Anaks Labs
Article 04

Beautiful, and invisible

The most disorienting thing a practice owner can be told is that their expensive, handsome, recently finished website is unreadable. It sounds like a sales line. This article is about why it is not a contradiction, and about the group of sites we deliberately included in our audit to test whether it was.

Published · Anaks Labs

One page, two separate properties

Whether a website is good and whether a website can be read are independent facts about it. That is the entire idea, and everything below is a consequence of it.

A page has an appearance, which is what a visitor judges: the photography, the balance, the typography, how quickly it settles, whether it feels like the kind of place you would let near your teeth. It also has a body of text that anything other than a human eye has to work from. Those two are produced by the same build and they are not the same object, and it is entirely possible to invest heavily in the first while the second quietly empties out.

Nobody in the process is behaving badly when this happens. The designer delivered what was asked for. The developer built it faithfully. The owner approved something that looked right, because it did look right. There is no point in the chain at which anybody sees the missing half.

The group we included to test this

Our audit ran across 182 clinic sites in the United States, and they were not one sample. A hundred and twenty-two came from a random draw of the federal provider registry. The other sixty were a comparison set: clinic sites that other people — design blogs, marketing agencies, nine separate round-ups — had already picked out and published as good examples. Not our choices, and not awards.

We included that second group for exactly one purpose, which was to find out whether being well made protected against any of this. It did not.

Anaks Labs field work. The two groups were drawn in completely different ways and we never pool them into a single percentage. Sample definition, both groups, and the limits →

Where the readable half goes

Three patterns account for most of what we saw. Each of them is invisible from the visitor's side, which is the only side the person paying for the site ever looks at.

  • The headline is set as a picture

    The best sentence on the site — the one the practice argued about for a week — is drawn as an image so that the type is exactly right. On screen it is perfect. As text it does not exist, and the single most important sentence on the page becomes an empty box to anything reading it.

    We measured this on four per cent of the registry sites and on under two per cent of the comparison set. It is not the common problem. It is the clearest one, and where it happens the loss is total rather than partial.

    Anaks Labs field work: 4.1% of the 122 registry sites, 1.7% of the 60-site comparison set. Base, definition and scope →

  • The words live in something that moves

    The panel that rotates through your six services, the accordion, the carousel of testimonials. A visitor swipes and reads. Very often those words are not part of the page when it arrives — they are put in afterwards — and what a visitor swipes through was never handed over at all.

  • The page never states, in a form anything can carry away, what the business is or where it is

    This is the common one and it is the least dramatic. On more than half the registry sites, nothing on the page declares that this is a dental practice in a particular town in a way a machine can lift. The address is very often printed there for a person to read. Printed and declared are different things.

    Anaks Labs field work: 57.4% of the 122 registry sites. The base and what counted →

What the difference measures out at

Because we are not allowed to publish anybody's real clinic site, we built one. Specimen Dental does not exist: it is a practice we invented, and both halves of it are ours — one homepage written in the shape the audit kept turning up, and the same content built the way we build. It is two pages we wrote, not a survey of the industry, and every number in this section is a measurement of those two pages only.

On the first of them, three of the five section titles arrived as real text and two arrived as pictures. On the second, seventeen titles out of seventeen arrived as text and none as a picture. Same practice, same services, same hours, same headline.

Measured on the two specimen builds we wrote. Definitions for every row are printed beneath the table itself. The two builds, and what each row counts →

The reason we show a specimen rather than a client is that we have no rebuild we are permitted to publish, and we would rather invent a before than anonymise somebody's real one. Both builds are live pages you can open and scroll.

Open both builds, side by side, with the full count →

Why you cannot see this on your own site

Because your website is not failing. It loads, it is legible, it works on a phone, the photographs are good and the contact form sends. Every symptom described above is consistent with a page that behaves perfectly for every human being who has ever opened it, including the one who paid for it. There is no error message, no broken link and no bad review. The absence is only observable from a position no visitor ever occupies.

Which is why the check is worth running deliberately rather than waiting to notice something. It does not need a tool and it does not need us.

Read next