Digital carbon
What this website weighs, how we measured it, and why we publish the number.
Why we measure this
Every page you load moves data across a network and puts work through a device at each end, and all of that uses electricity. That electricity is generated somewhere, and generating it produces carbon emissions. The more data a page moves and the more work it asks of the devices handling it, the more electricity it draws and the more carbon it causes. The weight of a page is therefore a reasonable proxy for its emissions, which is the principle Sustainable Web Design (opens in a new tab) sets out in full if you want the longer version.
Ease of access also comes into this. A lighter page loads much faster on patchy signal when you're on the move or on public wifi. The two goals pull in the same direction, so we set a weight budget and measure against it rather than guessing.
We publish the results here because a figure anyone can check encourages transparency and accountability, and nudges others to do the same with their own sites. The numbers below are generated from our own measurements, re-run whenever anything changes the weight of a page.
How this site was tested
Whenever we make changes to the site, all 20 pages of the built site are loaded in a real browser at 1440x900@1x, each in a fresh context with an empty cache, so what is recorded is a first visit rather than a flattering repeat one. We count the bytes actually transferred, compressed as they are served (dist/, brotli q5), including anything loaded from a third party.
Those bytes are then converted to grams of CO2e using the Sustainable Web Design Model v4 (opens in a new tab) via CO2.js 0.19.0 (opens in a new tab), which is the same model behind the A+ to F digital carbon rating scale. Nothing is estimated by hand and no third-party service is called: the measurement runs locally against the production build, and the page you are reading renders from its output.
Carbon report
- Median page
- 0.030g CO₂e251 kB transferred
- Digital carbon rating
- A+92% lighter than the 2.4 MB average page
- Pages measured
- 2018 rated A+, 2 rated A
- Lightest and heaviest
- 235 kB – 470 kB/404 and /blog/karo-makes-the-most-of-your-data
- Scripts 60%
- Fonts 21%
- HTML 9%
- Images 8%
- CSS 3%
- Other <1%
2.4 MB of 5.5 MB (43%) comes from third parties.
Emissions estimated from the transferred bytes with the Sustainable Web Design Model v4 (opens in a new tab).
Hosting
This site is a set of static files, served by Cloudflare. There is no server rendering a page per request and no database behind it, so a visit costs a file transfer and nothing else.
Cloudflare is listed as running on renewable energy by the Green Web Foundation, whose directory is the usual reference for this. That covers the hosting end of the journey only: the network in between and the device in your hand draw whatever their local grid supplies, which is why the figures above use a global average rather than claiming a clean grid throughout.
Methodology in detail
The exact settings behind the figures above, as recorded by the measurement run on 17 September 2026:
- Model
- Sustainable Web Design Model v4 (opens in a new tab), applied by CO2.js 0.19.0 (opens in a new tab). The model's own page sets out how bytes become grams.
- Grid intensity
- Calculated with model defaults (global average), so we are not assuming a cleaner grid than the average visitor is likely to be on.
- Viewport
- 1440x900@1x. One viewport, so a phone loading fewer or smaller assets is not reflected here.
- Cache state
- Measured cold (first visit, fresh browser context per page). Returning visitors transfer considerably less than these figures show, because fonts, styles and scripts are already on the device.
- What was measured
- The built site (dist/, brotli q5), compressed as it is served in production, not the uncompressed source.
- Rating scale
- The digital carbon ratings (opens in a new tab) published by Sustainable Web Design, from A+ down to F, banded by grams of CO2e per visit.
What these numbers are not
They are modelled, not metered. Nobody can measure the electricity a single page view actually draws across a network and a device they do not own, so every figure of this kind is an estimate built on shared assumptions. The value of publishing it is in the direction and the comparison, not in the third decimal place.