The 202408 dataset is live

[Barry Pollard‏]

Hi CrUX users,

This is your monthly announcement that the latest dataset has been published to BigQuery.

The 202408 (August 2024) dataset is now available and it covers 18,394,950 origins, an increase of 1.2% over last month. Here's a look at origins' Core Web Vitals performance this month:

• 64.8% of origins (↑ 1.1%) had good LCP

• 78.2% of origins (↑ 0.2%) had good CLS

• 85.0% of origins (↑ 0.8%) had good INP

• 50.1% of origins (↑ 2.0%) had good LCP, CLS and INP

Chrome has recently made a change to INP to no longer include pointerup events used for scrolling (for example when you scroll on mobile using your finger). This not only improved INP by excluding these events, but also resulted in a drop in coverage of INP. Some sites may notice the optional INP metric is no longer available in CrUX.

The Google Publisher Tag ad library released an INP optimization yielding on out-of-viewport Ad insertion logic reducing negative INP impact. Mobile INP pass rates improved from 55.58% to 62.12% for origins using GPT. Publishers can also further optimize ad insertion by using the GPT threadYield config option to yield even in-viewport ads.

These changes—and the continued improvement in LCP—have contributed to the 2% increase in Core Web Vitals pass rates this month. It’s great to see more than half of origins are now passing Core Web Vitals!

As warned for the last few months, First Input Delay (FID) is now deprecated and we have now removed FID from CrUX. The BigQuery dataset with August data published today will be the last month with FID data. The CrUX API will have FID removed from it today.

We have similarly also removed Effective Connection Type (ECT) from CrUX, again as previously announced. The BigQuery dataset with August data published today will be the last month with ECT data. The ability to query the CrUX API by ECT will be removed from today. The course ECT buckets (low-2g, 2g, 3g, 4g) have been replaced in the CrUX API with a more fine-grained RTT latency p75 metric value. Note that this metric measures the internet connection latency of users of your site, rather than latency of your site directly. It is calculated based on recently observed application-layer RTT measurements across recently active connections, not just connections to your site. Removing ECT as a dimension, and replacing it with a non-dimensional metric, means more origins will meet the threshold for CrUX data in the future. This will mean we will be able to bring CrUX insights, including split desktop and mobile data, and navigation breakdowns to many more sites.

Cheers,

Barry