Core web vitals for conversion (Not just SEO)

Most teams treat Core Web Vitals as an SEO compliance task. The conversion data tells a different story. Pages that hit good thresholds across LCP, INP and CLS convert noticeably better than pages that just pass the orange band. In our client portfolio, all-green Vitals correlate with roughly 2x the conversion rate of pages still in the orange or red zones.

What Google measures (the quick recap)

Here are the key things to understand. None of them are complicated on their own, but together they change how you approach this.

• LCP (Largest Contentful Paint): how fast the main content appears. Good: under 2.5s.
• INP (Interaction to Next Paint): responsiveness to user input. Good: under 200ms.
• CLS (Cumulative Layout Shift): visual stability. Good: under 0.1.

Our internal budgets

Use the list below as a working reference. Each item is something specific you can check or action directly.

• LCP under 2.0s on 4G mobile (Google's threshold is 2.5s).
• INP under 150ms (Google's threshold is 200ms).
• CLS under 0.05 (Google's threshold is 0.1).

We tighten Google's thresholds because the gap between 'good' and 'excellent' is where the conversion lift lives. A 2.4s LCP passes Google but still feels slow on mobile. A 1.4s LCP feels instant.

What moves the needle most

Here is the process broken into clear steps. Take them one at a time and check each off before moving on.

1. Compress and lazy load images. Use AVIF or WebP, never raw PNG hero shots.
2. Reserve space for hero, fonts and embeds. CLS kills mobile conversion.
3. Defer third party scripts. Tag managers and chat widgets are usually the worst INP offenders.
4. Inline critical CSS for above the fold content.
5. Self-host fonts with font-display: swap and a maximum of two families.
6. Use fetchpriority="high" on the LCP image.

The conversion data behind the budgets

The details matter here. The points below separate what actually moves the needle from what is just noise.

• A 1 second delay in load reduces conversions by up to 7 percent (Akamai/Aberdeen).
• Bounce probability rises 32 percent when load goes from 1s to 3s (Google/SOASTA).
• 53 percent of mobile users abandon pages that take over 3s to load.
• On a Framer build we shipped, LCP dropped from 3.1s to 1.4s and conversion lifted 22 percent in the next month.

INP: the metric most teams underestimate

INP replaced FID in March 2024 and is the hardest of the three to fix. The usual culprits are tag managers, chat widgets, A/B test scripts, and large React hydration costs. Audit your third-party script bundle quarterly. If a script is not actively earning its INP cost, remove it.

CLS: the cheapest fix in the stack

CLS is almost always solved by reserving space. Set explicit width and height on every image. Reserve a min-height on font swap regions. Avoid injecting cookie banners, ads or hero text after first paint. These are 30 minute fixes that move both Vitals and conversion.

AEO bonus: Vitals affect AI citation too

Slow pages get crawled less often and are less likely to be selected as citation sources by AI engines that prioritise responsive, well structured pages. The same investment that lifts conversion also lifts your odds of being quoted inside ChatGPT, Perplexity and Google AI Overviews.

A 2 week Vitals sprint

Follow the sequence below from start to finish. The order matters because later steps assume the earlier ones are done.

1. Days 1-3: audit field data in Search Console + CrUX. Pick the worst-scoring page template.
2. Days 4-7: ship image optimisation, font self-hosting, and explicit width/height on images.
3. Days 8-11: audit and remove or defer third-party scripts.
4. Days 12-14: validate field data, set monitoring (Vercel Analytics or RUM), and document budgets.

Sources and further reading

Before going further, get clear on the following. Each point below covers a different part of the picture.

• web.dev official documentation on LCP, INP and CLS.
• Akamai/Aberdeen research on load time and conversion.
• Google/SOASTA mobile speed benchmarks.
• FrameCraft client conversion data (n=30 engagements).