Twelve mockups, twelve reasons a site was failing.

The situation. Six providers across two offices. Hours were published in a header graphic, again inside a third-party booking widget, and again on the map listing — all three different after a schedule change nobody propagated. Patients were arriving at a closed office.

What we built. One canonical fact set covering both offices, and a page per location generated from it: address, hours, parking, providers and accepted insurance, all as selectable text. Structured data emitted from the same fact set, so the markup cannot drift from the page. The booking widget embedded beneath the information rather than in place of it.

How we verified it. Hours identical in the page text, the structured data and the map listing. Each location has its own URL and address block. Every page renders completely with JavaScript disabled.

The situation. Two decades of completed work living entirely inside downloadable capability statements and project PDFs. The site itself was four pages of copy that could have described any engineering firm in the country, and no page said which states the firm was licensed in.

What we built. Project pages carrying scope, sector, location, structural system and completion year as text, with the original PDF offered beside each one rather than instead of it. A licensure page naming each jurisdiction and licence type. Team pages with credentials and the sectors each engineer leads.

How we verified it. Every jurisdiction, credential and project fact appears as text in the served HTML and was confirmed by the firm before publication. Structured data for the organisation and named staff matches the visible pages exactly.

The situation. The entire site was a JavaScript shell wrapped around a third-party booking engine. Room descriptions, photographs, the cancellation policy and even the street address lived on the vendor's domain. With scripts blocked, the inn's homepage was blank.

What we built. A static site on the inn's own domain carrying room types, rates policy, directions, parking, accessibility notes, house rules, breakfast times and check-in windows as readable text. Photography served from the inn's own hosting. The booking engine reached from a clear route, doing what it is good at.

How we verified it. Every policy and property fact is readable with JavaScript disabled. Footer address, structured data and map listing are identical. The site would still describe the inn correctly if the booking vendor disappeared tomorrow.

The situation. A single “Areas We Serve” page listing eleven towns inside one paragraph. The phone number appeared once, rendered inside a footer image, so a caller on a phone had to scroll the whole page and retype it by hand. Overnight emergency cover — the firm's main advantage — was mentioned nowhere.

What we built. A tappable phone link in the header on every page and screen size. Genuine per-area pages, written rather than spun, stating the services offered there, realistic response expectations and whether emergency cover applies. Service-area structured data matching the pages, with no towns claimed that the firm does not serve.

How we verified it. The number is a tappable link inside the first screen on a phone and identical everywhere it appears. Each area page contains distinct written content. Markup, pages and the owner's own list of towns all agree.

The situation. Product descriptions, ingredients, allergen statements and sourcing existed only inside marketplace listings and a social profile, in three versions that did not match. The bakery's own domain redirected to the social page, so retail buyers looking the company up found nothing to point anyone at.

What we built. An owned site with a page per product carrying ingredients, allergens, net weight, shelf life and sourcing as text. A stockists page listing shops and marketplaces as channels rather than as the source of record. A wholesale page with case configurations and a route to a real person.

How we verified it. Allergen and ingredient text is now written once, matches every channel, and was confirmed in writing by the producer before launch. Product structured data mirrors the page. The domain resolves to the company's own site.

The situation. Four hundred parts distributed as a downloadable spreadsheet, a single contact form as the only route to a quote, and certifications shown as logo images with the issuing body and scope written nowhere. An engineer searching a part number found nothing at all.

What we built. Category and part pages generated from the distributor's own data, with specifications as text and tables and part numbers in the headings and the URLs. A certifications page naming each standard, issuer, scope and expiry, linked to the registries where each can be checked. A quote route that captures part numbers and quantities and reaches a person.

How we verified it. Every part number is reachable at a crawlable URL with its specifications as text. Certification claims name a verifiable issuer and scope. The spreadsheet is still downloadable — it is simply no longer the only way in.

The situation. Marketing copy implied nationwide practice while the firm was admitted in three states. Attorney pages listed names and photographs but no bar admissions, no year of call, and no practice area. For a firm whose whole product is credibility, none of it was verifiable on the site.

What we built. An admissions page naming each jurisdiction and bar number. Attorney pages with year of admission, practice areas, reported matters and languages. Practice-area pages describing what the firm actually does in each, written by the partners who do it, replacing the borrowed copy.

How we verified it. Every jurisdictional claim on the site matches the bar records the firm supplied. Person structured data lists only credentials stated on the page. No claim of reach beyond the states named.

The situation. Emergency availability, surgery days and drop-off windows were spread across a graphic, a PDF new-client packet and a social post from the previous year. The practice is not a 24-hour hospital, but nothing on the site said so, and worried owners were arriving at a locked door at night.

What we built. Service pages stating what is treated and when, with surgery days and drop-off windows as text. An after-hours page saying plainly that the practice is not a 24-hour hospital and naming the two nearest emergency clinics with directions. A new-client route that requests records instead of starting a phone chain.

How we verified it. Hours and emergency status are text on the page, in the structured data, and on the map listing, in agreement. The after-hours page names real external clinics the practice confirmed and links to them.

The situation. Every practical question — what to wear, how long a first visit takes, whether a referral is needed, which plans are accepted — was answered by the front desk, thirty times a day, because none of it was written anywhere. The site had four pages and a contact form.

What we built. A first-visit page answering each of those questions in order. An insurance page listing accepted plans with the date each was last verified. Condition pages describing what treatment actually involves. Therapist pages with specialisms, so a patient can ask for the right person by name.

How we verified it. Each answer on the site matches what the front desk says, confirmed by the practice manager line by line. Plan lists carry a verification date. Nothing about coverage is promised — the page says what is accepted and what to check.

The situation. An image-driven portfolio with almost no text: project names over full-bleed photographs, a single line of description, and no programme, site, area, year, client or consultant anywhere. Beautiful on a fast connection, and blank to anything reading the HTML.

What we built. A project record beneath every image set — programme, location, floor area, completion year, structural and services consultants, awards where they exist. A studio page with the practice's size, registrations and the way it works. The photography kept exactly as it was, now with words underneath it.

How we verified it. Every project remains fully described when images fail to load. Registration claims name the states the practice supplied. Page weight fell despite the added text, because the images were resized properly.

The situation. A four-page brochure site and a support inbox that answered the same seasonal questions hundreds of times: which filing applies to me, what do you need from me, and when. All of it existed — in the partners' heads and in last year's email threads.

What we built. A deadlines page as a real table: filing, who it applies to, the documents required, and the date the firm needs them. Service pages for each engagement type. An industries page describing the sectors the firm actually serves, and a licensing statement naming the state board.

How we verified it. The table is text, not an image or a PDF, and is reviewed by a partner each season. Dates on the page match the firm's internal calendar. No claim of registration the firm does not hold.

The situation. Producer, region, process and elevation appeared only in a photograph of the bag. Wholesale buyers asked for a spec sheet by email every week, and the site could not answer the one question that decides a purchase: what is in the bag right now.

What we built. A page per lot naming the producer, region, altitude, process, varietal and roast profile as text, with the roast-day schedule stated by weekday. A wholesale page with case sizes and lead times. A subscription page that explains what arrives and when before asking for a card.

How we verified it. Lot details are text and change when the lot changes, from one source the roaster updates. Product structured data matches the visible page. Photography of the bags is still there — it is no longer carrying the information.