Page Speed vs. Conversion Rate: What Our Data Reveals

The standard line goes something like this: "Every second of load time costs 7% in conversions." It gets repeated in every marketing conference, every website optimization guide, every pitch deck for performance tools. And it's not wrong — but it's misleading in a way that causes people to optimize the wrong thing.

We have PageSpeed data from thousands of landing pages analyzed through roast.page. Real Google Lighthouse scores, real Core Web Vitals, correlated against our 8-dimension conversion scores. Here's what the data actually shows — including the parts the speed-optimization industry doesn't want to talk about.

The Correlation Exists — But It's Weaker Than You Think

Yes, faster pages score higher on average. Pages with LCP under 2.5 seconds (Google's "Good" threshold) score an average of 54/100 on our overall conversion assessment. Pages with LCP over 4 seconds average 46/100.

That's a real difference — 8 points. But here's the thing: 8 points out of 100 means speed accounts for a fraction of conversion outcomes. The Top quartile pages average 72. Getting from 46 to 72 requires a lot more than faster loading.

The Threshold Effect: It's Not Linear

The most important finding in our data: the speed-to-conversion relationship has a massive threshold effect.

• 5+ seconds → 2-3 seconds: Huge improvement. Pages that move from "Slow" to "Good" on LCP see an average 6-point improvement in overall score and a dramatically better user experience.
• 3 seconds → 2 seconds: Noticeable improvement. Worth doing if you can — about 2-3 points of overall improvement.
• 2 seconds → 1.5 seconds: Marginal. Less than 1 point of score improvement.
• 1.5 seconds → 1 second: Effectively unmeasurable in our data. The conversion difference between a 1-second and 1.5-second page is lost in the noise of other variables.

This has a practical implication: if your LCP is 4 seconds, speed optimization is your single highest-priority fix. If your LCP is 1.8 seconds, you'll get more conversion lift from rewriting your headline than from shaving another 300 milliseconds.

Which Core Web Vital Matters Most?

We looked at each metric individually. The ranking by correlation with overall conversion score:

1. LCP (Largest Contentful Paint) — Strongest correlation

LCP measures when the biggest element loads — usually your hero image or headline. It has the strongest correlation because it directly determines when visitors can start processing your message. If the hero loads slowly, the 5-second clock starts ticking before your value proposition is even visible.

Google's thresholds: Good (<2.5s), Needs Improvement (2.5-4s), Poor (>4s). In our data, 54% of pages fall in "Needs Improvement" or "Poor."

2. CLS (Cumulative Layout Shift) — Second strongest

CLS measures how much the page jumps around while loading. High CLS doesn't just annoy visitors — it actively undermines trust. When a button moves while you're trying to click it, or a testimonial pushes your CTA off screen, it feels broken. Pages with CLS under 0.1 ("Good") score 3.2 points higher on Trust than pages with CLS over 0.25.

The trust connection surprised me. I expected CLS to affect the Technical dimension, which it does. But it also correlates with Trust scores because visual instability signals a lack of polish and professionalism.

3. FCP (First Contentful Paint) — Moderate correlation

FCP measures when something — anything — first appears on screen. It matters because a blank white screen creates anxiety. Even a loading indicator is better than nothing. But once you have a Good FCP (under 1.8s), further improvement has diminishing returns.

4. INP (Interaction to Next Paint) — Weakest correlation

INP measures how fast the page responds to interactions. On landing pages specifically, INP matters less than on web applications because landing pages are primarily read-and-click experiences, not interactive tools. That said, a laggy CTA button that takes 300ms to respond does create a subliminal "this feels broken" impression.

The "Fast But Broken" Paradox

This is the finding that should give speed-obsessed teams pause. In our dataset, 12% of pages with a PageSpeed performance score of 90+ still score below 40/100 overall.

Fast but broken. Pages that load instantly but have vague headlines, no social proof, invisible CTAs, and copy that reads like product documentation. They pass every technical test and fail every conversion test.

I saw a developer portfolio page that scored 99 on PageSpeed — static HTML, minimal CSS, no JavaScript. A technical masterpiece. It also scored 28 on conversion because the headline said "Welcome to My Portfolio," there was no CTA, and the only social proof was a GitHub contributions graph. Blindingly fast. Completely ineffective.

The inverse also exists: pages with mediocre PageSpeed scores (60-70 range) that score 70+ overall because their messaging is razor-sharp, their trust signals are overwhelming, and their CTA path is frictionless. Speed didn't kill them because they crossed the "good enough" threshold and nailed everything else.

Industry-Specific Speed Benchmarks

Speed expectations vary by industry because page complexity varies:

• SaaS: Average LCP 2.8s. SaaS visitors expect fast, polished experiences. Pages under 2s score notably higher.
• E-commerce: Average LCP 3.4s. Product images are the main bottleneck. Image optimization is the single biggest lever.
• Healthcare: Average LCP 3.1s. Slightly more tolerant of speed because the decision is high-stakes regardless.
• Real estate: Average LCP 3.8s. Heavy property galleries tank performance. The worst industry for LCP in our data.
• Agency/consulting: Average LCP 2.4s. Tend to be content-light, so speed is naturally better.

What Actually Predicts Conversions?

If speed accounts for a minority of conversion outcomes, what predicts the rest? From our dataset of thousands of pages, the variables ranked by impact on overall score:

1. Headline quality (Copy & Messaging dimension) — The single strongest predictor. A specific, outcome-driven headline is worth more than a 50% speed improvement.
2. CTA effectiveness — Copy specificity, contrast, and placement. See our CTA analysis.
3. Trust signals — Presence and quality of social proof above the fold. Pages with strong trust signals compensate for moderate speed issues.
4. Visual design quality — Professional design signals credibility. A fast-loading page that looks amateur loses trust instantly.
5. Page speed — Real and measurable, but fifth on the list. Not first.

Practical Guidance: When to Optimize Speed vs. Other Factors

Optimize speed first if:

• Your LCP is over 4 seconds (you're in the "losing visitors before they see your message" zone)
• Your CLS is over 0.25 (your page is visually unstable and actively undermining trust)
• Your PageSpeed performance score is under 50 (there are likely easy wins with image compression and script deferral)

Optimize messaging/conversion first if:

• Your LCP is under 2.5 seconds (you're in the "good enough" zone)
• Your overall conversion score is under 50 despite decent speed (the problem is strategy, not performance)
• You're spending more time on speed optimization than on page testing (diminishing returns territory)

The fastest path to higher conversions for most pages: fix your headline and CTA copy first (highest impact, lowest effort), add trust signals above the fold (second-highest impact), then optimize speed (necessary but not the primary lever). Run your page through our page speed analyzer to see where you stand on technical performance, or use the full analyzer for the complete conversion picture.

Speed matters. But it matters less than what you say to visitors once they arrive — and the data from thousands of pages backs that up.