A scoring grid says what is required; it does not say whether it is delivered. This tool measures what a public page actually costs to load: the bytes actually downloaded, the number of requests and the third-party dependencies — scripts, fonts, images served from other domains. It then fills in a score out of 20, where every point is justified by the measured fact. Free, no account.
The measurement is language-independent; the procurement criterion generator on this site is written for French public procurement.
The measurement is made by a server, not by your browser: the measured address and the result are recorded server-side to track how the tool is used. Nothing is stored in your browser — no cookie, no local storage. The site counts visits with a cookie-free tool that collects no personal data (GoatCounter).
Example: the monthly barometer of France’s 20 largest cities.
| What is scored | Points | The measured fact |
|---|
The measured facts are returned by the measuring service in French and are shown verbatim (“Ko” = kilobytes, “requetes” = requests, “hote(s) tiers” = third-party host(s)).
| Resource | Weight |
|---|
What this measurement does not tell you
Since 2 October 2026 (engine v4): lazy-loaded images declared in the HTML are counted, and their share is shown in the report.
Quoted verbatim from the measuring service, in French: the cost on the visitor's device, the JavaScript executed after loading, and any carbon estimate are not measured.
The thresholds in this score are Perennova's. It is not an official standard.
What next? A measurement tells you where a page stands; it does not tell you what to require from a supplier. To write the criterion and the scoring grid for your tender (French public procurement), the generator on the home page produces both from six answers. To be told when the scoring grid or the thresholds change, leave an address.