HTTP Archive’s latest Core Web Vitals Technology Report compares seven content management platforms, and the headline finding cuts against a common assumption: neither page weight nor Lighthouse lab scores reliably predict how a platform actually performs on Core Web Vitals in the real world. The platform that should have struggled the most, Shopify, finished third. The platform with the lightest pages and the best Lighthouse scores, Astro, only finished fourth.

That gap between lab predictions and field results is the real story here, and it has direct implications for anyone deciding what to build on or how to prioritize performance work.

What the report actually measures

The comparison blends two different data sources. The Chrome UX Report (CrUX) supplies anonymized, real-world field data from actual Chrome users who’ve opted into sharing usage statistics — this is what determines whether a site gets a “good” Core Web Vitals score in Google’s eyes. HTTP Archive supplies lab-based data, crawling and testing sites to generate Lighthouse scores and median page-weight figures.

Pairing the two datasets is what makes this comparison useful: it lets you see whether a platform’s lab-test performance (Lighthouse) and its theoretical payload size (page weight) line up with what real users on real devices and real connections actually experience (CrUX). Often, they don’t.

The rankings, and the platform that breaks the pattern

Duda topped the field with roughly 85% of sites earning a good Core Web Vitals score, paired with one of the lightest median page weights in the comparison at about 1.78 MB. Wix followed at 80%, with a heavier median page weight around 2.55 MB but still lighter than several platforms that scored worse.

Then the pattern breaks. Shopify placed third at roughly 79% of sites scoring well — a genuinely surprising result, given that ecommerce sites typically carry the kind of third-party scripts, tracking pixels, and dynamic storefront features that tend to drag performance down. Shopify’s numbers back up that expectation on paper: it posted the heaviest median page weight in the entire comparison, at 3.77 MB, and one of the weakest Lighthouse scores, at 47. By every lab-based signal, Shopify sites should have finished near the bottom. In the real world, measured by actual users, they didn’t.

The rest of the field: Astro landed fourth at 67%, despite having the single lightest median page weight (roughly 1.65 MB) and the best Lighthouse score (68) of any platform tested. Drupal came in fifth at 64%, holding steady at that number from January through April 2026 without meaningful movement in either direction. Joomla ranked sixth at 58%, despite a page weight (2.53 MB) that was actually lighter than several higher-scoring platforms — and it posted the lowest Lighthouse score of the group, at 43. WordPress finished last at 49%, with both a heavy median page weight (approximately 2.63 MB) and a near-bottom Lighthouse score (44).

Why Shopify’s real-world scores don’t match its lab scores

The likely explanation is that Lighthouse and CrUX are measuring different things. Lighthouse audits are sensitive to factors like JavaScript payload, unused JS, render-blocking resources, image inefficiency, and network waterfall structure — all measured under synthetic, throttled lab conditions. CrUX measures something different: how the page actually behaves for real users, including the effects of caching, CDN delivery, repeat visits, and a wide range of real devices and connection speeds.

Shopify’s infrastructure appears to be optimized specifically for that second category. Stable rendering behavior, layouts coded to avoid shifting, fast initial interactivity, and aggressive CDN-level resource delivery are exactly the qualities that reward a platform in field data even when its raw payload size and lab scores look weak. In other words, Shopify may ship more bytes than its competitors, but it appears to be unusually disciplined about how efficiently those bytes get delivered and rendered once they reach a real browser on a real network.

Astro’s advantage may say more about its typical use case than its architecture

Astro’s combination of the lightest pages and the best lab scores, yet only fourth place in real-world results, is worth sitting with. Part of the explanation may be selection bias: sites built with Astro tend to be simpler by nature — blogs, documentation, marketing pages — rather than the complex, interactive applications that stress-test a platform’s architecture. WordPress performs reasonably well out of the box too, for similar reasons on simple builds. It’s plausible that Astro’s out-of-the-box advantage would narrow if it were tested against sites carrying the same functional complexity Shopify or Wix routinely handle.

Page weight and Core Web Vitals correlate, but weakly

Looking at the full dataset, lighter median page weight does generally track with better Core Web Vitals scores — Duda and Astro both combine low weight with strong-to-decent field performance, and WordPress combines heavier pages with the weakest field performance. But the correlation isn’t tight enough to treat as a rule. Joomla proves the point in one direction (light pages, weak scores) and Shopify proves it in the other (heavy pages, strong scores). We explored a similar disconnect between raw file size and real outcomes when covering why Google says page weight alone is a misleading metric — the same principle appears to hold at the platform level here.

The more accurate framing is that page weight is one input into performance, not the determining factor. What seems to matter more is how efficiently a platform’s architecture handles the complexity a real site actually needs to carry.

Where Core Web Vitals scores actually break down

The three metrics that make up Core Web Vitals each fail for different, specific reasons, and understanding those failure points matters more than chasing a lighter page weight in isolation.

• Largest Contentful Paint (LCP) suffers when platforms load oversized images, delay discovery of the primary above-the-fold image, or let render-blocking CSS and JavaScript compete with the main content for priority. A site can have a modest total payload and still deliver a slow LCP if the browser can’t quickly identify what to render first.
• Interaction to Next Paint (INP) degrades when third-party scripts, tracking tags, chat widgets, sliders, and heavy JavaScript execution occupy the browser’s main thread. Every additional feature on a page is competing for the same limited execution time, which is exactly why feature-heavy ecommerce platforms would be expected to struggle here.
• Cumulative Layout Shift (CLS) breaks when content moves after it has already rendered — images without reserved dimensions, late-loading ads, embedded media, or dynamically injected interface elements pushing the page around while a visitor is trying to interact with it.

Shopify sites carry a lot of the exact ingredients that typically damage all three of these metrics, and their Lighthouse scores reflect that. Yet their field performance holds up regardless, which strongly suggests the platform is winning the execution battle even while losing the raw-payload battle. That’s consistent with what we found when looking at how Googlebot’s byte limits interact with crawling architecture — the size of a resource matters less than how well the platform manages the delivery and rendering pipeline around it.

The real takeaway

Core Web Vitals performance isn’t a contest to see which platform ships the smallest pages. It’s a contest to see which platform handles real-world complexity without letting that complexity translate into a slow, unstable, unresponsive page for the person using it. Anyone comparing platforms — or auditing their own site’s Core Web Vitals performance against open-source and proprietary alternatives — should weigh execution quality at least as heavily as raw page weight.

Frequently asked questions

Does a lighter page always mean better Core Web Vitals scores?

No. There’s a general correlation, but it’s weak. Shopify posted the heaviest median page weight in this comparison and still finished third in real-world Core Web Vitals scores, while Joomla had lighter pages than several better-performing platforms and still finished near the bottom.

Why did Shopify score well on Core Web Vitals despite poor Lighthouse scores?

Lighthouse measures lab conditions like JavaScript payload and render-blocking resources, while Core Web Vitals field data measures real user experience including caching, CDN delivery, and repeat visits. Shopify’s infrastructure appears optimized for stable, fast real-world delivery even though its raw payload and lab audit results are weak.

Which CMS ranked best for Core Web Vitals?

Duda ranked first with about 85% of sites earning a good Core Web Vitals score, followed by Wix at 80% and Shopify at 79%. WordPress ranked last among the seven platforms compared, at roughly 49%.

Should site owners still try to reduce page weight?

Yes, but not as the only lever. Reducing unnecessary payload helps, but this data shows that how efficiently a platform renders and executes what it loads — avoiding layout shifts, minimizing main-thread blocking, prioritizing the right content first — matters just as much as the total size of the page.

Why did WordPress rank last in this comparison?

WordPress combined a relatively heavy median page weight (about 2.63 MB) with one of the weakest Lighthouse scores in the group, and that combination showed up in its real-world Core Web Vitals results too, with only about 49% of sites scoring well.