Your Lighthouse score is 95. Your visitors say the site feels slow. Both can be true at the same time.
This gap trips up Bergen County business owners every week. They run a PageSpeed test, see a green number, and assume performance is solved. Meanwhile actual visitors bounce. The score isn’t lying — it’s just measuring something different from what your customers feel.
Lab vs field: the foundational distinction
Lighthouse is a lab test. It runs in a simulated environment with a throttled network, a controlled CPU, and a single cold page load. Same conditions every time. Reproducible. Clean.
Real User Monitoring (RUM) is the opposite. It measures what actually happened — on a real visitor’s phone, on real Wi-Fi or 4G, scrolling and clicking through real content. No simulation. No averaging across ideal conditions. Just what people experienced.
Google calls field data what it sees across millions of Chrome users. That’s the dataset that affects your rankings — not the lab score on your screen.
Why your 95 score might be lying
Four reasons the lab and the field disagree:
- Third-party scripts load after the snapshot. Chat widgets, analytics, marketing tags, retargeting pixels — many load 3-5 seconds in. Lighthouse captures the page before they arrive. Real visitors don’t.
- Dynamic content isn’t measured. Lazy-loaded sections, infinite scroll, AJAX cart updates — none of that hits the lab test.
- Interaction is barely tested. Lighthouse measures one cold load. Visitors click menus, fill forms, expand accordions. That’s where INP problems hide.
- One device, one connection. Lab tests use a mid-tier Android profile. Your audience runs everything from a 2018 iPhone on spotty LTE to a desktop on fiber.
The metrics that actually matter
Three Core Web Vitals. Measured in both lab and field — but the field numbers are what count.
- LCP (Largest Contentful Paint) — under 2.5 seconds at the 75th percentile of real visitors.
- INP (Interaction to Next Paint) — under 200 milliseconds. Replaced FID in March 2024. Measures how snappy clicks and taps actually feel.
- CLS (Cumulative Layout Shift) — under 0.1. Catches the layout jumps that make visitors mis-tap.
The 75th percentile threshold matters. Google doesn’t care if your average visitor has a good experience. It cares that 75% of them do.
Where to find your field data
- Chrome User Experience Report (CrUX) — the public dataset Google pulls from real Chrome users. Free, anonymized, the source of truth for ranking signals.
- Google Search Console → Core Web Vitals report — the same CrUX data, sliced by URL group and pass/fail status. Check this monthly.
- GA4 web-vitals events — install the web-vitals JavaScript library and pipe LCP, INP, CLS into GA4 as events. Slice by device, browser, page template. This is your own RUM.
The diagnostic when lab and field diverge
When Lighthouse says 95 and CrUX says half your visitors fail LCP, trust the field. Lab tests are useful for catching what changed between deploys. Field data is what tells you whether real people are leaving.
One Bergen County client came to us with a 92 Lighthouse score and a steadily dropping conversion rate. CrUX showed 64% of real visitors failing LCP. The culprit: a chat widget that loaded 4 seconds after the lab snapshot finished. Invisible to PageSpeed. Brutal to actual customers on mobile. We delay-loaded it behind a scroll trigger — field LCP pass rate jumped to 91% within two weeks.
How AJD handles this
Every performance engagement starts with both views. Lab score for the controlled baseline. CrUX and Search Console for what real Bergen County visitors are actually getting. When they disagree, the field wins — every time.
Whether you work with us or not, install the web-vitals library this week. Pipe the events into GA4. In 28 days you’ll have a real picture of what your visitors experience — not a synthetic one.
Lab score green but visitors still bouncing? We’ll pull your CrUX data, find the gap, and fix the field metrics that actually move rankings.
