SquishPixel Blog · Core Web Vitals · Web performance · Front-end

Webflow sites pass Core Web Vitals in the field, but multi-page lab tests tell a harder story

According to DEV Community, a new benchmark compares two ways of measuring Webflow performance. The first is field data from the Chrome UX Report (CrUX), which records what real Chrome users experience. The second is lab data from Lighthouse, which tests pages under throttled conditions. The author argues that neither replaces the other.

On the field side, the article cites rebuilds of the HTTP Archive Core Web Vitals Technology Report for September 2026. They put about 71.3% of Webflow mobile origins as passing all three Core Web Vitals, against about 52.8% of all measured origins. Among Webflow sites in the top one million origins, the pass rate falls to about 49.8%. Loading, measured by Largest Contentful Paint (LCP), is the metric that fails first.

The lab side is a sample from the author's own monitoring tool, Watcher: 109 Webflow client sites, with reports generated between 1 September and 15 November 2026. Each report averages results across a median of 15 pages per site, covering the home page, CMS templates and conversion paths. The article says the median mobile Performance score was 55 and the desktop median was 78. Only 3 of 109 sites reached a mobile average of 90.

LCP was the structural problem. Only two sites averaged 2.5 seconds or better across the selected pages. Layout shift (CLS) was the bright spot. Interaction delay (INP) sat just above the lab "good" line at the median. The most common Lighthouse opportunities were unused JavaScript and unused CSS. Server response almost never appeared as a named opportunity.

Why it matters

A site that passes in the field can still carry slow templates. The article notes that CMS item pages, blog lists and form pages may not get enough traffic to show up in CrUX at page level, yet visitors still meet them after the first click. Testing only the home page can hide that.

The findings also point to where effort pays off. In this sample the lab problems were mostly front-end weight, not hosting. Large hero images, site-wide scripts and animation libraries are the likely first targets. Note that the sample is the publisher's own client base, so treat it as a rough guide rather than a ranking of platforms.

Source: DEV Community. This brief summarises their reporting; read the original for the full detail.