How this is calculated

Google's Chrome UX Report records how long real visitors actually waited for a page to render its main content, and reports the share of those visitors in three bands: under 2.5 seconds, 2.5 to 4 seconds, and over 4 seconds. These are measurements of real people on a specific page over the previous 28 days.

This is the part that makes the estimate defensible. It is not derived from the Lighthouse score, and a company cannot dispute it without disputing Chrome's measurements of its own site.

Google found that 53% of mobile visits are abandoned when a page takes longer than three seconds to load. That figure comes from Google's “The Need for Mobile Speed”.

We multiply the share of a page's real visitors who got a slow load by the rate at which Google says such visitors leave. The result is an estimated share of visits lost to load time.

Chrome's bands split at 2.5 and 4 seconds, but Google's finding is anchored at three seconds, so the middle band straddles the threshold. We count all of the over-4s group and half of the 2.5-to-4s group as being past three seconds. That halving is an approximation, and it is the single largest source of imprecision here. It is why every figure on this site says estimated.

It is not a measurement of any company's actual traffic. We have no access to anyone's analytics, and we never publish revenue estimates, because inventing numbers about someone else's business would be dishonest.

It is an estimate built by applying an aggregate finding to one page's measured distribution. Treat it as an indication of scale, not a figure to put in a board report.

Lighthouse is a lab simulation run on throttled hardware. It is useful, every company here has its four scores published, and you can verify any of them yourself on PageSpeed Insights. But it describes a simulated visit, not real ones.

The two disagree, and that is the point. Ahrefs scores 30 on Lighthouse performance, one of the worst on the board, yet only about 5% of their real visitors get a slow load, so they sit near the bottom. Ranking on the lab number would have contradicted the headline figure on most rows.

Chrome only reports field data for pages with enough traffic to anonymise. Where there is no sample we cannot say anything about a company's readers, so that blog is unranked and shows a Lighthouse score only.

We do not fall back to lab timings for these. An early run of this board produced a simulated 41 second load for one such page. Publishing that would have been indefensible.

The starting list is large, well known companies. The rule is that we only add companies big enough that appearing here is funny rather than damaging. Smaller sites appear only when somebody submits them.

Any listed blog is removed on request, no questions asked. This is a point about page speed, not a pressure campaign.

This board was built by the founder of Superblog, a blogging platform, and Superblog is on it, scored on the same scale as everyone else. Our own blog scored 52 on mobile in July 2026 before we fixed it, and we publish that starting number too.

A scoreboard run by a vendor is worth nothing if the vendor exempts itself, so we did not. We are not first, and nothing about a company's position here is influenced by whether they are a customer.

Fix your blog and press recheck. There is no payment option and there never will be, which is the entire point. Rechecks are limited to one per blog every ten minutes because results do not move faster than that.

Note that Chrome field data updates on a 28-day rolling window, so a real fix takes weeks to move the ranking even though the Lighthouse score responds immediately.

• Google Chrome UX Report, via the PageSpeed Insights API
• Google, “The Need for Mobile Speed”
• Google and SOASTA, “Mobile page speed new industry benchmarks” (2017)