CLS discrepancy between CrUX dashboard & Search Console?

[Gemma Reeves]

Can someone help me understand why discrepancy exists between CrUX dashboard & Search Console? (when they both say they come from the same data) Thanks

[Barry Pollard]

The CrUX Dashboard looks at origin-level data only.

Google Search Console (GSC) attempts to group pages into "page groupings".

So take for example a site and say it has the following page views:

• / (i.e. the home page) - 10,000 page views with passing CLS
• /blog/page1 - 5 page views with failing CLS
• /blog/page2 - 5 page views with failing CLS
• /blog/page3 - 1 page view with passing CLS
• /other - 1 page view with failing CLS.

In that case, the CrUX dashboard may show the CLS as passing because those 10,000 home page views are the vast majority of the whole origin.

However, GSC may show the following:

• Group 1 - 10,000 page views, including the home page with page-level data and /other without page-level data (as insufficient traffic). This group is passing CLS overall.
• Group 2 - the 3 blog pages with 3 example pages, all without page-level data (as insufficient traffic). There may also be 500 blog pages in total in those group but 497 of those aren't visited at all. This group is failing CLS overall.

Note, the CrUX team (of which I'm part) don't work on GSC and don't have any additional visibility into how it groups pages, so take the above (extreme!) example as just an example of our understanding of how they display this data.

So The CrUX Dashboard makes it look like everything is good, but GSC says one group is passing, one is failing, and that 500 our of 502 pages (99.6% of pages) on your site are failing (as 500 pages are in the failing group) and only 2 are passing.  As I say, this is an extreme example with one page so heavily outweighing the others.

GSC is only available to site-owners that have verified their ownership, whereas the CrUX Dashboard is based on public data, and hence shows less detail.

By grouping pages into "groups", GSC is trying to be helpful to show you pages that likely have similar root causes so you can tackle them separately. For example, maybe it's your /blog pages that all can be fixed with one common fix to that page template to address the issue for all pages in that group. In my experience this usually works pretty well for well-trafficked sites.

However GSC's page grouping is also prone to confusing people too. For example, the following scenarios can be confusing:

• This (extreme) example, where CrUX and PageSpeed Insights says all is fine at an overall origin-level, but GSC says 99.99% of pages are not fine. Both are "correct" based on their view of the data, but it can seem contradictory and so confusing.
• When the example pages with sufficient traffic to have page-level data are passing, but the group overall is failing (as it's being dragged down by the long tail of other, slower, pages). Then the example pages aren't actually that helpful to diagnose issues!
• When the pages in a group are actually as similarly fast as other pages that are passing when tested, but they are just visited so infrequently they often are not cached on infrastructure and so take a bit longer to show to real users (compared to cached pages which display quicker for most users).
• When the pages in a group are actually as similarly fast as other pages that are passing when tested, but just so happen to be visited by users on slower connections/devices (e.g. if the blog article is for "the top cheapest mobile phones", then it might appeal to those currently using slower devices and hence the page does load slower for those readers).
  

I hope that helps and note all GSC questions will need to be directed at Search. Again, above is my understanding based on public documentation of how this works and my own experience.

[Gemma Reeves]

Thanks for the in-depth explainer, I've re-read a few times but not sure I can make that logic fit our scenario where CruX shows 10% poor CLS and GSC shows 0. Makes it difficult to troubleshoot as no clue which page template might be the culprit

[Barry Pollard]

What do you mean "CruX shows 10% poor CLS". Can you show what you mean with a screenshot?

--
You received this message because you are subscribed to the Google Groups "Chrome UX Report (Discussions)" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chrome-ux-repo...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/chrome-ux-report/b559ae4e-426d-4300-871c-dfe8cbbbc451n%40chromium.org.

[Barry Pollard]

OK, so we recommend measuring Core Web Vitals at the 75th percentile:

To ensure you're hitting the recommended target for these metrics for most of your users, a good threshold to measure is the 75th percentile of page loads, segmented across mobile and desktop devices.

Tools that assess Core Web Vitals compliance should consider a page passing if it meets the recommended targets at the 75th percentile for all three of the Core Web Vitals metrics.

There will ALWAYS be some users that experience worse performance and it's not possible to guarantee that 100% of users can make any reasonable measure of "good". The world is a messy place, with a mix of users, connection, devices, and even random glitches.

So we say that if 75% of your page views meet the Core Web Vitals thresholds then you pass Core Web Vitals. There's more detail on why we picked the 75th percentile here, but roughly we want to know the majority of your users are satisfied, but also that measure isn't too much impacted by outliers.

This screenshot shows that 80% of your page views on your origin (i.e. your site) meet the good threshold, and therefore you pass CWVs, so it's not surprising that Google Search Console says all your pages are passing (i.e. all your pages have at least 75% of page views that are passing the Core Web Vitals thresholds).

The reason the CrUX dashboard shows the full breakdown is it can be useful to see if you're improving or getting worse over time. Here you can see you are degrading month on month, and although you are passing in each month, if this trend continues then you will start to fail. So you might want to find out why you are degrading and fix it before that happens!

On Thu, 15 Aug 2024 at 14:36, Gemma Reeves <ge...@avforums.com> wrote:

[Gemma Reeves]

Thank you this has been really helpful!