Core Web Vitals for Lead Generation: Which Metric Actually Moves Form Fills

TL;DR

• Deloitte's 2020 study of 30M sessions found a 0.1s mobile speed gain lifted lead-gen form submissions by 21.6%.

• INP (Interaction to Next Paint) is the metric your form embeds, chat widgets, and tracking scripts are quietly failing.

• A good INP score is 200ms or less at the 75th percentile of all interactions, not just the first one.

• Triage in this order on a marketing site: fix the LCP image, then the INP-killer third-party script, then CLS from embeds and fonts.

• Measure with CrUX field data on a 28-day window. Lab scores from Lighthouse will lie to you about real user pain.

The short answer

If you run a lead-gen site and you're trying to figure out which Core Web Vitals metric actually moves form fills, the answer is INP first, LCP second, CLS third — in that order, for most marketing pages we audit. The reason is simple: lead-gen pages don't sell from the scroll, they sell from the click. The moment someone taps your form field, your chat bubble, or your CTA, that's the interaction Google now measures, and that's the moment your third-party stack stutters. Core web vitals for lead generation is a different problem than it is for ecommerce, and the public advice rarely makes that distinction.

The one piece of lead-gen-specific evidence worth anchoring to: Deloitte's 2020 Milliseconds Make Millions study, run across roughly 30 million user sessions, found that a 0.1-second mobile speed improvement lifted lead-generation form submissions by 21.6% and cut mobile bounce by 8.3% (web.dev case study). One tenth of a second. That's the number to remember.

1. Why ecommerce speed stats mislead lead-gen site owners

Most speed studies you've read measure revenue per visitor on a product page. The user lands, scrolls, taps add-to-cart, checks out. Five interactions, all of them on first-party code the store owns. Speed wins are linear: shave 200ms, conversions go up by some predictable slice.

Lead-gen pages don't work like that. A typical B2B marketing page has one job: get a form fill or a meeting booked. The interactions that matter aren't scroll and tap-to-buy. They're focus-on-form-field, type, submit. And almost every one of those interactions runs through third-party code you don't own: a HubSpot or Marketo form embed, an Intercom or Drift chat widget, a Calendly iframe, a Segment tag manager firing six analytics scripts behind it.

Which is why the typical "we got our Lighthouse score to 95" win doesn't show up in pipeline. Lighthouse runs in a lab, on a cold load, with no user interaction. Your real bottleneck is what happens 1.8 seconds after the page paints, when the visitor taps the email field and your chat widget's main-thread JS blocks input for 340ms.

2. LCP, INP, CLS translated into form-fill behavior

A quick operator's translation, because the official definitions are written for engineers and the marketing posts are written for nobody in particular.

LCP (Largest Contentful Paint) is how fast your hero — usually the headline image or video — renders. On a lead-gen page, this is the "did this brand load fast enough that I trust it" moment. If LCP is over 2.5 seconds, you're losing the visitor before they read the value prop.

INP (Interaction to Next Paint) is how fast the page responds when the user actually does something. On a lead-gen page, this is the form-field tap, the dropdown open, the chat bubble click. INP replaced FID in March 2024, and the change matters: FID only measured the first interaction. INP measures the slowest one. So your form, which loads three scripts after the user taps it, finally gets graded honestly.

CLS (Cumulative Layout Shift) is how much the page jumps around as it loads. On a lead-gen page, the killer is the embedded form that pushes everything down 400ms after paint, so the visitor taps the wrong button. Or the late-loading webfont that re-flows the headline.

LCP is about whether they stay. INP is about whether they convert. CLS is about whether they rage-quit.

3. INP: the metric your form embeds and chat widgets are failing

Here's the thing about INP that the public discourse hasn't caught up to. A good INP score is 200ms or less at the 75th percentile, measured across all interactions on the page, not just the first one (web.dev INP guide). For a marketing site with a HubSpot form, an Intercom widget, a cookie banner, and a heatmap tracker, hitting that number is hard. We've audited marketing sites where the LCP is a clean 1.9 seconds and the INP is 540ms because the chat widget's event listener blocks the main thread every time the user tabs between form fields.

The pattern is consistent. The visitor lands, the page paints fast, they like what they read, they tap the email field — and then the form takes a third of a second to respond because three vendor scripts woke up at once. That delay doesn't show up in your Google Analytics. It shows up in your form-completion rate, two layers downstream, and you blame the copy.

INP optimization on a marketing site is mostly a vendor-script audit. Which widget is firing? When? Is it deferred? Is it lazy-loaded only when the user scrolls near the form? If you've never asked those questions, the answer is almost certainly no.

4. The 0.1-second evidence, lead-gen edition

The Deloitte 2020 study is the cleanest lead-gen-specific evidence in the public record. 30 million sessions across travel, retail, luxury, and lead-gen verticals. Lead-gen-specific finding: a 0.1-second mobile speed improvement lifted form submissions by 21.6% and cut mobile bounce by 8.3% (web.dev case study).

Two things matter about that number. First, it's mobile-specific, which is where most lead-gen traffic actually lives in 2026. Second, 0.1 seconds is a real engineering target. It's not a 3-second-to-1-second hero claim. It's the kind of delta you get from deferring one tracking script, or compressing one hero image, or moving your form embed from synchronous to async.

In our client work, typically the first three fixes on a marketing site recover somewhere between 150ms and 400ms of LCP and 100ms-300ms of INP. If Deloitte's curve holds — and we hedge that, because their study is six years old and the third-party stack has gotten heavier since — the form-fill lift is real and measurable in a 28-day window.

There's a second reason this matters more in 2026 than it did in 2020. Pew Research's July 2025 analysis found that when a Google AI Overview appears, only 8% of visits result in a click to a traditional result, versus 15% without one (Pew). The clicks you still earn have to convert. Speed isn't a vanity score anymore. It's a revenue input.

5. Triage order: what to fix first on a marketing site

Most Core Web Vitals checklists are 40 items long. That's the wrong shape for a marketing team. Here's the order that has worked on the lead-gen sites we've audited.

Marketing-site triage order. Pull field data, fix the failing metric, re-measure after the 28-day CrUX window.

Start with the LCP image. Nine times out of ten on a Webflow, Framer, or WordPress marketing site, the LCP element is the hero image, and it's either too large, not pre-loaded, or served from a slow origin. Replacing a 1.2MB hero with a properly sized AVIF served from a CDN typically recovers 400-900ms of LCP. (If you're on Framer specifically, the platform handles a lot of this for you — we've written about the SEO trade-offs of Framer elsewhere.)

Next, the INP killer. Open Chrome DevTools' Performance panel, record a session where you tap the form field and the chat widget, and look for the long task. Usually one vendor script is doing 70% of the blocking work. Defer it, lazy-load it, or replace it. Intercom and Drift both have lazy-load options that almost no one turns on.

Third, CLS from late-injected embeds. Reserve the height of your form embed with a CSS min-height so the page doesn't jump when the form arrives. Same for hero video, social proof carousels, and webfont swap.

Then re-measure. Not in Lighthouse. In CrUX field data. Which brings us to the next beat.

6. Field data vs lab scores: measuring what Google measures

The single biggest mistake we see marketing teams make with Core Web Vitals is measuring with Lighthouse and declaring victory. Lighthouse is a lab tool. It runs once, on a cold cache, on a simulated device, with no user interaction. Google ranks you on field data — what real users on real devices actually experience over a rolling 28-day window. That's the Chrome User Experience Report, or CrUX.

The practical version: pull your CrUX data from PageSpeed Insights or the CrUX dashboard. Look at the 75th percentile, because that's the threshold Google uses. If your INP at p75 is 280ms, you're failing, even if your Lighthouse score says 92. Lighthouse can't see your INP failure because Lighthouse doesn't interact with the form.

Wait 28 days after each fix before declaring the win. CrUX is a trailing 28-day window, so a fix shipped on day one doesn't fully reflect in the field data until day 29. We've watched teams panic on day three and ship a rollback for a fix that was going to work fine.

7. When Core Web Vitals stop being the bottleneck

There's a point where speed work stops paying. We've seen it. A marketing site with an LCP of 1.6 seconds, INP of 180ms, CLS of 0.04 — all comfortably in the green — and a form-fill rate of 1.2%. The next 100ms of speed isn't the lever. The copy is. Or the offer. Or the form length. Or the fact that the form asks for company size, job title, and phone number on the first screen.

The operator's read: Core Web Vitals are a hygiene metric. Get them into the green, then stop tuning. The next conversion lift comes from offer clarity, form friction, and the redesign questions we've written about here. If your CWV is already green and your form-fill rate is still flat, the problem is no longer technical. It's the page itself.

This is also where the conversation widens. A marketing site that converts isn't a fast site. It's a fast site with a sharp offer, a single CTA per page, and a form that asks for the minimum viable information. The speed work is the entry ticket. The conversion work is the website design problem.

FAQ

Which Core Web Vital matters most for form conversions?

INP, in most lead-gen audits we run. LCP affects whether the visitor stays long enough to read the offer, but INP determines whether the form responds when they tap it. Most marketing-site INP failures come from third-party form embeds and chat widgets blocking the main thread, not from first-party code.

What's a good INP score for a marketing site in 2026?

Google's threshold is 200 milliseconds or less at the 75th percentile, measured across all interactions on the page over a 28-day window. That's a hard target for sites with multiple vendor scripts. Audit your form embed and chat widget first — they're typically the largest contributors to INP failures.

How do I fix LCP on a lead-gen website without a developer?

The highest-impact fix is the hero image. Compress it, convert to AVIF or WebP, set explicit width and height attributes, and add fetchpriority="high" to the image tag. On platforms like Framer, Webflow, and Squarespace, the image-optimization settings handle most of this if you enable them. That alone often moves LCP under 2.5 seconds.

Does page speed actually affect form fills on B2B sites?

The cleanest public evidence is Deloitte's 2020 Milliseconds Make Millions study of about 30 million sessions, which found a 0.1-second mobile speed improvement lifted lead-generation form submissions by 21.6%. The study is six years old, so we hedge it — but the direction and magnitude have held up in client work.

Should I use Lighthouse or PageSpeed Insights to measure Core Web Vitals?

Use PageSpeed Insights and read the CrUX field data section, not the Lighthouse lab score. Google ranks pages on real-user field data over a rolling 28-day window. Lighthouse runs once in a simulated lab environment and doesn't measure INP from real interactions, so it routinely shows green when your real users are failing.

A close

Run this one this week: pull your homepage CrUX data from PageSpeed Insights, find your weakest metric at the 75th percentile, and ship one fix. Measure again in 28 days. If it's INP, the fix is almost certainly a vendor script.

If you'd rather hand the triage off, let's talk.