Quite a lot has already been written about the importance of Google Core Web Vitals as a new ranking factor for pages. A special feature of the Web Vitals is that they are calculated from real user data measured on the page. This field data is obtained from the Chrome User Experience Report (CrUX). This is welcome in principle, as it is much more likely to measure the real user experience on a page.
But what happens when this field data suddenly changes significantly due to an unforeseen traffic peak? We were able to observe this case recently on a website we manage.
The website in question offers factual advice pages on a popular special topic. There is a steady (and evenly distributed) stream of traffic from interested parties, all of whom seem to have relatively good internet connections (for Germany) and reasonably up-to-date devices. Accordingly, the Core Web Vitals averages for this website were consistently rated in the "green zone".
The website is set up in such a way that most of the advice pages use a common page template, i.e. they are technically similar. Only distribution pages and news articles are generated using a different template and therefore differ somewhat structurally from the advice pages.
On one of the advice pages, the experiences of a well-known TV celebrity with the special topic are discussed. Because of its catchy name, this celebrity page is usually one of the highest-traffic pages on the site - but otherwise it doesn't really differ from the other advice pages.
In February, the TV celebrity had a major TV appearance one evening, which was apparently also about the special topic covered on the page. Directly during and shortly after the television contribution the access numbers rose rapidly on the prominent side. Within a few minutes, there is additionally about twice as much traffic on the web server as there usually is throughout the day. Practically only the affected celebrity page (found via Google search for the celebrity name) is called up.
The unexpected peak load on the page is absorbed relatively well thanks to Varnish caching, but the server itself is already close to the load limit due to the many simultaneous calls and can only process some of the calls with a slight delay.
Initially, this traffic load had a rather positive effect on the website: the total access figures went up in a presentable manner and a relatively large amount of advertising was delivered and clicked on the page.
With a few days delay, however, surprising negative effects on the Core Web Vital values in the Google Search Console become apparent.
We can assume that many of the peak visitors accessed the site directly from the couch using their cell phones or tablets (second screen). Therefore, the field data for the celebrity site was suddenly collected with many very slow connections and outdated processors. This equipment was strikingly different from the devices and connections otherwise used by the website's regular audience (normal computers with good connections).
Relatively soon after the incident, Search Console reports that there were problems with the Core Web Vitals on this particular page, especially with the time it took to build the "Largest Contentful Paints" (LCP).
These criticized performance problems could not be easily recreated with a normal desktop browser. Only when a "slow 3G" connection in combination with a 4x slower processor was chosen for the performance simulation, the criticized values could also be reproduced locally.
More and more pages with the same template receive the "malus" of bad user data from the single page, even though the TV viewers had called up virtually no other pages on the server apart from the actual celebrity page.
After about 14 days, Core Web Vitals values for all similar advice pages were rated as "improvable" instead of good. However, the news and mailing list pages on the server with other page templates remained in the green.
Since there were no more reports about the prominent performance killer in the next few days, the values also began to recover quickly, however, thanks to the regular audience with their better devices. Nothing drastic was changed in the basic page loading or server settings during the affected period.
Fortunately, the incident did not really have a negative impact on rankings or traffic of the website. At this point, Core Web Vitals are also not yet official ranking criteria.
- Traffic peaks on individual pages can apparently have an unexpectedly strong effect - with some time lag - on the overall ranking of Core Web Vitals for an entire website (at least for all similarly constructed pages).
- The effect dissipates relatively quickly as field data sources return to normal.
- Even if the Core Web Vital values look good in Search Console, you should test the lab values by simulating poor Internet connections and devices. It is not said that the user composition will always remain the same.
Please feel free to share this article.