Field data is performance data collected from real visitors on their own devices and networks. Lab data is collected in a controlled environment with a predefined device and network setup. The two often disagree because they measure different things.
How do the two types differ?
web.dev describes lab data as the output of tools such as Lighthouse, run under identical conditions every time. Field data, also called real user monitoring (RUM), reflects the actual mix of devices, connections and locations and is reported as a distribution. The score is the 75th percentile.
PageSpeed Insights shows both. According to Google, the field part comes from the Chrome User Experience Report (CrUX), the lab part from Lighthouse, and the field part covers the previous 28-day collection period. Field data offers fewer metrics, while lab data helps with debugging.
| Metric | Measures | Core Web Vital? | Field or lab | “Good” per web.dev |
|---|---|---|---|---|
| LCP (Largest Contentful Paint) | Render time of the largest image, text block or video visible in the viewport, from the start of navigation | yes (loading) | field and lab | 2.5 s or less |
| INP (Interaction to Next Paint) | Latency of interactions across the whole visit until the next paint; one value that all (or nearly all) interactions fall below | yes (interactivity), since 2024-03-12 | field; in the lab only with simulated interactions, Lighthouse does not measure INP | 200 ms or less |
| CLS (Cumulative Layout Shift) | Largest burst of unexpected layout shifts during the entire lifecycle of the page | yes (visual stability) | field and lab | 0.1 or less |
| TTFB (Time to First Byte) | Time from the start of navigation until the first byte of the response arrives | no (helps diagnose LCP) | field and lab | 0.8 s or less |
| FCP (First Contentful Paint) | Time from the start of navigation until any part of the page's content is rendered | no (helps diagnose LCP) | field and lab | 1.8 s or less |
| TBT (Total Blocking Time) | Total time after FCP during which the main thread was blocked long enough to prevent input responsiveness | no (lab proxy for INP) | lab; web.dev does not recommend field measurement | less than 200 ms on average mobile hardware |
| Speed Index | How quickly content is visually displayed during page load (from a video of the load) | no | lab (Lighthouse) | Threshold according to Lighthouse: mobile up to 3.4 s “fast” (web.dev gives no value) |
| FID (First Input Delay), historical | Time from the first interaction until the browser can begin processing event handlers | until 2024-03-12, replaced by INP | field only | 100 ms or less |
According to web.dev, the 75th percentile of page loads is assessed, mobile and desktop separately. A page passes the Core Web Vitals when LCP, INP and CLS are “good” at the 75th percentile. Chrome tools ended FID support on September 9, 2024. Sources: web.dev: Web Vitals; web.dev: Largest Contentful Paint (LCP); web.dev: Interaction to Next Paint (INP); web.dev: Cumulative Layout Shift (CLS); web.dev: Time to First Byte (TTFB); web.dev: First Contentful Paint (FCP); web.dev: Total Blocking Time (TBT); web.dev: First Input Delay (FID); web.dev Blog: Interaction to Next Paint is officially a Core Web Vital (2024-03-12); Chrome for Developers: Lighthouse – Speed Index. Last checked: .
Why do the numbers differ?
web.dev lists several reasons:
- Lab tests cannot predict when and how users interact, so Interaction to Next Paint (INP) can look fine in the lab and poor in the field.
- Field data includes cache hits, back/forward cache restores, and different viewport sizes, which can change which element counts as the Largest Contentful Paint (LCP).
- Field Cumulative Layout Shift (CLS) covers shifts over the whole lifetime of a page, not only during loading.
- Personalization and A/B tests change what visitors see.
Which one should you use?
web.dev recommends basing the decision on whether to optimize on field data, because it shows what users experience. Lab data is useful for finding and reproducing the cause, for example on a slow device.
Example
A page scores green in Lighthouse, yet PageSpeed Insights flags INP in its field section. Visitors interact with menus and filters in ways the lab run never performs.
Sources
- Why lab and field data can be different (and what to do about it) – web.devweb.dev
- About PageSpeed Insights – Google for Developersdevelopers.google.com
- Chrome UX Report (CrUX) – Chrome for Developersdeveloper.chrome.com
Last updated: