Product

Site Speed Is a Revenue Decision: Core Web Vitals in 2026

Abstract cover art for the article: Site Speed Is a Revenue Decision: Core Web Vitals in 2026

Speed is a line item on your revenue forecast. When a page loads slowly or stalls after a tap, fewer people reach checkout, fewer fill in the demo form, and the ones who do arrive a little less convinced. The published evidence is strong enough that performance belongs in the same meeting as pricing and positioning, with an owner and a budget. It also doesn’t force a choice between fast and beautiful. Most slow premium sites are slow because of a handful of specific design and build decisions, and each of those decisions has a cheaper version that looks the same.

What Google measures in 2026

Google’s Core Web Vitals are three field metrics, collected from real Chrome users rather than a lab test. Largest Contentful Paint (LCP) measures how long the biggest piece of content in the viewport takes to render. Interaction to Next Paint (INP) measures how quickly the page visibly responds after someone clicks, taps or presses a key. Cumulative Layout Shift (CLS) measures how much the layout jumps around unexpectedly.

INP is the newest of the three. It replaced First Input Delay on March 12, 2024, and the change matters more than it sounds. FID only looked at the delay before the browser began handling the first interaction. INP observes interactions across the whole visit and, for most sites, reports the slowest one. A page that loads quickly and then freezes when someone opens the menu used to pass. Now it doesn’t.

Metric Good Needs improvement Poor
LCP (loading) 2.5 s or less 2.5 s to 4.0 s Over 4.0 s
INP (responsiveness) 200 ms or less 200 ms to 500 ms Over 500 ms
CLS (visual stability) 0.1 or less 0.1 to 0.25 Over 0.25

Two details decide whether you actually pass. Google assesses each metric at the 75th percentile of page loads, segmented by mobile and desktop, and a page passes only when it meets all three targets. So the bar is not your median visitor on office Wi-Fi. It is the visitor on a mid-range phone on a train.

Most of the web still misses it. The HTTP Archive’s 2025 Web Almanac found that 48% of mobile sites had good Core Web Vitals, up from 36% in 2023. On desktop the figure was 56%. LCP is where mobile falls down most: 62% of mobile pages had a good LCP score, against 74% on desktop.

What the studies actually show

The most quoted study is Milliseconds Make Millions, commissioned by Google and carried out with Deloitte Digital and Fifty-Five. It analyzed mobile data from 37 brands in retail, travel, luxury and lead generation, covering 30 million sessions over four weeks. Using a regression model, the researchers estimated what a 0.1-second improvement across four speed metrics was associated with. For retail, conversions rose 8.4% and average order value rose 9.2%. Travel conversions rose 10.1%. On luxury sites, progression from product page to basket rose 40.1%. Lead generation sites saw a 21.6% lift from the first form step to submission.

Read the study itself before you put those numbers in a board deck. It is correlational, built on Lighthouse lab measurements joined to analytics data. The luxury figure describes add-to-basket, which the authors note luxury shoppers often use as a wish list. And the lead generation section reports that overall conversion rates on mobile and desktop went down when speed improved, even as form progression went up. The direction is clear. The exact multipliers are not something to promise your CFO.

Controlled tests are more useful, and Google has published several. Vodafone ran an A/B test in which the only difference between two versions of a landing page was Web Vitals optimization. A 31% improvement in LCP came with 8% more sales, a 15% better lead-to-visit rate and an 11% better cart-to-visit rate. Rakuten 24 ran a month-long A/B test of an optimized landing page against the original and reported a 53.37% increase in revenue per visitor and a 33.13% increase in conversion rate.

Responsiveness has its own evidence. redBus worked on INP by debouncing scroll handlers, keeping input state local to the component and fetching 10 search results per request instead of 30. INP on that lazy-loading interaction fell from between 870 and 900 milliseconds to between 350 and 370, and the company reported a 7% overall increase in sales. On stability, The Economic Times brought CLS from 0.25 to 0.09 and LCP from 4.5 seconds to 2.5 seconds, and reported bounce rates down 43% overall.

One caution applies to all of these. Case studies get published when the result is good. Treat them as proof that the effect is real and sometimes large, then measure your own.

Where design decisions show up in the numbers

Engineering gets handed the performance bill, but many of the line items were written in the design file. Here is how the usual premium-site choices map onto the metrics.

Hero video and large imagery

For a video element, LCP uses whichever comes first, the poster image load or the first frame. A full-bleed autoplay hero with no poster, started by a JavaScript player, is close to the worst case. Give the video a small, well-compressed poster that works as an image on its own, serve the file from your own origin, and keep the player script off the critical path.

Still images carry the same risk. The Web Almanac found that 16% to 17% of pages lazy-load their LCP image, which Google’s guide to optimizing LCP says will always add delay. Lazy-load what sits below the fold. Never the hero. Put the hero image in the HTML where the browser finds it early, and give it a high fetch priority.

Fonts

A display face is often the most distinctive part of a brand, and it is usually worth keeping. What costs you is how it loads. If your LCP element is a headline set in a web font, preload that font and avoid font-display settings that hide text while it downloads. Google’s font best practices recommend serving WOFF2 only and tuning the fallback font’s metrics with size-adjust so the swap doesn’t move the layout. In practice, one custom family in two weights, subset to the characters you use, covers most brand systems. Four families in six weights is a design choice with a measurable cost.

Animation

Motion is where premium sites earn their feel, and where they often lose INP. Google’s animation guide is blunt about it: restrict animations to transform and opacity, and avoid any property that triggers layout or paint. Animating top, height or margin makes the browser recalculate the page on every frame. Sticking to transforms also keeps motion out of your CLS score, because transform animations don’t register as layout shifts.

The bigger risk is JavaScript. Scroll-linked effects, custom cursors and page transitions that run heavy scripts on the main thread delay the response to every tap. Before you approve a motion concept, ask what happens when someone taps a button halfway through it.

Third-party scripts

According to the 2024 Web Almanac, 92% of pages use at least one third party, and the median page among the top 1,000 sites uses 66. Every chat widget, heatmap tool, testing snippet and embedded player competes with your own code for the main thread. Embeds are the easy win. Lighthouse’s guidance on facades gives the example of a YouTube placeholder that weighs 3 KB, with the 540 KB player loaded only when someone interacts with it. The same guidance lists facades for live chat tools.

How to budget for speed without losing the look

Performance gets cut when it is treated as polish at the end. It survives when it is a constraint from the start, like a grid or a color palette.

Set the budget before design starts. Google’s performance budgets guide describes three kinds: quantity limits such as image weight or number of font files, timing milestones, and rule-based scores. For a marketing site, a workable combination is the Core Web Vitals thresholds at the 75th percentile on mobile, plus a few limits your designers can see while they work, such as hero asset weight and the number of font files.

Spend it where the brand lives. You rarely need every section to animate. Pick the few moments that carry the identity, usually the hero, a signature transition and one interactive product moment, and build those properly. Everything else can be static and quick. Visitors remember a site by its best moments, and they notice lag everywhere.

Give the number an owner after launch. Speed decays. Marketing adds a tag, and six months later someone uploads an uncompressed hero. Put field data from Search Console or your own real-user monitoring into the same monthly review as conversion rate, and require a performance check before any new script goes live.

When this advice matters less

If your site is a thin front door for enterprise deals that close through a sales team over six months, a tenth of a second will not show up in revenue you can measure, and the money may do more elsewhere. The same goes for internal tools where users have no alternative. Even then, watch INP inside the product, because a sluggish interface shapes how people feel about you every working day. If you already pass all three thresholds on mobile, further gains get expensive fast. And search rankings are a weak reason to chase perfect scores. Google says good Core Web Vitals align with what its core ranking systems seek to reward. That is a reason to care, not a reason to trade away content quality for a higher Lighthouse number.

A next step you can take this week

Open the Core Web Vitals report in Google Search Console, or run your three highest-revenue pages through PageSpeed Insights and read the field data, not the lab score. Write down mobile LCP, INP and CLS for each. For any metric outside the good range, find the element responsible: the LCP element, the slowest interaction, the block that shifts. Then trace it back to the decision that created it. Usually you will find an asset or a script that can change without changing how the page looks. Fix that first, wait for a month of new field data, and compare conversion rates across the same periods.

Sources