I know how many elements, bytes and headings our site has, because I have counted all of those. I did not know how tall it is. Page height is the one dimension a visitor experiences directly: it is the scrollbar they see before they read a word, and it is the thing that decides whether your closing argument is on screen or four flicks of a thumb away.

So I measured every page in the sitemap at two widths, in a real browser, with the simplest instrument there is:

document.documentElement.scrollHeight // the whole document window.innerHeight // one screen of it 

78 of the 79 URLs, twice: once at 1280x900 and once at 390x844, which is a phone. The homepage is excluded because the browser profile I measure with is signed in and bounces to the dashboard, so it never appears below.

The whole site, end to end

1280x900 390x844
shortest page 1,130 px 1,522 px
median page 3,812 px 5,983 px
tallest page 9,167 px 12,742 px
median page in screens 4.24 7.09
all 78 pages stacked 304,981 px 456,550 px
that, in screens 339 541
pages that fit in one screen 0 0
pages over five screens 24 59

Narrowing the viewport from 1280 to 390 makes the median page 1.55 times taller, and the per-page ratio runs from 1.29 to 1.68. That is the whole cost of a single-column reflow, measured: nothing is hidden, nothing is added, the same text simply stands up.

I checked that "nothing is hidden" rather than assuming it. Total characters of rendered text across the 78 pages: 378,216 at 1280, and 378,216 at 390. Identical. No display: none content branch, no phone-only summary, no desktop-only table. One document, two shapes.

The 304,981 pixel figure has a daft but useful translation. CSS defines its pixel as 1/96 of an inch, so at that nominal density the desktop site is about 81 metres of paper and the phone site is 121 metres. A phone's pixels are physically much denser than that, so the real ribbon is shorter, but the number is a fair measure of how much reading is being asked for.

What a visitor gets before scrolling

For each page I walked every text node, took its bounding rectangle with a Range, and counted the characters whose box starts inside the first screen.

above the fold 1280x900 390x844
characters, median page 1,319 854
share of the page's text 26.6% 17.5%
best page 88.9% 60.6%
worst page 10.2% 6.9%

A quarter of the argument on a desktop, a sixth on a phone. The best-served page is the cookie policy, 88.9 per cent above the fold, because it is 1,572 characters long in total and almost all of it fits. The worst is the review of the year at 10.2 per cent, which is the tallest page on the site and exactly what a long question list should look like.

Where the furniture sits is identical on every page of a given type, which is what a generated layout gets you:

marketing pages the three legal pages
header height 57 px 64 px
first link inside main 120 px (the breadcrumb) 354 to 689 px
h1 top 164 px 112 px
main share of document height 88.4% (median) 94 to 96%

Two header heights, two h1 positions, 78 pages. The h1 is above the fold on every single page at both widths, which is the one thing in this whole audit I would have bet on and won.

The 1,211 pixel footer

Then the number that made me open the inspector. The footer is one component, rendered at one of two sizes:

footer 1280x900 390x844
75 marketing pages 449 px 1,211 px
3 legal pages (compact footer) 65 px 93 px

On a phone the full footer is 1.43 screens tall on its own. As a share of the document it runs like this:

page footer share, desktop footer share, phone
/features/multi-venue 18.9% 32.4%
/tools 18.5% 32.2%
/solutions/corporate-events 18.6% 31.4%
median page 11.7% 20.2%
review of the year 4.9% 5.1%
/legal/privacy 2.2% 2.3%

So a reader on a phone who reaches the end of our shortest feature page has spent a third of that page's height on a navigation block, and the pages where that hurts most are precisely the short, focused ones a search result lands on. The three legal pages prove the alternative exists in our own codebase: they get the compact footer, and it is 93 pixels.

I am not going to delete the footer. Every one of our pages is two clicks from the homepage, which I have measured separately, and the footer is a large part of why. But "the footer is one shared component" and "the footer is 32 per cent of the page a phone visitor scrolls" are different facts, and only one of them shows up in a bundle report.

The hub page with no heading for 3,157 pixels

One outlier worth its own line. The first h2 inside main, measured from the top of the document:

page first h2
median page 351 px
/pricing 1,340 px
/solutions 3,157 px

On the venue hub you scroll three and a half screens, past eleven venue cards, before the document offers you a second-level heading. The cards are links and they are the point of the page, but a reader skimming by heading gets nothing to skim until they are most of the way down. That is a structure problem the HTML validator, the accessibility audit and the Lighthouse run all pass without comment, and it only shows up when you measure in pixels instead of in tags.

Check it yourself

Any page, any browser, no tooling:

const h = document.documentElement.scrollHeight const f = document.querySelector('footer').getBoundingClientRect().height ;({ screens: +(h / innerHeight).toFixed(2), footerShare: (100 * f / h).toFixed(1) + '%' }) 

Run it on the multi-venue page, then open the device toolbar, set 390x844 and run it again. The footer share roughly doubles, and nothing about the page changed except its width.

Two things I am doing with this. The compact footer that the legal pages already use goes onto the short feature and solutions pages, which is a one-line change to a prop and takes a third of a phone screen back. And /solutions gets a heading before its card grid, because a page where the first heading is 3,157 pixels down is not really a hub, it is a long introduction with links at the end.