What the numbers mean

How much data a page on this website usually makes you download. For each page we measure, we take three parts first visit and one part returning, as most people arrive fresh and some come back. We take the middle page on a phone and on a laptop, and average the two.

The first page you open on a website you have never visited before.

A page reached by a click from another page on the same website, when some of its files are already saved.

The lightest and heaviest of our visits: up to three on each device, or two when they agree within 5%. Some websites load different things on different visits; where they do, we show what.

The homepage and four pages of the same kind for every organisation in the industry, so like is compared with like (for a bank: mortgages, credit cards, everyday accounts and contact). Where a website has no page for a task, we measure the pages it has. We never open sign-in, cart or legal pages.

A letter from A+ (Boomtown) to H (Monster), for a typical page view. A+ to F are the same letters as the Website Carbon Calculator; G and H are ours, for very heavy pages. See how the grades work.

Across an industry, the average is the median: half the websites are lighter, half heavier.

Where the website sits among the organisations in its industry measured the same month, from the lightest to the heaviest.

The homepage’s first visit, split by who supplies each file: the organisation itself, or other companies, such as chat, analytics and advertising. Unknown source is weight we could not trace to a named company.

The same page split by kind of file: images, scripts, fonts, styles and other. Video is weighed on its own and never counted in the weight.

The five largest single files on the pages we measured, on a first or returning visit, and whose they are.

Ways the website could be lighter, each measured by making the change to a copy of a file we received. There are two kinds:

• Fix: nothing a visitor would notice, such as a smaller copy of an image for phones.
• Choice: removes or delays something, such as loading a chat window only when someone opens it.

We show the five biggest and say how many we found. Anything under 50 KB is not listed. Every saving is an estimate.

Both count equally. Websites often send laptops heavier images, so we measure both and show each.

An estimate of the carbon from 100,000 typical page views: the electricity used by data centres, networks and visitors’ devices, and a share of making them. We use the Sustainable Web Design model, with New Zealand’s grid for electricity used here and the world average for data centres. It is a model, so the true figure could be several times higher or lower. It goes down as the data goes down.

Figures from Google’s Chrome UX Report: how real Chrome users in New Zealand reach the website (phone, computer, tablet) and how quickly its main content appears for most visits. See Chrome’s real-visitor figures.

The dates we measured that website.

Our bot obeys every website’s robots.txt and every block. A few websites are measured manually by one of our team, by the same method, and the profile says “Measured manually by one of our team”.