What the homepage of France’s 20 largest cities weighs
Perennova barometer · edition 1 · measured on 2 October 2026. One method for all: we download the page and everything its HTML declares, then count bytes, requests and third-party hosts. No carbon estimate. Every figure links to its dated report. Repeated every month, so the trajectory can be read.
What the measurement shows
- 17 pages measured out of 20; three sites refuse the measuring server (see below).
- Median weight: 3,445 KB. The mean is 3,935 KB.
- 12 pages out of 17 exceed 2 MB (2,048 KB). Only two stay under 1 MB: Le Havre (884 KB) and Grenoble (949 KB).
- A 1-to-14 gap between the lightest (884 KB) and the heaviest (at least 12,584 KB, measurement cap reached).
- 29% of all measured weight comes from lazy-loaded images (
loading="lazy"). A browser only loads them on scroll: a first-screen-only measurement does not see them. - Median Perennova scale: 6 out of 20. These are our own published thresholds, not an official standard.
The table
* Paris: the 12 MB measurement cap is reached after 32 of 41 sub-resources; the real weight is higher.
Three sites not measured
- Rennes (metropole.rennes.fr): the site answers 403 to the measuring server.
- Reims (www.reims.fr): the site serves the measuring server a 222-byte holding page (anti-bot check), not the homepage.
- Saint-Denis (La Réunion) (www.saintdenis.re): the site answers 403 to the measuring server.
Refusing a declared bot is a legitimate operating choice. It is neither a fault nor a score: these three cities are simply not ranked.
Follow the barometer
Your address is only used to send you the barometer, once a month. It is deleted on request to perennova@proton.me.
Do you run another website? Measure your page: same method, dated report, free, no account.
Why these twenty cities
They are the twenty most populous communes of France (Insee, municipal populations as of 1 January 2023, list taken from Wikipedia). All exceed 50,000 inhabitants: the French law of 15 November 2021 (known as REEN) requires them to define a responsible digital strategy by 1 January 2025 (Maire-info, 5 September 2022, in French).
The weight of a homepage is not that strategy. It is a measurable, dated fact that can be compared from one month to the next.
Method, and what it does not say
- The measuring server downloads the HTML, then the scripts, stylesheets, images (including lazy-loaded ones) and
url()files that this HTML declares. Caps: 200 sub-resources, 12 MB, 60 seconds. - It does not run JavaScript. Whatever a script loads afterwards (carousels, maps, videos, trackers) is not counted. In a browser the page therefore weighs at least this much, often more.
- The weight shown is for the whole page, not the first screen: lazy-loaded images are counted, and their share is shown.
- It measures neither CPU, memory nor battery on the device, and does not judge design intent. It produces no carbon estimate.
- One measurement per site, on a single day, with no cache or cookies, from a server that identifies itself as
PerennovaScan. A homepage changes: the dated report holds for its date, not beyond. - “Third-party host”: any hostname different from the page’s, including a subdomain of the same organisation.
- This barometer is not a compliance audit against the French RGESN guidelines.
Prior work. Greenspector published a ranking of 46 French city and metropolis websites in 2020 and again in 2022, measured on a real device (energy, data, CPU, memory): their study is richer than this table. This one measures less, but it is repeated every month and every figure carries its dated evidence.
Right of reply. You run one of these sites and dispute a measurement? Write to perennova@proton.me: the measurement is redone and the correction published here, dated.
Measurements and page produced with an AI agent, reviewed and published by Perennova. Engine: perennova-scan v4.