Methodology

HOW WE MEASURE — AND WHAT THE NUMBER IS WORTH

Every carbon figure on this site is an estimate produced by a published model. This page sets out exactly what we record, how we convert it, and — just as importantly — where the number stops being reliable.

1. What we measure

We load your page in a headless Chromium browser on our own infrastructure — never from your visitors' devices, and with no impact on your live site. The browser starts from a clean profile with an empty cache, so nothing is pre-loaded.

Using the Chrome DevTools Protocol, we record the compressed size of every resource that finishes downloading — HTML, CSS, JavaScript, images, fonts, video, third-party tags — and total them. This is bytes actually sent over the wire, not uncompressed file size.

Recording stops shortly after the page load event, once the network has been quiet for a moment. Each scan is capped so a single slow site cannot stall the queue.

2. How we convert bytes into CO₂

We pass the byte total to the Sustainable Web Design Model (SWD), via the open-source @tgwf/co2 library maintained by the Green Web Foundation. We use its per-visit calculation, which spreads the estimated energy across datacentre, network transmission and end-user device, applies an average grid carbon intensity, and builds in the model's own assumption about the mix of new and returning visitors.

Before converting, we look your host up in the Green Web Foundation's public directory of providers running on renewable energy. If it is listed, we apply the model's green-hosting adjustment, which lowers the datacentre portion of the estimate. If it is not listed — or the lookup fails — we use the standard, higher figure rather than assume in your favour.

The output is grams of CO₂ equivalent per visit. We store that figure alongside the raw byte count, the per-resource breakdown, the hosting result and the timestamp, so any scan can be traced back to the measurement it came from.

3. The A–D rating scale

Ratings are fixed thresholds applied to the per-visit figure. They are our own bands, not an external standard or certification.

≤ 0.20 g

≤ 0.20 gLighter than most of the web

0.20 – 0.50 g

0.20 – 0.50 gAround or below average

0.50 – 1.00 g

0.50 – 1.00 gHeavier than average

> 1.00 g

> 1.00 gSignificantly heavier than average

4. Known limitations

These are the places our number is weakest. We would rather you know them up front than discover them in front of a client.

It is a model, not a measurement

No energy meter is involved anywhere in this process. We measure bytes and apply a published conversion model. The output is an estimate with a wide uncertainty range, not a metered reading of electricity consumed.

We assume global average grid intensity

We do not currently detect where your site is hosted or what powers that datacentre. A site running entirely on renewable energy receives the same estimate as an identical site running on coal. This is the single largest source of error for well-hosted sites.

Green hosting relies on a third-party directory

We check your host against the Green Web Foundation's public directory and apply their green-hosting adjustment when it returns a match. That directory is not exhaustive: a host genuinely running on renewable energy but not registered there will be scored as if it were not, so the figure is conservative rather than wrong. If the lookup fails, we also fall back to the conservative figure.

One page load, cold cache

Every scan starts from a fresh browser with an empty cache. Real visitors arrive with assets already cached, browse several pages per session, and use a wide mix of devices and connections. The model applies its own blended new/returning visitor assumption, but it cannot know your actual traffic pattern.

Bytes only

Server-side work that does not produce network traffic is invisible to us — database load, background jobs, image processing, model inference. Two sites transferring identical bytes can have very different backend energy costs.

Embodied carbon is excluded

Manufacturing the phones, laptops, servers, routers and cables involved carries a substantial carbon cost. The model covers operational energy only, so the true lifecycle footprint of any web page is higher than the figure we report.

Public pages only

We scan unauthenticated pages as an anonymous visitor. Anything behind a login, paywall or geo-restriction is not measured.

Late-loading third parties may be missed

We record traffic until the page load event plus a short idle period. Scripts that fire later — on scroll, on consent, on interaction — fall outside the measurement window and are not counted.

Numbers are not comparable across tools

Different carbon calculators use different models, assumptions and visitor profiles. Our figure is comparable to our own figures, measured the same way over time. Comparing it against a number from another tool is not meaningful.

5. What this number is — and isn't — for

Suitable for

— Tracking whether a specific page is getting heavier or lighter over time

— Comparing pages and sites that we measured the same way

— Prioritising which pages to optimise first

— Evidencing the effect of optimisation work, before and after

— Communicating progress in voluntary sustainability updates, with the limits stated

Not suitable for

— Regulatory filings of any kind

— CSRD / ESRS Scope 3 greenhouse gas inventory disclosure

— Audit-grade or assurance-grade carbon accounting

— Carbon offset purchasing or retirement calculations

— Substantiating a “carbon neutral” or “net zero” claim

To be explicit: Ecossio is a measurement and reporting tool. We are not an assurance provider, an auditor, or a certification body, and our output is not a substitute for greenhouse gas accounting performed under the GHG Protocol or an equivalent framework. If you are preparing a regulated disclosure, treat our figure as one supporting input for your sustainability adviser to assess — not as the disclosure itself. Nothing on this page is legal or compliance advice.

6. Changes to this methodology

If we change how we measure or convert — a different model version, a new grid intensity source, adjusted rating bands — we will record the change and its date here. Historical scans keep the basis they were measured on, so a trend line never silently changes shape underneath you.

25 September 2026 — green hosting adjustment added

We began checking hosts against the Green Web Foundation directory and applying the model's green-hosting adjustment. Sites on a listed green host will show a one-off drop in their reported figure from this date — the site did not change, the basis of the estimate did. Scans recorded before this date were calculated without the adjustment and have been left as they were.

Last reviewed: September 2026.

Questions about any of this

If you need detail we haven't covered — the exact model version, how a specific figure was produced, or the raw scan data behind a report — write to support@ecossio.com and we will send it. You can also read our guide to reporting website carbon.