Core Web Vitals: fix the lab PageSpeed score or the field data in Search Console first?

Lots of audits kick off with a PageSpeed score in the 40s, and then the team burns a sprint chasing it. Lab scores come from one simulated, throttled device, so they can look awful while real visitors have a perfectly fine experience. The field data in Search Console's Core Web Vitals report, built from 28 days of Chrome user data, is much closer to what Google uses as a signal. So start with the URL groups flagged poor there, not the homepage score. The catch is that low-traffic sites often have no field data at all, which leaves you with lab numbers only. For those sites, which lab metric do you actually move first?

Before you accept that there's no real-user data, check the field section in PageSpeed Insights. When a single URL doesn't get enough visits, PSI often falls back to origin-level CrUX data for the whole domain. It's coarse, but it still tells you more than a lab run does.

If you really are lab-only, start with LCP. It's the most stable metric in a simulated test, and on most stores it comes down to one element, usually the hero banner or the main product image. Open the LCP element in the diagnostics and look for three things: it's lazy-loaded (it shouldn't be), it's a CSS background instead of an img tag, or it's waiting on a slider script before it can load. Fixing those usually moves the number more than anything else on the page.

Agreed on LCP, and the lazy-loading check is the one people miss most. Theme hero sliders often lazy-load the first slide too, so the LCP image gets delayed even though it's right there above the fold. Take loading="lazy" off that image and you'll usually see the difference before you touch anything else.

On sliders the problem is often worse than lazy loading. Many themes keep the first slide's real URL in data-src or data-lazy and let JavaScript swap it in after load, so the preload scanner never sees it early. Check the raw page source, not the DOM in devtools, and look for an actual src on that img tag. If it's missing, the fix belongs in the theme. If it's there, add fetchpriority="high" and remove loading="lazy", so Chrome ranks it above the other images competing for bandwidth.

Jun's fix only helps if the LCP element is actually the image, so check what PSI reports first. On a lot of service sites, like a clinic or plumber landing page, the hero is a big headline over a background, and the LCP is just the h1 text. Text like that usually waits on the web font, not the image, so fetchpriority does nothing for it. In that case the fix is font loading. Look at the LCP element label in the diagnostics before you touch the slider.