Core Web Vitals: Why a Faster Site Is a More Profitable One
Speed is the part of user experience nobody puts on a roadmap until it starts costing money. A page that takes four seconds to show its main content is not just technically slower than one that takes two. It is a page where more people leave before they see the offer, fewer people click, and fewer people buy.
Core Web Vitals are Google's attempt to make that feeling measurable. They are three field metrics that describe loading, interactivity and visual stability. This post explains what they measure, what the published business results actually show, and where the causal claims get weaker than the headlines suggest.
What Core Web Vitals actually measure
Google defines three stable metrics, each covering one part of the experience. The thresholds below are the good targets, as documented in Core Web Vitals on web.dev.
| Metric | What it measures | Good target |
|---|---|---|
| Largest Contentful Paint (LCP) | How long until the largest image or text block is visible | 2.5 seconds or less |
| Interaction to Next Paint (INP) | How quickly the page responds to taps and clicks | 200 milliseconds or less |
| Cumulative Layout Shift (CLS) | How much visible content jumps around while loading | 0.1 or less |
The targets apply at the 75th percentile of page loads, not the average. If a quarter of your visitors still wait more than 2.5 seconds, you have not passed, even if your median looks fine. Google recommends segmenting mobile and desktop because they often fail for different reasons.
What the business results show
The most useful evidence comes from companies that published before-and-after numbers. Google collects many of them in its Core Web Vitals business results collection.
| Company | What changed | What followed |
|---|---|---|
| Vodafone Italy | 31% better LCP | 8% more sales |
| Tokopedia | 55% better LCP and about 4 seconds better TTI | 23% longer average session; 35% higher CTR; 8% higher conversion |
| NDTV | LCP cut from 3.0s to 1.6s | 50% lower bounce rate, alongside other product changes |
| Nykaa | 40% better LCP | 28% more organic traffic from tier-2 and tier-3 cities |
| Agrofy Market | 70% better LCP | 76% less load abandonment |
| redBus | CLS from 1.65 to 0; TTI from about 8s to 4s | 80-100% higher mobile conversion across markets |
| iCook | 15% better CLS | 10% more ad revenue |
Read the table as directional evidence, not a formula. These companies usually changed several things at once, and some results were measured during broader product work. NDTV explicitly notes that other product changes happened alongside its Web Vitals work, so the 50% bounce reduction cannot be attributed to LCP alone. The honest claim is correlation plus a plausible mechanism.
That is why Vodafone's team recommends a server-side A/B test as the way to measure impact. If you want to know whether speed caused the lift, hold everything else constant and test the faster version against the slower one.
Why speed turns into money
The mechanism is simple: slow pages create more opportunities to leave. NDTV's 55% improvement in LCP correlated with a 50% reduction in bounce rate. Tokopedia's performance work correlated with 35% higher click-through and 8% higher conversion. Those are not abstract scores; they are people reaching the next step instead of abandoning the page.
Speed also changes how much content people consume and how much advertising they see. iCook improved CLS by 15% and reported 10% more ad revenue. The layout stopped moving, ad slots behaved predictably, and the experience became easier to use.
Search is the quieter channel. Nykaa's 40% LCP improvement correlated with 28% more organic traffic from tier-2 and tier-3 cities, where slower devices and networks make performance problems more visible. Google has said page experience, including Core Web Vitals, is part of its search ranking systems. It is one signal among many, not a lever that turns page two into page one. Treat it as a tiebreaker that matters most when two pages are otherwise equally relevant.
For a business, the point is never the metric. It is the behaviour the metric predicts: a bounce, a session, a click, a sale.
The outage lesson: performance is also capacity
On July 16, 2018, Amazon's Prime Day opened at 3 p.m. ET and shoppers almost immediately hit glitches. The mobile app and desktop site crashed, users saw error pages, and some reached checkout only to get an error. Amazon's stock briefly lost the 1.5% gain it had made that morning, Axios reported.
Internal documents obtained by CNBC show what happened behind the error pages. Amazon had not secured enough servers for the traffic surge. It switched to a simplified fallback front page, temporarily cut off international traffic, and added servers manually. A breakdown in an internal system called Sable cascaded into Prime, authentication and video playback.
Amazon still called it the biggest shopping event in its history, with more than 100 million products purchased. A company that size can absorb a bad hour. Most businesses cannot. You do not need Prime Day traffic to hit the same wall: a campaign, a product launch or a newsletter can send ten times the usual traffic to a page built for one times.
Core Web Vitals do not measure server capacity directly, but the same neglect shows up as slow LCP, failed requests and interfaces that stop responding. Performance is not only a front-end cleanup. It is capacity planning plus the discipline to keep the page fast when traffic arrives.
What actually moves the numbers
LCP: make the main content arrive sooner
Most LCP problems are one of four things: slow server response, a large hero image, render-blocking CSS or JavaScript, or a late-loading font. The fixes that show up repeatedly in the case studies are server-rendering the critical HTML, preloading the hero image, compressing images and serving modern formats, using a CDN, and deferring non-critical scripts. Vodafone improved LCP by 31% with server-side rendering, image optimization and less render-blocking JavaScript.
CLS: reserve space before content arrives
Layout shift usually comes from content that arrives without reserved space: images without dimensions, ads, embeds, cookie banners and web fonts. Reserve the slot before the content loads. iCook improved CLS by making ad unit sizes less variable and pre-allocating fixed slots. redBus fixed component slots and removed unoptimized tag insertion scripts, taking CLS from 1.65 to 0.
INP: keep the main thread free
INP measures the delay between a tap and the next frame. Long JavaScript tasks are the usual cause, especially third-party tags, hydration and event handlers doing too much work. redBus optimized third-party scripts and split responsibilities into smaller services, cutting TTI and TBT. Tokopedia built a script controller, code-split its bundle and deferred non-critical JavaScript. INP replaced First Input Delay in 2024 because FID only measured the first interaction, not the whole visit.
How to measure without fooling yourself
- Start with field data. CrUX, Search Console's Core Web Vitals report and PageSpeed Insights show real-user data. Lighthouse is a lab tool: useful for diagnosis, but not a substitute for what customers experience.
- Look at the 75th percentile, split by mobile and desktop. Sitewide averages hide the slow cohort that costs you the most.
- Do not use Lighthouse to judge INP. It cannot measure interaction because there is no user input. Use Total Blocking Time as a lab proxy, then confirm the result in the field.
- Install the web-vitals library and send the three metrics to your analytics. Aggregate by page template, device and country, not just sitewide.
- Tie every metric to a business outcome. If checkout conversion does not move when LCP improves, you may have fixed a page nobody was abandoning.
- Beware one-off tests. A fast page on your laptop and a fast page on a three-year-old phone on mobile data are different products.
A 30-day plan that stays honest
- Pick one template that carries traffic and revenue: the homepage, a product page or checkout.
- Record a baseline at the 75th percentile on mobile for all three metrics, plus bounce rate and conversion for the same period.
- Fix the LCP element first. It is usually the largest visible win and the easiest to explain to stakeholders.
- Reserve space for images, ads and embeds with explicit dimensions or aspect ratios.
- Audit third-party scripts. Remove what nobody uses, defer what can wait, and load the rest after interaction.
- Run a server-side A/B test for two to four weeks. If the faster version wins, add a performance budget to CI and monitor it weekly.
The bottom line
Core Web Vitals are useful because they turn a vague complaint into a number a team can act on. They are not a score to maximize for its own sake. The companies in the case studies did not win because they crossed a threshold; they won because they removed delays that were costing them clicks, sessions and sales.
Start with the page that makes money. Measure the delay your customers actually experience. Fix that, then check whether the business moved.
References
All links were checked at the time of writing.
- Core Web Vitals, web.dev
- Largest Contentful Paint, web.dev
- Interaction to Next Paint, web.dev
- Cumulative Layout Shift, web.dev
- Core Web Vitals business results, web.dev
- Tokopedia case study, web.dev
- NDTV case study, web.dev
- Axios: Amazon Prime Day shopper surge crashes its website
- CNBC: Internal documents show how Amazon scrambled to fix Prime Day glitches