Core Web Vitals and conversions: what measured cases show

Core Web Vitals are Google's three measures of what real visitors experience on a site: loading, responsiveness and visual stability. Each has a "good" and a "poor" threshold, and a page passes only when it reaches "good" on all three, at the 75th percentile of page loads, with mobile and desktop measured separately³.

Largest Contentful Paint (LCP) measures when the largest visible element finishes rendering. Interaction to Next Paint (INP) measures how long the page takes to respond to clicks, taps and key presses. Cumulative Layout Shift (CLS) measures how much content jumps around on screen while someone is reading.

The 75th percentile is what matters: a site is rated "good" on a metric only when at least 75% of visits meet the threshold⁴. A healthy average is not enough if a quarter of your visitors, usually on mobile and on a slow network, get a poor experience.

Fewer than half. According to HTTP Archive's 2025 Web Almanac, only 48% of sites have good Core Web Vitals on mobile, versus 56% on desktop⁵. The figure is global, drawn from the Chrome UX Report (CrUX), the dataset of real Chrome user experience. On mobile, LCP is the metric that fails most often, well ahead of INP and CLS.

In numbers: 62% of mobile pages reach a good LCP, versus 74% on desktop, and 13% land in the poor range on mobile⁵. For INP, 77% of mobile sites are good⁵. For CLS, 81%⁵. If you are stuck on LCP, you are in the majority; if you pass all three, you are ahead of more than half the web.

The most useful case studies are the ones that isolate speed. Vodafone's is the cleanest: two versions of the same landing page, visually and functionally identical, with traffic split down the middle. The optimized version had a 31% better LCP and produced 8% more sales, a 15% lift in lead-to-visit rate and an 11% lift in cart-to-visit rate¹.

The other cases are before-and-after comparisons with no control group. Treat them as strong evidence, not proof of causation:

QuintoAndar is the closest match to a large company operating in an emerging market. The housing platform had the worst INP among its competitors and, after the work, pages with a poor experience fell from 32% to 6.9%². Part of the gain came from removing third-party pixels and breaking up long JavaScript tasks.

Deloitte's "Milliseconds Make Millions" study, run with Google, measured the effect of a natural one-tenth-of-a-second improvement in mobile site speed. Over 4 weeks, it analyzed retail, travel, luxury and lead generation brands across Europe and the United States¹⁰. The data is global, with no country breakdown.

1. 01
   Retail: conversions up 8.4% and average order value up 9.2%¹⁰.
2. 02
   Travel: conversions up 10.1% and average order value up 1.9%¹⁰.
3. 03
   Luxury: page views per session up 8.6%¹⁰.
4. 04
   Lead generation: bounce rate improved by 8.3%¹⁰.

The study's limit is built into its design: it is observed correlation, not a controlled test. That makes it useful for sizing the bet, while Vodafone's A/B test is what proves it. If your site sells little on mobile, these percentages carry less weight for you.

We recommend fixing whichever metric fails, in the order the web fails most: LCP, then INP, then CLS. The logic is simple. A page passes only when it reaches "good" on all three, and on mobile LCP is the most common bottleneck, with 62% of pages good versus 77% for INP and 81% for CLS⁵. Your own field data still outranks the global average.

1. 01
   Measure with field data, not just lab tests. Use the Core Web Vitals report in Search Console or PageSpeed Insights, split mobile from desktop, and write down each metric's value at the 75th percentile.
2. 02
   Fix LCP first if it is above 2.5 seconds³. In the cases above, the gains came from server-rendering critical HTML, resizing the hero image and cutting render-blocking JavaScript, as Vodafone did.
3. 03
   Fix INP if it is above 200 milliseconds³. Break up long JavaScript tasks, defer anything that does not respond to the click and audit third-party pixels and tags, as QuintoAndar did.
4. 04
   Fix CLS if it is above 0.1³. Set width and height on images: on mobile, 62% of pages fail to set dimensions on at least one image⁵.
5. 05
   Prove the gain with an A/B test when traffic allows. Vodafone sent 50% of traffic to each version¹, and that design is what separates the effect of speed from everything else.

Open PageSpeed Insights, paste the URL of the page that drives the most revenue or leads, and read the field data section, not the lab score. Write down LCP, INP and CLS on mobile and compare each one with the thresholds in the table. The first metric outside "good" is your first task.

Then cross-check with your analytics tool: record that page's current conversion rate, by device. That is your baseline. Without it, any speed gain becomes a matter of opinion, and the internal case for funding the next fix has nothing to stand on.