Cloudflare, Netlify and Vercel Sell Web Speed. In Google's Field Data, Their Own Sites Are Among the Heaviest of 49 Software Companies We Checked.

Cloudflare, Netlify and Vercel are in the business of making the web fast. Cloudflare runs a global network that speeds up other people's sites; Netlify and Vercel exist to deploy fast web front-ends. In Google's own field data, their own sites are among the heaviest of the 49 major software companies we checked, with Cloudflare measured at a 10.9-second load on mobile, the slowest in the set. Across all 49, just four pass Core Web Vitals.

For software companies this is not an idle curiosity. Speed is where real users live, it shapes conversion and trust, and Core Web Vitals has been a Google ranking signal since 2021. Google's research links slower pages to sharply higher abandonment, and every heavy reading is a moment a prospect or customer waits on a screen. That the companies most associated with modern web engineering post some of the heaviest numbers is the surprising part, and it has a clean technical explanation rather than a failure of skill.

The measurements come from the Chrome User Experience Report, Google's public record of what real Chrome users experienced on real devices and networks, the same field signals Google's ranking systems use. It reflects how much a modern software front-end does, not the quality of the tools these companies ship, and it lands the same way for everyone: publicly, on real phones.

We compiled 49 of the most widely used software and SaaS companies and queried the Chrome User Experience Report for each brand's registrable domain, taking its busiest origin, which for a software company is typically the application itself rather than the marketing homepage, measured on mobile at the 75th percentile. A missing INP counts as failing.

A company passes only when all three metrics land in the good range at the 75th percentile: Largest Contentful Paint under 2.5 seconds for loading, Interaction to Next Paint under 200 milliseconds for responsiveness, and Cumulative Layout Shift under 0.1 for visual stability. Every value below is a real reading in Google's public dataset, and the four that pass show the bar is reachable even for a busy software surface.

There is a common assumption that the companies building the web are fast at the web. We put it to a direct test, pulling real-world Core Web Vitals for 49 major software companies, from Oracle and SAP to Figma, Notion and Stripe. Eight percent pass. Forty-five of the forty-nine do not clear the bar on mobile in the May 2026 dataset.

The more interesting result is which four pass. Not the newest, most design-forward developer tools, but Oracle and SAP, two long-established enterprise platforms, plus two focused utilities, SurveyMonkey and Zendesk. What they share is restraint on their busiest property: lighter pages, less client-side rendering, fewer heavy interactive surfaces. It is not about age; it is about how much the page has to do before a user sees and can use it. Many of the tools that win design awards and lead developer conversation show heavier field data, which reflects how much interactive work their busiest surfaces do.

Three of the heaviest readings in the sample belong to companies whose work centers on web infrastructure. Cloudflare shows the highest loading time in the set, a 75th-percentile LCP of 10.9 seconds. Netlify reads at 7.7 seconds. Vercel, the maker of Next.js, shows a 400 ms INP on responsiveness. Even companies that sell web performance run product sites heavy enough to feel it, and their field data is public in CrUX like everyone else's.

The pattern behind the numbers is clear, and it is a genuinely hard engineering problem, not a lack of care. Only 12.5% of these sites have good loading (LCP) and only 29% have good responsiveness (INP), while 82% have good layout stability (CLS). That is the signature of a heavy JavaScript application. Software companies run some of the most JavaScript-heavy properties on the web: canvas engines, real-time collaboration layers, live dashboards and graphics pipelines all run in the browser, with marketing stacks, analytics and tag managers layered on top.

The busiest property for a software company is usually the logged-in application, where all of that weight lives, so when you measure a software company's real-world field data you are mostly measuring its app, and its app is heavy by design. That is also why the more focused tools read lighter: a survey tool or a help-desk widget does not need a canvas engine, a real-time collaboration layer or a graphics pipeline. The most capable software is often the heaviest, and Core Web Vitals measures delivered experience rather than ambition.

The full per-domain table shows the whole set at once, sorted by load time. Layout stability holds up across nearly every brand; loading and responsiveness are where the heavy-app signature shows and where the pass rate is lost.

Pass rate across the full 49-brand sample is 8.2% (4 of 49): good LCP 12.5%, good INP 29.2%, good CLS 81.6%.

For software companies the lesson is practical. If Cloudflare at a 10.9-second LCP, Netlify at 7.7 and Vercel at a 400 ms INP post these readings, any engineering-led company can, because the pressure is the same everywhere: capability ships faster than it gets optimized, and only 12.5% of these 49 sites hold a good LCP while 29.2% hold a good INP. The busiest, most-used surface, usually the logged-in app rather than the marketing homepage, is exactly where a performance budget repays itself, through main-thread discipline and code-splitting. Oracle, SAP, SurveyMonkey and Zendesk already clear the bar on a busy product surface, which is proof the 8% pass rate is a choice rather than a ceiling.

For anyone buying software, the reading cuts the other way. A vendor's own field data is not a reliable signal of how fast the product will feel. A heavy product-site reading, like the 45 of 49 that fail here, mostly means the product site is heavy, as most in this category are. Figma at 8.2 seconds and Notion at 6.5 still ship tools their users rate highly. Judge the tool on your own Core Web Vitals once it is in your stack, and treat that number, not the vendor's site, as the one worth moving.