Core Web Vitals and Conversion: What the Data Actually Shows

The SEO argument for site speed is real but modest. The conversion argument is much larger, and it is the one worth taking to your finance team.

What we see in the data

Across marketing sites and storefronts we have rebuilt, mobile conversion improves roughly 6% to 9% for every second removed from Largest Contentful Paint, up to about the 2 second mark. Past that the returns flatten sharply. This is correlation from rebuilds where other things changed too, so treat it as a direction rather than a coefficient.

Where the time actually goes

Teams reach for image optimisation first because it is the easiest thing to do. It is rarely the biggest problem. In the audits we run, the ranked culprits are consistent:

• Third-party scripts: chat widgets, heatmaps, A/B testing tools, review platforms and six analytics tags nobody remembers adding.
• Render-blocking CSS and fonts loaded without a swap strategy.
• Client-side rendering for content that could have been static HTML.
• Unoptimised hero images, which matter, but usually less than the three items above.

Cumulative Layout Shift is the quiet killer

LCP gets the attention, but layout shift does more direct damage to conversion. A form that jumps as a late-loading banner injects itself above it produces mis-taps, abandoned fields and a user who no longer trusts the page. Reserve space for everything that loads late: images with explicit dimensions, ad slots, cookie banners, embedded video.

A practical order of operations

• Get field data from Chrome UX Report or your RUM tool. Lab scores mislead on real device performance.
• Audit and remove third-party scripts. Defer everything that survives until after first interaction.
• Serve content as static HTML wherever it can be static. Static or incrementally rendered pages beat client-side fetching every time.
• Set explicit dimensions on every image and reserve space for late-loading elements.
• Self-host fonts with font-display: swap and preload the one face used above the fold.
• Only then optimise images and enable modern formats.

A 90 in the lab and a 40 in the field means you optimised for the test, not the customer.