Web vitals missing on Crux API since update ?

    "origin":"https://kr.louisvuitton.com",

    "formFactor": "PHONE"

}

From time to time i even get straight up a 404.

I was thinking that it could be due to low traffic on chrome users but after checking i can see that other subdomains with lower traffic on chrome dont' have the issue.

Any idea on why this is happening ?

Kind regards

Quentin

Thanks to both of you for letting us know.

I have looked into the origin mentioned in Quentin's post (https://kr.louisvuitton.com). My understanding is that this particular origin is affected by not quite hitting our requirements for "sufficiently popular" (https://developer.chrome.com/docs/crux/methodology#popularity-eligibility).

Unfortunately this can play out in a somewhat confusing manner: Because LCP has slightly fewer recorded measurements than FCP, we may not be able to include it when we do include FCP.
But then also, when we're right at the edge of the popularity requirement and need to drop the data for a form factor, we require that the overall data (=all form factors) has at least 80% of the traffic in it (https://developer.chrome.com/docs/crux/methodology#filtering).
So this can have the effect of dropping (for example) phone-specific data when the desktop data can't be included but has more than 20% (e.g. imagine it's 70% phone and 30% desktop and desktop can't be included). But in that case ... the overall data would be available.

For the specific example origin "https://kr.louisvuitton.com": Whereas ~yesterday, when we were serving 2024-04-09, we were missing LCP for form factor 'phone' for the first issue, ~today, we're serving 2024-04-10 and so we're returning 404 for phone for the second issue (probably, I haven't checked this 100%) - but if you make a request without form factor { "origin":"https://kr.louisvuitton.com" } you actually get overall data today (including LCP).

Hope this helps a bit. If you do see things that don't make sense, please point them out and examples are definitely useful and appreciated.

Good wishes,

Johannes.

I think this is a limitation of this dataset - so if you have your own RUM it's probably a better way to compare such country specific sites, or ratios between form factors.

Good wishes!

Thank you!

I do think this origin is also affected by the popularity thresholds, and this is also true for the form factors.

Quickest way to see it is in the CrUX History API, or via CrUX VIs, e.g. this link shows the CWV history:

and clicking between "Desktop" and "Phone" in the controls area one can see that there were stretches when these form factors had individual reporting.

And, there is a "Form Factors" metric, it's in this view at the bottom (can be expanded):

There is some data at times, but because frequently there wasn't enough traffic to include the detailed data points, there's no fractions or it ends up being 0% and 100% which is distortion that comes from our aforementioned filtering (up to 20% can be dropped).

Btw - another way to do a quick check on "popularity" according to CrUX is rank magnitude (https://developer.chrome.com/blog/crux-rank-magnitude). In this case, the rank magnitude is 50M (smaller means more popular). This was my query:

So, I'd say this is a limitation with CrUX / what we can publish, not with your site.

You could measure things with your own Real User Monitoring (RUM).

This is Barry's article on the topic: https://web.dev/articles/crux-and-rum-differences

Good wishes,

Johannes.