Nobody in this trade wants a website. They want the phone to ring.

Twenty-two sites were measured in a real browser — twelve trade sites, nine well-designed businesses from other categories, and a lead-generation marketplace as a control. Not read, not judged by eye. Rendered and measured.

The findings were worse than expected, and they are the whole basis for how these sites get built:

• 6 of 12 trade sites have no click-to-call above the fold on a phone at all.
• 3 of 12 have no <h1>, and of the nine that do, only three name the town they work in.
• 0 of 12 let you send a photograph of the broken thing.
• 1 of 12 marks up its reviews so search engines can read them.
• The median trade homepage is 4.2MB. The worst two measured 14MB and 23MB — delivered to someone standing outdoors next to the broken thing, on cell data.

Hand-written HTML, CSS and JavaScript on static hosting. No WordPress — that means plugin updates, security patches and a 2am phone call from a client who will not pay for any of it.

Every build is checked by an acceptance harness before it ships: contrast composited for transparency, every tap target measured at 44px, the page rendered with images disabled, and load time measured on a throttled mobile profile rather than on a desk. The reference build passes 25 of 25 and ships at 215KB, reaching first paint in about a second on a throttled 4G connection.

Each check has been watched fail on a page built to break it. A test that has never gone red is not evidence.

It will not invent anything about your business. Not a review, not a price, not a licence number, not an opening time. Anything not yet supplied renders as a visible placeholder that cannot ship by accident, and a build tool lists every one of them.

It will not promise a ranking. Nobody can. What it will do is stop the self-inflicted losses — a page that 404s a URL you already ranked on, a set of near-identical town pages, reviews no search engine can read.

It will not sell you a site and skip the Google Business Profile. For a local trade the profile carries the large majority of the calls. Selling the site alone would be underdelivering on the only number the client actually counts.

My Treasure Valley Handyman — Robbie York, Meridian. Two dark directions on the same four screens, so they compare like for like: home, a service page, the estimate flow, and the desktop home. Both are built on a design system generated from his own logo, and both carry his real phone number, his real hours and the six towns he actually serves.

Everything not yet supplied is a visible bracketed placeholder — the review text, the star rating, the job photographs. None of it is invented, and none of it can ship by accident.

Both are live, responsive sites rather than pictures of sites — open them on a phone and a desktop and you get the two different designs he drew, not one stretched to fit.

Direction A →  ·  Direction B →  ·  the artboards

The research, the reference build and the verification tooling are finished. The first client site is built and running on a private preview, and is not launched — it is waiting on photographs and a licence number from the owner, and on his decision between two design directions.

So this door is open on the practice more than on a portfolio. There is one client and the work is real and paid. The design is below; the launched site is not public until the owner supplies photographs and chooses his direction.