Loading
Measured 12 October 2026
Not a survey and not an opinion. We requested 4,519 company homepages the way a crawler does, read what came back, and counted. 3,002 had at least one thing worth fixing. 1,098 had nothing at all — which is the number most reports like this leave out.
4,519
Homepages requested
Every company in our own outreach database that had a website on file.
4,100
Answered
The other 419 did not respond at all. Every figure below is out of the ones that did.
73%
Had something worth fixing
3,002 sites. The commonest single finding was "an enquiry path with gaps".
27%
Had nothing worth fixing
1,098 sites we would not change. We say so because a report where everyone fails is a sales document.
16 findings, counted
Every row is a measurement, a consequence, and the thing we would put in its place. The order is how common each one is, not how much we would like to sell you.
67% of sites · 2,740 of 4,100
What it costs
A form but no tap-to-call, or a phone number but nothing to type into. On a phone, each missing route is an enquiry that depends on the visitor being determined.
What we build
Every route the device offers, so getting in touch costs one tap rather than a decision.
55% of sites · 2,270 of 4,100
What it costs
Nothing on the page states who the company is in a form an engine can parse, so assistants fall back to third-party directories — which is where wrong facts come from.
What we build
The company described in the page itself, in the form assistants read.
55% of sites · 2,257 of 4,100
What it costs
Someone ready to talk has to send a message and wait for a reply before a time exists. Every round trip loses the people who were ready at that moment.
What we build
A slot bookable from the page, confirmed immediately, with the brief already attached to it.
55% of sites · 2,249 of 4,100
What it costs
Not a fault by itself, but it usually means the front end has not been opened in several years — and whatever else has not been revisited with it.
What we build
A front end rebuilt on current tooling, so page speed stops being inherited.
47% of sites · 1,940 of 4,100
What it costs
Fetched while the page is still drawing, including the ones nobody scrolls to. On a phone connection that is the difference between a page appearing and a page arriving.
What we build
Images deferred, so the first screen does not wait for the last one.
33% of sites · 1,355 of 4,100
What it costs
No form, no input, no tap-to-call, no WhatsApp, no email link. Every enquiry depends on the visitor working out how to reach you unprompted.
What we build
One enquiry route that works on a phone, and an assistant that takes the brief before it reaches anybody.
31% of sites · 1,264 of 4,100
What it costs
Nothing in the markup states what the page is about. Engines and screen readers both fall back to guessing from the rest of it.
What we build
The page saying what it is, in the element everything reads first.
31% of sites · 1,252 of 4,100
What it costs
An engine cannot tell what any of them show, and a screen reader announces them as nothing.
What we build
Images described — an accessibility fix and a visibility one at the same time.
26% of sites · 1,054 of 4,100
What it costs
The sentence under the company name in a search result is whichever fragment the engine decides to lift from the page.
What we build
The description written rather than guessed at, on the pages people arrive on.
15% of sites · 611 of 4,100
What it costs
Measured before a single image or font loads. Above a second and a half it is a hosting problem rather than a design one, and it costs the visitors who never see the page.
What we build
The page rendered ahead of time, so the first byte arrives immediately.
5% of sites · 225 of 4,100
What it costs
A page containing <center>, <font> or bgcolor was hand-built long ago and patched since. It will not survive a redesign of anything.
What we build
The same page on current markup, so the next change does not mean touching all of it.
4% of sites · 155 of 4,100
What it costs
Too little for a search engine to understand the business, and too little for an assistant to describe it to someone who asked. Thin pages get summarised by guesswork.
What we build
A homepage that states the offer in text, so engines quote it rather than guess.
2% of sites · 102 of 4,100
What it costs
Requested the way a crawler does, it returns an empty shell. Search engines mostly cope now; AI assistants frequently do not, and nor does anything else reading the page without running it.
What we build
A page whose text is in the HTML, so an assistant asked about the company can describe it.
2% of sites · 94 of 4,100
What it costs
Browsers label it "Not secure" in the address bar, next to the company name, before a visitor has read a word.
What we build
HTTPS with the certificate and the redirects handled properly.
1% of sites · 57 of 4,100
What it costs
No viewport tag, so the page renders at desktop width on a phone and has to be pinched to read. Most traffic is mobile.
What we build
A layout that fits the screen it is opened on, phone first.
1% of sites · 42 of 4,100
What it costs
Invisible to those assistants whatever else is on the site. It usually arrives with a plugin or CDN default rather than by decision.
What we build
Crawl rules set deliberately, so the assistants you want quoting you are allowed in.
How it was measured
Stated plainly, because a study that hides its method is a marketing claim with a number attached.
The homepage only, requested over the public internet the way any crawler would, with no login and nothing submitted. We did not test a form, a checkout, a mobile layout, or anything behind a sign-in.
Not a human impression and not a score out of a hundred. Each finding is a specific condition in the markup or the response, so the same page gives the same answer every time, and you can check any of them yourself.
A single slow response can be a cold cache or our own network. Every site that looked slow was read again seconds later, and the finding was dropped unless it reproduced — which it did not for 470 of them.
We measured servers and markup, not sales. Nothing here says a finding costs a percentage of anything, because we have never seen these companies’ numbers and neither has anyone else publishing figures like these.
Some sites return an empty shell and assemble themselves in the browser. We count those as what they are rather than claiming they have no contact form — the shell is not evidence that anything is missing.
These are the companies in our own outreach database. The counts are published; the list is not, and no individual site is identified anywhere on this page.
Read on 12 October 2026. The numbers on this page are generated from the audit data on every build, so they cannot drift away from what was actually measured. If you want the method in more detail, it is how we work on everything.
Your site
The same read, on your address, in about six seconds. You get the findings whether or not you tell us who you are.
Takes a few seconds. No email needed, and nothing is sent to you unless you ask.
If the read turns up something worth fixing, we will build you a working version of that page — at no cost, with nothing to sign, and yours whether or not we go further.