You check your home page with GreenTracer and get one figure. A colleague runs the same page through another website carbon calculator and gets a different one, perhaps with a different letter grade beside it. It's tempting to assume one of the tools is wrong. Usually neither is. They are answering slightly different questions, with different published assumptions, and once you know where to look the gap is easy to explain.
Almost every calculator follows the same basic recipe: measure how much data the page sends, turn that into energy, then turn energy into carbon using a figure for the electricity grid. Each step involves a choice. This article walks through the five that matter most.
Which version of the model
Most tools use the Sustainable Web Design Model (SWDM), but not necessarily the same version. The current version 4 page says the authors “updated most data sources” and now separate operational and embodied energy. Its energy figures add up to 0.300 kWh per gigabyte across data centres, networks and user devices. The archived version 3 page used a single figure of 0.81 kWh per gigabyte and a grid intensity of 442 g CO₂e per kWh.
The difference is large. In its CO2.js v0.16 release guide, the Green Web Foundation says version 4 estimates are “on average two-thirds lower” than comparable version 3 estimates, and recommends re-baselining older figures rather than mixing the two. A tool still on version 3 will usually show a much higher number for the same page.
Some tools don't use the SWDM at all. The Green Web Foundation's CO2.js documentation describes a second model, OneByte, which counts data centres and networks only. It notes that the two models “return slightly different results” because their system boundaries differ. CO2.js uses SWDM v4 by default, but version 3 can still be selected. GreenTracer uses SWDM v4, as CO2.js does.
What was actually measured
Before any model runs, each tool has to load your page and count the bytes. Three choices change that count.
- First visit or an average visit. The SWDM includes a return visitor ratio and a data cache ratio, because a returning visitor's browser may already hold some files. CO2.js offers a perVisit function that includes the model authors' assumptions about visitors and caching; the version 3 page assumed 25% of visitors return, loading 2% of the data. A tool that applies a caching discount will show less than one that doesn't. GreenTracer measures one initial, uncached load and applies no discount.
- Which device. A page can send different images to phones and desktops. GreenTracer runs Google PageSpeed Insights on mobile and desktop and uses the heavier run. Another tool may test only one. See why desktop and mobile results differ.
- What counts as loaded. GreenTracer's test doesn't scroll, click or accept cookie banners, so lazy images and consented tags aren't counted. A tool that scrolls the page, or waits longer, may count more.
Even the same tool can see a slightly different page from one day to the next; that's covered in why two checks of the same page can differ.
Grid intensity and green hosting
Energy becomes carbon through a grid intensity figure. The SWDM v4 default is the global average of 494 g CO₂e per kWh, but the model says users “may choose to replace the default grid intensity with regional, country, or local values”. A tool that uses your country's grid, or your visitors', will differ from one that uses the global average. GreenTracer uses the global average because a page test can't know where visitors are.
Hosting is the other lever. Version 4 has a green hosting factor between 0 and 1. On a host the Green Web Foundation lists as green, GreenTracer removes the data-centre operational share, which takes the estimate from about 0.148 g to about 0.121 g CO₂e per MB, about 18% lower. If one tool recognises your host as green and the other doesn't, or treats it differently, the figures will part company. Our article on what green hosting changes goes into detail.
Different grading scales
Letter grades add another layer. The Sustainable Web Design digital carbon ratings are labelled for SWDM v4 and run from A+ at 0.040 g per page view or less to F above 0.359 g. The page says the scale is designed to give a “like for like” comparison between web pages. GreenTracer uses these boundaries; you can see them all on the grades page.
A tool may use its own scale instead, or grade on an older model. Then a C in one tool and a B in another may describe almost the same grams. Compare the grams, not the letters.
How to compare fairly
If you want a number you can track, or share with a client, the rule is to change one thing at a time:
- Use the same tool for every check you intend to compare, including before-and-after checks.
- Use the same page address, exactly. The home page with and without a campaign link or a trailing path may not be the same page.
- Check the model and version. Look for “SWDM v4” or equivalent in the tool's method notes. If one says v3, expect a higher figure.
- Compare the bytes first. If two tools report very different page weights, the gap is in the measurement, not the model. Lighthouse's total byte weight is shown in KiB, so convert before comparing; our page weight guide explains the units.
- Write the method down next to the figure: tool, date, model version, whether the host was treated as green.
None of these tools is metering your page; every figure is an estimate, for the reasons in why website carbon is an estimate. What makes an estimate useful is consistency. Pick one method, stick with it, and check again after each change.
Sources
- Sustainable Web Design: estimating digital emissions (SWDM v4)
- Sustainable Web Design: estimating digital emissions, version 3 (archived)
- Sustainable Web Design: digital carbon ratings
- Green Web Foundation: methodologies for calculating website carbon (CO2.js)
- Green Web Foundation: CO2.js perVisit function
- Green Web Foundation: CO2.js v0.16 release guide (SWDM v4)
- Chrome for Developers: Avoid enormous network payloads (Lighthouse)
