PageSpeed CrUX data updates cadency

[simone luise]

Hello there,

I'm happy to see that there is a community behind the strange world of "Core Web Vitals" that, as a developer, is my pain and also a wonderful discover.

However, I need to you to help me only to understand two concepts that I've seen there:
https://developer.chrome.com/docs/crux/methodology/tools?hl=en

In my understanding the CORE thing is the period, but THE MUST IMPORTANT THING is that CrUX insight need real user tracking data to works.

So my question is, I've a site that 80% of the traffic is on Phones but I've an issue on CLS on DESKTOP, I'm assuming to have fixed it (the mitigation on Google Search Console) is "pending", but I'm asking to myself:

If I've released the fix on 8th may when I should have a feedback on Different instruments?

• Google Search Console
  I've to wait 28 days from the request to validate the fix?
• CrUX vis
  The last date is 9th may so I'm expecting to have some improvements? Or 1 day of good (hope) is not enough to mitigate 27 days of BAD trackings?
• pagespeed.web.dev
  The report told me that the score refers to a period that was ended on 13th May, so as before I'm expecting to have an increase? Or 4 days is not enough? or also the pagespeed refers to CrUX data that is updated only on 9th?

We have delivered so many fixes and improvements but we seems to have no results, I'm confident that you know it's just a matter of time, but I would like some confirmation.
In the site we have a 0.4-0.5 CLS issue on all desktop pages BUT, with the release of 8th may, I'm sure to have fixed the hugest CLS issue that affect our sticky desktop header (so every page on every scroll on desktop)... in my live tests using performance devtool, before release I've 0.4-0.5 CLS issue (as highlighted by all instruments) but after the released fix the CLS is near to 0 (when third party scripts not add strange content).

Any help is appreciated.

[Barry Pollard]

All of the CrUX data, in all three tools, is based on 28-day aggregations.

Or 1 day of good (hope) is not enough to mitigate 27 days of BAD trackings?

Yes this is it basically. You often don't have to wait the full 28 days to see some improvement (especially in the histograms as opposed to the 75th percentile number). Also as the 75th percentile high-level number is based on 75% of your users, after 21 days out of 28 you should definitely see some impact. But you may not get the full impact until the full 28 days has passed.

Before joining Google, I wrote about this all here: https://www.smashingmagazine.com/2021/04/complete-guide-measure-core-web-vitals/#the-chrome-user-experience-report-crux. The main change since then is we've made the histograms easier to see in PageSpeed Insights when you click Expand View:

So if you have really fixed the issue, you should see the number of "Good" pages tick up, and "Poor" tick down. And eventually the p75 number at the top will switch too when 75% of your users are good.

Note also that CrUX Vis only updates weekly on Mondays. And each of those will be the previous 28-days prior to that. For more information on what the various tools show, see here: https://developer.chrome.com/docs/crux/methodology/tools

--
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 visit https://groups.google.com/a/chromium.org/d/msgid/chrome-ux-report/abd07121-fdc8-4f85-89fa-75d4c298dc8bn%40chromium.org.

[simone luise]

Hi mate, may thanks to your quick response, appreciated.

I've read your article and It's awesome, more or less you describe what is in my confused head, so now I'm in the correct path (thanks to you).
Bad news for me... only 23% of our site traffic is made using Chrome (5% using Chrome - Desktop - a few tens of thousands), so probably we should have data availability issue on Chrome Search Console.

I think that we have to add the CrUX tracking on the page (we're full of useless third party script, we can add one useful) in order to have fresh data for our debugging, because it's the point, is hard to debug something that happen to a percentage of users without having any clear informations about the real bug... only best practice and real life testing (not enough seems).

I hope to be useful for some other in the future as you are for me now.
I'll be back for some updates when I've them.

Many thanks!

[simone luise]

Hello mates,

as described in the article that Barry (I'm falling in love with you) are exactly what was happen on my CLS KPIs.
Thats the CLS of our homepage after the 14/05 fix released on production:

- 0.44 CLS (14/05) - BAD
- 0.43 ------ (15/05)
- 0.43 ------ (16/05)
....
- 0.43 ------ (18/05)

- 0.42 ------ (19/05)

- 0.41 ------ (20/05)
- 0.41 ------ (21/05)
- 0.36 ------ (22/05) - first real drop
- 0.15 ------ (23/05) - finally "NEED IMPROVEMENTS"

- 0.11 ------ (24/05)
- 0.01 ------ (25/05) - finally "GOOD"
- 0.08 ------ (26/05)
- 0.07 ------ (27/05) - today :) 

Oy my understanding the "CLS kpi" is useful but, the real KPI to keep monitored after a fix are the % of GOOD, NEED IMPROVEMENTS and BAD...
on my experience the GOOD % improve 1% every two days also if the KPI overall not changed.
We start with a GOOD of 60% and now we have 77% and I'm expecting to see this number growing up (around 85%)...

Here the result:
https://cruxvis.withgoogle.com/#/?view=visstability&url=https%3A%2F%2Fjp.triumph.com%2F&identifier=url&device=DESKTOP&periodStart=27&periodEnd=-1&display=p75s
(we have to work on so may other issue on the site but I'm exited to have a real impact on CWV finally).

[Barry Pollard]

Good job! And that’s why we make the tri-bin histograms available in CrUX.

You can also see these in CrUX Vis:

https://cruxvis.withgoogle.com/#/?view=visstability&url=https%3A%2F%2Fjp.triumph.com%2F&identifier=url&device=DESKTOP&periodStart=27&periodEnd=-1&display=both

Having your own RUM allows you to look at different percentiles (not just 75th percentile that CrUX uses) and timespans (not just the 28-day timespan that CrUX uses) which also helps. Just be aware they can be a lot more variable - which is one of the main reasons why we choose the p75/28-day metrics we use.

To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/chrome-ux-report/5d9b32b5-5db5-45ba-9759-79272862d77cn%40chromium.org.