Three typical starting positions

Starting position

An electrical contractor with 35 staff in a mid-sized town. The website was seven years old, barely usable on a phone and took over eight seconds to load. The business was desperate for fitters and apprentices but received almost no applications through its own site. Searching for the core service plus the town name put them on page three, behind two directories and four competitors.

The actual problem

Not the visuals. The site had no subpage for any of its services, everything sat on the homepage. For a search engine there was no way to tell what the business stood for. On top of that, the careers section consisted of a PDF and a generic info address.

What we would do

• Built a service structure: one page per trade, each with local context
• Added structured data for the business, opening hours and service area
• Fully completed the Google Business Profile and linked it to the site
• Replaced the PDF download with a three-field short application
• Rebuilt on a fast foundation, bringing load time under two seconds
• Set up a process where fitters ask for a review after finishing a job

Where it typically goes wrong

Application forms with seven fields and a mandatory CV upload routinely hit abandonment rates above 80 percent. Tradespeople apply from a phone, often on a lunch break, and do not have a CV as a PDF to hand. We cut it back to name, contact and a free-text box. The CV gets requested on second contact.

Starting position

A medical centre with four locations and online appointment booking. Their accountant had flagged the Austrian accessibility act (BaFG), in force since 28 June 2025 for electronically provided consumer services. Nobody internally could judge how far the site was from compliance.

The actual problem

An audit in a practice of this kind typically turns up 40 to 60 violations of WCAG 2.2 AA. The serious one is almost always the booking flow: calendar widgets assume mouse clicks and cannot be operated by keyboard. For blind patients, booking online is therefore impossible. Legal exposure and exclusion of real users coincide here.

What we would do

• Full audit, automated plus manual with keyboard and screen reader
• Rebuilt the booking flow, fully keyboard operable and correctly labelled
• Raised every colour contrast to at least 4.5 to 1 without abandoning the brand
• Forms with visible labels and errors shown next to the field
• Published an accessibility statement and a route to report barriers
• Trained the editorial team so new content does not undo the work

What matters most here

The heaviest part is rarely the engineering, it is the images. Sites that have grown over years easily carry several hundred photos with no usable alternative text. We generate drafts automatically and have them reviewed by clinical staff. Without that review, medically inaccurate descriptions ship, and those are worse than none.

Starting position

An engineering firm with 22 staff and a highly specialised core discipline. The market is small, the purchase decisions are large. The website was technically fine but editorially a self-description: a lot about the company, little about the problems it solves.

The actual problem

A firm like this usually does not rank for the technical term describing its core service. Worse: ask common AI assistants for providers in that field and the company is never named once, despite being one of very few specialists in the German-speaking market. The expertise exists but is nowhere recorded in a form a machine can connect.

What we would do

• Built a topic cluster: one core page, surrounded by deeper articles
• Shifted the content from self-description to problem solving
• Added structured data linking company, service and field of expertise
• Turned reference projects into standalone, linkable pages
• Wrote articles so individual sections can be cited on their own
• Measured AI answer visibility monthly against a fixed set of test questions

What is unusual here

Search volume is the wrong metric. Terms like this get searched a few hundred times a month at most. What matters is the quality of enquiries. So in engagements like this we report qualified first conversations from the start, not visitor counts.