Skip to content
CarbonlessBeta

Methodology

Every number on this site is an estimate. This page sets out how we produce them, what we assume, and where they fall short, so you can judge the figures for yourself.

How we measure

A headless Chromium browser loads your page from a data center with a desktop viewport, and records every network request through the Chrome DevTools Protocol. We count wire bytes, meaning what actually crossed the network after compression, rather than decompressed sizes. The page is loaded twice: once with an empty cache, then again in the same session. The difference shows how well your caching works in practice, instead of assuming a fixed number.

How bytes become CO₂e

We use the Sustainable Web Design Model v4 through CO2.js, the same model behind Website Carbon and Ecograder. It splits energy across three segments (data centers 22%, networks 24%, user devices 54%), covers both operational and embodied emissions, and applies a world average grid intensity of 494 gCO₂e/kWh. Our headline figure assumes 75% new and 25% returning visitors, using the cache reload ratio we measured on your page. Verified green hosting, according to the Green Web Foundation, removes the operational share of the data center segment.

Ratings

Grades follow the official SWD Digital Carbon Ratings: percentile boundaries derived from the HTTP Archive corpus, running from A+ at 0.040 g or less (the 5th percentile) to F above 0.359 g. Because the scale is shared across the industry, these grades are comparable with other credible tools. The boundaries are anchored to the June 2023 HTTP Archive crawl, the latest the ratings publisher has released, so "cleaner than X%" describes that reference corpus rather than the web as of today.

What this cannot tell you

  • Data transfer is a proxy. A megabyte of JavaScript costs your visitors more CPU than a megabyte of image, the model does not see CPU work, and network equipment draws power largely regardless of traffic.
  • The model is attributional and calibrated to global averages. It likely overestimates for efficient infrastructure and underestimates for inefficient setups, so treat the figures as comparable indicators rather than exact accounting.
  • We measure one page, from one location, on a desktop profile. Your real audience has its own devices, networks and electricity grids.
  • A host that is not registered with the Green Web Foundation shows as unknown, which is not the same as not green.
  • Green hosting comes only from the Green Web Foundation's verification. When a site discloses its upstream infrastructure in a carbon.txt file, we report what it says and check the disclosed providers against the same dataset, but a self-declared file never changes a rating or a gram figure. Where the verification covers a CDN in front of the site rather than the server behind it, the report says so.

Why act on it anyway

The uncertainty sits in the absolute numbers, not the direction. Fewer bytes, fewer requests, effective caching and cleaner hosting reduce energy use under every model. Every fix we suggest comes from something we measured on your page, and each one is mapped to the W3C Web Sustainability Guidelines.