What we measure
- LCP (Largest Contentful Paint): time to render the largest above-the-fold element. Good ≤ 2.5s.
- INP (Interaction to Next Paint): worst-case responsiveness to a user interaction. Good ≤ 200ms.
- CLS (Cumulative Layout Shift): unexpected layout movement during page load. Good ≤ 0.1.
Lab data and field data, side by side
Lab data (Lighthouse-style synthetic) is fast and reproducible but doesn't reflect real users. Field data (Chrome UX Report) reflects real users but is delayed and only available for sites with enough traffic. We show both, label which is which, and explain when to trust each.
The fix list
For each failing metric we list the offending resources: which image is bloating LCP, which third-party script is blocking INP, which font swap is causing CLS. Findings ship with the exact URL, the byte size, the recommended format, and a one-line engineering remediation.
How CWV relates to ranking
Google has confirmed CWV is a ranking signal, but a small one - tiebreaker territory. The bigger payoff is conversion: every 100ms shaved off LCP correlates with measurable conversion lift in published ecommerce studies. Optimize CWV for users; ranking improvement is a bonus.
Common fixes we surface
- Convert hero images to AVIF or WebP, set explicit dimensions, preload.
- Defer or remove blocking third-party scripts (analytics, chat, A/B testing).
- Add font-display: swap and self-host critical fonts.
- Reserve space for ads and embeds to eliminate CLS.
- Replace long-running JavaScript handlers with debounced or async patterns to fix INP.
<link rel="preload" as="image" href="/hero.avif" fetchpriority="high" />
<img src="/hero.avif" width="1280" height="720" alt="…" decoding="async" />The 5-point CWV checklist
- Stop measuring FID. Confirm your monitoring uses INP.
- Identify the LCP element on every template — image, headline, hero card.
- Audit every third-party script for INP cost; defer or replace the worst offender.
- Reserve space for every async-loaded element: ads, embeds, late-rendered cards.
- Re-test on field data after each ship; lab data lies about real-user conditions.