Why it matters
- A site's average hides the differences between countries. One country's network and phones differ from another's, and a site that is fast in Riyadh may be slow in Cairo or Rabat.
- Google measures the page by its real visitors, so a site that does not pass at most of its visitors does not pass in the page experience assessment.
- The usual CrUX API has no country dimension; the difference shows only in the per-country BigQuery tables.
How to fix
- Start with the worst country: what are its network and devices? Test the page at a network speed and on a phone like theirs.
- Bring your server closer to your visitors: a CDN with points of presence in the region.
- Cut the size of images, fonts and scripts and defer what is not above the fold (see the page speed and image weight checks).
- Fix the dimensions of images and ads to avoid layout shift (CLS).
- Reduce heavy work on the main thread to improve responsiveness (INP).
How we detect
- We query your site's origin (the domain with its protocol) in the Chrome UX Report tables in BigQuery for nine Arab countries: Saudi Arabia, the UAE, Egypt, Kuwait, Qatar, Bahrain, Oman, Jordan and Morocco (
chrome-ux-report.country_<cc>.<yyyymm>), for the latest month whose tables are published, phones only. - For each country we compute the share of good visits: LCP up to 2500 ms, INP up to 200 ms, CLS up to 0.1, from the tables' histograms; and we ask the origin's popularity rank in each country with a second, smaller query.
- We run the query as a dry run first at no cost, and run it only if it reads less than the cap, and BigQuery refuses to bill more than the cap (
maximumBytesBilled). We keep the result for a day. - The tool runs when the operator has put in a service account and a project to pay for it; without them it stops with a notice and sends nothing.
- A country shows in the result only if the origin has enough data in it. It is a minor finding.