StartupQuickstart

July 27, 2026 · StartupQuickstart

Core Web Vitals for founders: what actually moves the needle

LCP, INP, and CLS in operator terms: what the current thresholds mean for rankings and conversion, the four causes behind most failures, and how to check your real scores in five minutes.

A person using a laptop and a phone at the same time

Somewhere in your inbox there is probably a report saying your website “fails Core Web Vitals,” attached to a redesign proposal. Before you sign anything, spend twenty minutes understanding what the three numbers measure, because the causes are usually narrower, and cheaper to fix, than the pitch implies.

Core Web Vitals are Google’s three user-experience scores. The ones that count come from real Chrome visitors: PageSpeed Insights reports them over the previous 28-day collection period, at the 75th percentile of visits. Google picked that percentile to reflect the most frustrating experiences on your site, so your score is set by the slower phones on hotel Wi-Fi, not by your MacBook on fiber. A page passes only if it meets all three targets.

What each number means to a visitor

  • LCP (Largest Contentful Paint): how long until the biggest element on the screen, usually your hero image or headline, is actually visible. This is the “is this site even working?” moment. Good: 2.5 seconds or less.
  • INP (Interaction to Next Paint): when a visitor taps a button or opens the menu, how long before the page visibly responds. INP watches every click, tap, and key press in the visit and reports roughly the slowest one. It replaced First Input Delay as a Core Web Vital in 2024, so older audits that grade you on FID are out of date. Fail it and people tap twice, trigger the wrong thing, and blame their phone. Good: 200 milliseconds or less.
  • CLS (Cumulative Layout Shift): how much the page jumps around while it loads. This is the score for “I went to tap Pricing and hit Book a Demo because a banner loaded above it.” Good: 0.1 or less, a unitless score where smaller is calmer.

Why the thresholds hit rankings and revenue

Rankings first. Google states that Core Web Vitals are used by its ranking systems, and in the same breath that it “always seeks to show the most relevant content, even if the page experience is sub-par.” Treat it as a tie-breaker. Great content on a slow page still outranks thin content on a fast one. Where it bites is commercial queries in which ten results say roughly the same thing, and there a position or two is real traffic.

Conversion is the bigger lever, and it doesn’t care about Google at all. A Deloitte study commissioned by Google found that a 0.1-second improvement in mobile site speed lifted retail conversions 8.4% and travel conversions 10.1%. Google’s own collection of case studies includes Vodafone, which improved LCP by 31% and saw 8% more sales. The mechanism is mundane: someone clicked your ad, the hero took six seconds on their phone, and they hit back. You paid for that click either way.

Four causes account for most failures

We’ve audited a lot of failing marketing sites, and nearly every red score traces to the same short list:

  1. An uncompressed hero image. The multi-megabyte PNG exported straight from Figma and shipped at full resolution to phones. Google’s LCP guide lists the fixes: serve the right size, compress it, and use modern formats like AVIF or WebP. In our audits, a properly sized hero is routinely a tiny fraction of the original file with no visible difference.
  2. Third-party scripts. The chat widget, two analytics trackers, a heatmap tool, and a retargeting pixel a marketer added years ago. Each one is JavaScript competing with your own page for the main thread, and long tasks on the main thread are exactly what delay a page’s response to a tap. Nearly every INP failure we see lives here, not in the site’s own code.
  3. Late-loading embeds that shift layout. The testimonial carousel, cookie banner, or embedded video that arrives after the text and shoves everything down. If no space was reserved for it, that jump is your CLS score.
  4. Cheap, slow hosting. Time to first byte is the first slice of LCP, and Google’s guide is blunt that “nothing can happen on the frontend” until it arrives. When a bargain host takes a second and a half to respond, most of your 2.5-second budget is gone before the browser has anything to render.

The typical failing site has all four at once: the Figma-export hero, five trackers nobody dares remove because nobody remembers what they do, an embed with no reserved height, all served from a host chosen in a hurry two years ago.

Our own defaults exist to make that list impossible rather than fixable: an image pipeline that resizes and compresses automatically, marketing pages rendered to static HTML with very little JavaScript, fonts self-hosted so there’s no third-party font request and no swap-jump, and no plugin ecosystem quietly adding scripts behind your back. None of it is heroic. Defaults do the work.

Passing is the goal; 100/100 is a hobby

One caveat the audit reports skip: the 0 to 100 “performance score” in Lighthouse is a weighted average of lab metrics from a simulated load, and PageSpeed Insights notes that lab and field values may differ. The real-visitor pass or fail is what counts. Google itself says that chasing a perfect score just for SEO “may not be the best use of your time.” Once you pass, returns diminish fast: the engineering week that takes LCP from 1.9 to 1.2 seconds would almost always earn more spent rewriting the headline it loads. And if you’re pre-launch with 40 visitors a day, none of them from paid or search, a template site that roughly passes is fine. Leave it alone until traffic is worth money.

How to check yours in five minutes

  1. Open PageSpeed Insights at pagespeed.web.dev. It’s free, from Google, no signup.
  2. Run your homepage and your two most-visited landing pages. Check analytics for what those actually are; it’s rarely what you’d guess.
  3. Read the top section, what real users are experiencing, and ignore the lab score below it for now. That field data is the 28-day record. Green on all three metrics means you pass.
  4. If a page has too little traffic for its own field data, PSI falls back to your whole site’s data, and if the site has too little as well, the section is missing. In that case speed isn’t your bottleneck yet. Treat the lab numbers as directional and move on.

Sites we build and run pass these thresholds by default, and the retainer includes watching the field data monthly, so a new tracker or an oversized image gets caught the week it lands. If you have a failing report and a redesign quote side by side, send us both and we’ll tell you which of the four causes you’re looking at.

Sources

Want systems like this built for you?

We build and run data pipelines, websites, and AI automation for startups.

Book an intro call