Page width gets measured all the time. I had never measured Nakodo's pages the other way, so I took every URL in the sitemap, overrode the device metrics with CDP, and read document.documentElement.scrollHeight at a phone width and a laptop width.
await page.cdp("Emulation.setDeviceMetricsOverride", { width: 390, height: 844, deviceScaleFactor: 1, mobile: true }); await page.goto(url); const r = await page.evaluate(() => ({ scrollHeight: document.documentElement.scrollHeight, header: document.querySelector("header").getBoundingClientRect().height, footer: document.querySelector("footer").getBoundingClientRect().height, h1: document.querySelector("h1").getBoundingClientRect().top + window.scrollY, })); 76 pages, twice, no errors.
The totals
| At 390x844 (a phone) | |
|---|---|
| Total height of the site | 663,300px |
| That in screens | 786 |
| Median page | 9,104px, 10.8 screens |
| Tallest | /how-it-works, 14,012px, 16.6 screens |
| Shortest | /alternatives, 3,202px, 3.8 screens |
| At 1280x800 (a laptop) | |
|---|---|
| Total height of the site | 423,716px |
| Median page | 5,771px, 7.2 screens |
| Tallest | /how-it-works, 8,834px |
| Shortest | /blog, 1,822px |
The median page is 1.57 times taller on the phone than on the laptop. The extremes of that ratio are interesting: the two template-heavy guides are the least affected at 1.45, because a block of monospaced message text wraps the same way at both widths, and /youtubers is the most affected at exactly 2.0, because its card grid goes from three columns to one.
Per word, the median page spends 6.15px of height on every word at 390px and 3.91px at 1280px. That is a decent rule of thumb for anyone estimating how long a page will feel before building it: on a phone, a 1,500-word page is about ten screens.
The footer is a constant
header height, all 76 pages, both widths: 65px footer height, all 76 pages, at 390px: 1,233px footer height, all 76 pages, at 1280px: 501px elements inside the footer: 515 One value. Not a range, not a median: every page on this site carries exactly the same 1,233px of footer on a phone, because the footer is one component with a landscape drawing in it, and nothing in it is conditional.
That makes the arithmetic easy and a bit uncomfortable. 1,233 x 76 = 93,708px of footer, which is 14.1% of the whole site's height, and 111 of those 786 phone screens. The median page spends 13.5% of its scroll on it.
The distribution is the problem rather than the number:
footer as a share of page height, at 390px /alternatives 38.5% /tiktokers 37.6% /instagrammers 37.3% /blog 37.2% ... /glossary 10.9% /guides/youtube-outreach-email-templates 10.9% /how-it-works 8.8% The four worst are all index pages, and index pages are exactly the pages where a reader has not yet decided to stay. /alternatives is 3,202px tall, of which 65px is header and 1,233px is footer, leaving 1,904px of actual page: the furniture is 68% the size of the content it frames. An earlier audit of this site found that the same index pages were the most duplicated by text as well, with two thirds of their words appearing elsewhere. Two different measurements, same six pages, same cause: those pages have almost nothing of their own on them.
Where the important things sit
Every one of the 76 H1s is above the fold on a phone. There are only three distinct starting positions on the whole site, 121px, 157px and 181px, which is the page-type template showing through: a breadcrumb or an eyebrow line is the only thing that varies.
The primary call to action is a different story:
pages with a "Start free" control: 74 of 76 (/privacy and /terms have none) above the fold at 390x844: 2 (/, and one blog post) above the fold at 1280x800: 4 median distance from the top of the page: 6,704px On a phone the median reader meets our main button nearly eight screens down. This is not obviously wrong for a content site: the guides are there to answer a question, and a button in your face before the answer is the pattern everyone complains about. It is a choice though, and until today it was an unexamined one. The two pages where the button is above the fold are the two pages someone arrives at having already decided to look at the product.
Nothing is skipped
$ grep -c "content-visibility\|contain-intrinsic" src/app/globals.css 0 On the tallest page, 471 of its 814 elements (58%) start below the fold, and 515 elements on every page belong to the footer, the vast majority of them the paths of one drawing. All of it is laid out and painted on first load. At these sizes that is survivable, and our first-paint numbers are fine, but content-visibility: auto plus a contain-intrinsic-size on the footer is a two-line change that would let the browser skip 1,233px of landscape on every page until it is scrolled into view.
The reason I have not done it already is that contain-intrinsic-size has to be close to the real size or scrollbars jump, and I did not know the real size. Now the real size is a constant, which is the easiest possible case: 1,233px at 390, 501px at 1280.
What I am changing
-
content-visibility: autowithcontain-intrinsic-size: 1233pxon the footer, measured rather than guessed, and the same treatment for the last sections of the two 14,000px pages. - The index pages get content. They are short, repetitive and mostly furniture, and this is the third separate measurement that has said so.
- A rule I will try rather than adopt: no page shorter than four screens on a phone. Below that, the page is a list of links to somewhere else, and a reader who scrolls past it sees more of our footer than of our site.
- The call-to-action position stays where it is on guides, and I will put the measurement in the repository so the next person who argues about it has 6,704px to argue with.
