Public audits & demonstrations

How do we measure?

We test each configuration of your site — the original and with our optimizations — 2–3 times for mobile and desktop, and take the median rather than a single run.

How much that median is worth is something we measured rather than assumed, and the answer is uncomfortable enough to print here: rounds of the same measurement, minutes apart on the same URL, disagreed by 9 points at the median on mobile, and 18% of them landed in two different Lighthouse bands. On top of that, PageSpeed can answer a repeat request for the same URL with its previous analysis, which turns some of our repeat rounds into copies rather than measurements. Both effects, with the numbers and what we are doing about them, are in this study. Read a single score here as a band, not as a point.

• PageSpeed Insights or Lighthouse. Depending on the run, we measure through Google's PageSpeed Insights API or a local Lighthouse — it's the same Lighthouse engine either way — under simulated throttling, so mobile numbers are deliberately harsher than desktop.
• Lab, not field. These are controlled lab runs, not real-user (CrUX) field data, so a score here can differ from what Google reports for your site in Search Console.
• Per page. Scores are per-page and illustrate the potential gain — they are not a guarantee for a specific site.

Our audit crawler

To run an audit we fetch the page being tested with a crawler that identifies itself honestly — so you can recognise it in your access logs:

• User-Agent. WebSpeed-Audit/1.0 (bot; +https://webspeed.pro). We do not disguise it as a browser.
• What it does. A small number of GET requests to the single URL submitted — one to check the page responds, one to detect the platform and preview which optimizations would apply. It does not crawl your whole site, submit forms, or touch anything behind a login.
• The score is not measured by our crawler. The before/after numbers are produced by Google PageSpeed Insights / Lighthouse, which fetch the page from their own infrastructure — the same request any speed-test tool makes. A rule aimed only at our crawler will not change the measurement itself.

To keep a site out of our public examples entirely, email us — see below. We only route a site through our proxy with the owner's consent (see Terms).