Core Web Vitals are real-user experience metrics. They do not measure whether a site is beautiful; they measure whether it loads, responds and stays stable. Google defines three: LCP, INP and CLS. The “good” thresholds are up to 2.5 seconds, 200 milliseconds and 0.1 respectively.
Lab vs. field
Lab tools such as Lighthouse simulate conditions and help diagnose. Field data comes from real users, aggregated in CrUX (Chrome User Experience Report) and Search Console. A site can look good in the lab and fail in the field, so priorities should come from real data where it exists.
LCP: what appears first
Largest Contentful Paint measures when the largest visible element finishes rendering. It is usually a hero image or a text block. Common causes:
- The main image is too heavy or has no defined dimensions.
- Loading is delayed by blocking CSS or JavaScript.
- A slow server or no cache on the first response.
- Fonts hiding text for too long.
Priority: optimise the hero image, preload it where it helps, reduce blockers and serve HTML quickly. This usually delivers the biggest gains with the least risk.
INP: what responds to input
Interaction to Next Paint measures the latency between an interaction and the next visible frame. Often the problem is not “the site loads slowly”, it is “the site blocks when I click”. Heavy JavaScript, long handlers and unnecessary work on the main thread are the usual suspects.
Priority: split long tasks, defer scripts that are not needed, remove dependencies and measure real interactions.
CLS: what jumps
Cumulative Layout Shift measures visual instability. Images without dimensions, injected ads, late banners and fonts that change size are common causes. The fix is often cheap: reserve space and set explicit dimensions.
How to prioritise without optimising blindly
- Measure the current state on mobile, using field data when available.
- Identify the problem element for each metric, not just the value.
- Fix what affects the most pages and the most traffic first.
- Validate in the lab and in production, comparing before and after.
- Repeat in a short cycle. Performance is maintenance, not a one-off project.
For a starting point, the free audit checks observable technical signals, including response time and basic elements. For a full plan, see SEO and performance.