Lab data vs field data: why PageSpeed Insights sometimes disagrees with reality
A PageSpeed Insights report shows two different kinds of data: a live, one-time “lab” test performed against your page right now, and “field” data — real visitors' actual experience, averaged over the past 28 days. They can disagree, sometimes sharply, and knowing why keeps you from chasing the wrong number.
What is lab data?
Lab data is a single simulated page load, run under fixed, controlled conditions (a set network speed and device profile) the moment you click “Analyze.” It's available instantly for any URL and shows exactly which resources and scripts slowed that load down, which makes it the right tool for debugging what's slow right now. It isn't perfectly repeatable, though: Lighthouse's own documentation warns that scores shift between runs even when nothing on the page has changed, so compare several runs before trusting a small difference. Its weakness is the flip side of that control: it doesn't reflect the range of real devices, networks, and locations your actual visitors use.
What is field data?
Field data is aggregated from real Chrome visitors who have usage statistics reporting and history sync turned on, collected via Google's Chrome UX Report (CrUX) over a rolling 28-day window. It reflects genuine variation across devices, connection speeds, and geographies — but it only exists for URLs with enough Chrome traffic to produce a valid sample, so brand-new or low-traffic pages often show no field data at all. It also lags: PageSpeed Insights refreshes it daily, but each day's number still covers the previous 28 days, so a fix you ship today shows up gradually and is only fully reflected about four weeks later.
Why do lab and field scores sometimes disagree?
Lab tests one fixed condition. On mobile, PageSpeed Insights emulates a mid-range phone (a Moto G Power) on a throttled slow-4G connection, and it loads the page with an empty cache. Field data is whatever mix your real visitors actually have. If most of them are on fast connections and newer devices, or are returning visitors with your files already cached, field numbers can look much better than the lab. If a meaningful share are on congested networks or older phones, a site can look fine in a lab test and still show a poor field LCP. Individual metrics can diverge too: heavy JavaScript that hurts a lab-measured metric like TBT might barely affect real-world INP if most actual interactions happen well after the page has finished loading.
Which one does Google actually use for ranking?
Google describes Core Web Vitals as metrics that measure real-world user experience, and says good Core Web Vitals align with what its core ranking systems seek to reward. The Core Web Vitals report in Search Console — the tool Google points site owners to — is built from CrUX field data, not from a lab test. So for Search, the field numbers are the ones to watch, and a lab score is best treated as a debugging aid. For a URL that doesn't yet have enough real-world traffic for a valid field dataset, lab data is the only data you'll have.
Which one should you actually pay attention to?
If field data exists for your URL, treat it as the more reliable signal for real, Search-relevant performance. Use lab data to diagnose and fix a specific issue in the moment, then check the fix's real-world effect in field data once enough time — and traffic — has passed for the 28-day window to catch up.
How does PageWeight show both?
Every PageWeight scan runs a live lab test using the PageSpeed Insights API. When Google has enough real-user data for the URL, the report also shows a “Real users” rating (fast, average, or slow) and INP, shown as “Response speed.” INP only ever comes from field data, because a lab test has no real clicks or taps to measure. Every other metric in the report is a lab measurement from that scan. Tracked pages keep a history of those lab scores, which is where run-to-run noise evens out into a trend.