Counting SPA navigations lifts the CrUX pass rate four points
Summary
Counting single-page app navigations lifts the share of origins passing all three Core Web Vitals by four points in October's CrUX data, and no site got faster.
Pull both CrUX series for the same URLs now and work out which routes explain the difference, before the metrics graduate in May 2027.
The October 2026 CrUX dataset passes all three Core Web Vitals on 60.3% of origins when single-page app navigations are counted, against 56.3% on the metrics Google reports today. Harry Roberts, an independent web performance consultant, put the two series side by side in Web-Perf Wednesday 018.
A soft navigation is a route change inside a single-page app that the user reads as a new page: the URL changes and the main content swaps, but the browser never loads a new document. The standard metrics treat that whole session as one page load, so a slow first paint and every later layout shift and interaction land on the same record. The experimental series splits the document at each soft navigation and scores each one on its own.
Every metric improves under that split, across the dataset’s 18 million origins.
| Metric | Standard | With soft navigations |
|---|---|---|
| Good LCP | 68.8% | 73.6% |
| Good CLS | 81.6% | 85.0% |
| Good INP | 85.2% | 87.0% |
| All three | 56.3% | 60.3% |
No site got faster between the two columns. Roberts is explicit that the measurement changed rather than the pages. Later views in a session reuse the resources and application state the first load already paid for, so their largest paint is cheap. CLS and INP also accumulate within each navigation rather than across the whole session, so a first-load shift or a sluggish first tap is no longer charged to every view that follows.
That accounting becomes the official one if Google graduates the experimental metrics in May 2027, the target date in Roberts’ write-up, with Search Console expected to adopt them afterwards. Until then the current series stays the official data. The origins in that four-point band would start passing in Search Console without a single deploy. The aggregate hides both directions, though, and the likelier picture is that some origins move the other way: an app whose route changes are slower than its initial load has nothing to gain from being scored per navigation.
The comparison can be run today. The CrUX APIs now also expose experimental_first_contentful_paint_with_softnavs and experimental_navigation_types_with_softnavs, and BigQuery gets them in the next monthly release. The soft navigation series itself is older: when Chrome first added it, single-page apps looked much faster than the reports their owners were being judged on. Roberts’ advice is to pull both series for the same URLs and form factors, record the size and direction of the difference, then identify the routes responsible for it.
His second check is whether your real user monitoring (RUM) provider counts the same navigations Chrome does. A dashboard that still treats the session as one page load will not match the series Google is about to promote. Vendor instrumentation has been wrong about Core Web Vitals before, as when a monitoring script kept fast interactions out of INP for years.
What to do
Pull both series for the same URLs and form factors you already track, and compare them metric by metric rather than on the pass rate alone. Where the two disagree, find the routes behind the gap. Then ask your RUM vendor whether its SDK detects soft navigations today and how it attributes LCP, CLS and INP across them. If the answer is no, that dashboard and Search Console will drift apart once the metrics graduate.