the origin: an origin combines measurements across its pages. Search Console’s Core Web Vitals report groups similar pages on your own site. For historical data, the CrUX History API returns 25 weekly collection periods by default, each covering an overlapping 28-day window. Earlier periods may have data even when the latest period does not; history does not have a lower eligibility threshold.

When to use lab and field data

Use field data to prioritize problems observed during real visits. Check all the Core Web Vitals, not just LCP, and remember that a good p75 does not mean every page load was good. Use lab tests to investigate causes and check changes before release. Your own real user monitoring can also provide diagnostic detail. If public field data is missing, keep testing in the lab rather than treating the gap as evidence that the site is fast or slow.