I build websites for law firms, so in September I ran 14 well-known US law firm homepages through PageSpeed Insights: national personal injury brands, Big Law, and a few regional firms. I expected the big-budget sites to win. They didn't, and the lab score turned out to be a poor predictor of who passed.
Full write-up with every site and number: Core Web Vitals for law firm websites (14 sites tested). This post is the developer version: what surprised me and what I'd fix first.
The headline: lab score and field result disagree
Google ranks on field data (real Chrome users, 28-day window, from CrUX), not on the lab score you see at the top of PageSpeed Insights.
- One site scored 91 in the lab and failed Core Web Vitals in the field. The culprit was layout shift (CLS 0.38), which a single lab run barely registered.
- Another scored around 51 in the lab and passed comfortably.
- Two Big Law sites had lab LCP of 30-45 seconds on the simulated slow phone, yet field LCP of about 1.5 seconds and a pass. Their real visitors are on fast hardware and fast networks. The lab simulates a mid-range phone on slow 4G; the field reports what actually happened.
So if you're auditing a site, look at the "Discover what your real users are experiencing" section first. The big circle is a diagnostic, not a verdict.
5 of 13 failed, and no two failed for the same reason
13 of the 14 sites had field data. Five failed:
| Failing metric | What it looked like |
|---|---|
| INP 295 ms | A heavy homepage where taps on the menu felt sticky |
| CLS 0.38 | Content jumping as late elements loaded (the 91-score site) |
| LCP 2.6 s | Just over the 2.5 s line on a big hero |
| LCP 2.7 s + CLS 0.88 | The worst of the set; lab score 38 |
| LCP 3.1 s | Slow hero image on mobile |
That spread is the useful lesson: "make it faster" isn't a plan. Each failure needs a different fix.
The one problem almost everyone shared
10 of 14 sites (including my own agency site, which scored 67 in the lab) had render-blocking resources as the top opportunity. That's CSS and synchronous scripts in <head> that the browser must download and parse before it paints anything.
On WordPress builds it's usually the same suspects: page-builder CSS for every widget on every page, font loaders, analytics and chat widgets loaded synchronously, and slider libraries.
The fix order I use now
Ranked by how much each usually moves the field result:
-
Render-blocking resources (LCP). Inline critical CSS, defer the rest,
defernon-essential scripts, and stop loading builder assets on pages that don't use them. -
Layout shift sources (CLS). Set
width/height(oraspect-ratio) on every image and embed, reserve space for cookie banners and chat bubbles, and usefont-display: optionalor size-matched fallbacks for web fonts. -
The hero image (LCP). Serve WebP/AVIF at the size actually displayed, add
fetchpriority="high", and never lazy-load the LCP image (a surprisingly common WordPress default). - Third-party JavaScript (INP). Chat widgets, call tracking and review carousels are the usual INP killers. Load them on interaction or after idle.
- Server response time (everything). Page caching plus a CDN. It rarely decides pass or fail alone, but it raises the floor for the other four.
One caveat from my own site: aggressive "delay JavaScript" settings in caching plugins can break things in ways a speed test won't show. On mine, it silently killed page-builder widgets for logged-out visitors. Always test as a logged-out user after changing optimization settings.
Checking a site in five minutes
- Run PageSpeed Insights on the mobile tab.
- Read the field section first. Pass or fail is decided there.
- If field data fails, note which metric, then scroll to the diagnostics that map to it (render-blocking and LCP element for LCP, layout shift culprits for CLS, long tasks for INP).
- Fix one metric at a time, then wait. Field data is a rolling 28-day window, so a fix takes up to four weeks to fully show.
If you want the full per-site breakdown, including which firms passed and why, it's in the original article. I also put the fix list into a free website scan tool for law firms.
Curious whether others see the same lab-vs-field gap on their client sites. My guess is it's widest on B2B and Big Law sites whose audiences are mostly on desktop and fast connections.
Top comments (0)