Introducing EcoPigs v3 and the Measured-Carbon Model (EMCM)

For over a decade, the entire website-carbon industry has measured roughly the same thing: how many bytes a page weighs, multiplied by a yearly average for how dirty the electricity grid is, printed as a single number with no margin of error. It is a reasonable proxy. It is not a measurement — and as carbon reporting becomes mandatory under CDP, SECR, B Corp and CSRD, the gap between "a reasonable proxy" and "a number you can defend in a disclosure" is the whole game.

Today we are publishing EcoPigs v3, built on the EcoPigs Measured-Carbon Model (EMCM) — our open, published methodology for measuring, not estimating, the carbon of a website.

What actually changed

EMCM does four things no estimate-only tool does.

It measures device energy, not bytes. Page weight cannot tell a 1 MB photo apart from a 1 MB piece of heavy, battery-draining JavaScript — but they have very different energy profiles. Where the EcoPigs badge is installed, we read the real work a visitor's processor does (via the Chrome DevTools Protocol) instead of guessing from file size.

It uses the live grid, not last year's average. Electricity is far cleaner at 3am on wind than at 6pm on gas. EMCM weights emissions by the carbon intensity of the grid at the time and place each real visit happened.

It calibrates against a real dataset. A new site with little data is grounded against the patterns we have measured across hundreds of similar sites; as it gathers its own real visits, the figure shifts to rely on its data. Thin-data sites get an honestly wide range, not a fabricated precise one.

It tells the truth about uncertainty. Every figure is a median with a 90% confidence range, and every part is labelled by how we know it: measured, modelled, or estimated. We never show a bare number as if it were exact.

The ladder: estimate to measurement

This is the part that matters commercially. A free EcoPigs scan gives you an estimate — a cold first visit, with the live grid and full lifecycle, but modelled device energy. Installing the badge upgrades that to a measurement of your real visitors' actual footprint.

Here is the proof, from our own site: a cold scan estimated 0.029 g of CO₂e per visit. Measured across real visitors, the true figure was 0.012 g — about 2.5× lower, because real people arrive with files already cached and a cold scan cannot see that. The letter grade barely moved; the reportable number moved a lot. That difference is exactly why a site putting a figure into a regulated disclosure needs the badge, not just the scan.

(That comparison is always a site against itself — its own scan versus its own real visits. We never imply the gap between two different sites is "what the badge did.")

The honesty doctrine

EcoPigs' whole position is honesty, so EMCM bakes in rules designed to survive a sceptic: no invented figures ("unavailable" beats a made-up number); ranges, not false precision; every number labelled by evidence; reproducible on demand, down to the inputs, coefficients and sources; and openness about our own weak spots, including where a coefficient is a conservative estimate rather than a measured value.

The clearest example is embodied carbon — the manufacturing footprint of the devices and infrastructure involved. It is typically the larger share of the total, and in v3.0 it is still modelled, not measured. So its uncertainty dominates the overall range, and we show that honestly rather than letting the measured operational part make the whole number look more precise than it is. Making embodied carbon measured too — from real device active-time — is the headline of our next release.

Deep, not broad — and we say so

EMCM is backed by real measurement: around 1,130 graded sites and 16,000+ measurements over 10 continuous months. But we are precise about what that is. It is a deep dataset — continuous measurement over many months — not yet a broad survey of the whole web; it leans toward the UK and certain platforms. So we publish findings scoped to what the data supports, always with the sample size shown, and we say "deepest," never "largest" or "the web."

Read the methodology

EMCM is published openly — that is what earns citation. For a plain-language overview, read Measuring the Real Carbon of the Web. For the full technical paper — the empirical-Bayes calibration, the novelty matrix, and every coefficient with its citation — read The EcoPigs Measured-Carbon Model, or see how any single score is calculated on the methodology page.

The shift in one line: from estimating a website's carbon to measuring it — and being honest about the difference.