CrUX: How much data is sufficient?

Question

URLs without much traffic, if tested with PageSpeed Insights tool have no own CrUX metrics - the test displays instead the metrics from the origin (domain). About the tested URL the tool mentions:

There is insufficient real-user data for this URL.

CrUX mentions

A page is determined to be sufficiently popular if it has a minimum number of visitors. ... An exact number is not disclosed.

I would like to know: how much data is suficient? Does somebody have experience values or estimations?

Edit:

I try to describe a background of my question:

In general, I'm looking for criteria to baste an index bloat. There are two metrics describing URLs which are counted as good/useful:

• An amount of URLs in search results, not according to the site-query, but rather to the amount displayed by Search Console,
• An amount of URLs with Top100-rankings, displayed by other tools, not always correlating with GSC.

Both metrics not always correlate, but knowing both helps to get an average.

In the Core Web Vitals report delivered by GSC there is, for mobile and desktop, the same amount of URLs (a sum of good, bad and needing improvement).

And this is my idea: belong URLs not listed in the GSC Core Web Vitals report to index bloat?

Answer 1

(Disclosure: I helped write those CrUX docs)

You can look at your analytics data to get a sense for how many unique visitors you typically need to be included in the CrUX dataset, but there are a few reasons why this won't always be accurate:

• Only Chrome users on supported platforms are included in CrUX. You'd need to filter your analytics to only these platforms to be comparable.
• Users must meet other eligibility criteria, like enabling usage statistics and syncing browser history. These settings are internal to Chrome, so site owners don't have a way to determine whether a user meets these criteria.
• User populations vary across sites, so imperceptible differences in user eligibility will limit how comparable sites' estimated popularity thresholds are to each other.
• The CrUX data pipeline does some complex data quality processing. Specifically, fuzzing and filtering can make it harder to reverse engineer the popularity threshold.

So in short, it's complicated. It's a byproduct of both the complexity of the CrUX dataset as well as deliberate design choices to protect sensitive data.

Answer 2

DISCLAIMER: Some time ago I read about the sample size needs to be >1000 in the last 30 days.

Like Rick said, only supported platforms and users who opted in are counted. So the number of visitors in the last 30 days needs to be higher.

Also the metrics are per-URL with fallback to the start-page. For example if you want to measure myblog.com/article123 and that URL has only a sample size of 600 and myblog.com has 2000, then the field-data of the start-page is shown.

Answer 3

I found an interesting article here: https://www.erwinhofman.com/core-web-vitals-faq/data/#how-many-visitors

Nobody knows. Tests I conducted in the past pointed out that your site needs around 400 pageviews a month per device (mobile, tablet, desktop), using Chromium to get a Core Web Vitals origin summary when using PageSpeed Insights. Do note that this might be the absolute minimum.