Skip to content

Work samples

Built, then measured

Every page we build or repair goes through the same testing as our clients' websites before handover. Here are four sample sites for fictional businesses, with every finding, including the awkward ones.

These businesses do not exist. Forms send nothing, the shop sells nothing. Photos: Unsplash.

Screenshot of the first screen of Holzwerkstatt AumĂĽller

Craft business with wardrobe planner

Holzwerkstatt AumĂĽller

A joinery that wants to win jobs through its website. A planner that turns dimensions and wood into a sketch and guide price instantly, and an enquiry that carries those dimensions along.

What is built in

  • Wardrobe planner: width, height, depth, wood and interior with a live sketch and guide price
  • The enquiry takes over the planner's dimensions, a photo of the room can be attached
  • Services as a tile grid, process with the duration of each step

Measured on 12 September 2026

automatically confirmed findings
0
items for review
5

Four were false alarms in our own software and are now being fixed. The fifth (visible focus) was checked on the skip link during review: visible.

Measured with the rules, keyboard, zoom and reflow, accessibility tree and dynamic content layers. The AI layer was not run for this measurement.

Found in the first run and fixed

  • Focus ring in the dark notice bar reached only 2.98:1 contrast (WCAG 1.4.11)
  • Logo initials were visible but missing from the link's accessible name (WCAG 2.5.3)
Open sample site: Holzwerkstatt AumĂĽller
Screenshot of the first screen of Physiotherapie Lindenhof

Practice with online booking

Physiotherapie Lindenhof

A practice whose patients are often older or in pain right now. Booking in three steps, a symptom finder and opening hours that show whether the practice is open right now.

What is built in

  • Three-step booking with free and taken slots and a live summary
  • “Where does it hurt?”: pick a body area, see the matching treatment, book directly
  • Open status in the header and today highlighted in the hours table

Measured on 12 September 2026

automatically confirmed findings
0
items for review
7

Six were false alarms in our own software, including text inside closed disclosure widgets that is not rendered at all. The seventh (visible focus) was checked during review: visible.

Measured with the rules, keyboard, zoom and reflow, accessibility tree and dynamic content layers. The AI layer was not run for this measurement.

Found in the first run and fixed

  • Two required fields were not identified as required anywhere (WCAG 3.3.2)
  • Focus ring in the notice bar barely visible in dark mode at 1.49:1 (WCAG 1.4.11)
  • Logo initials missing from the accessible name (WCAG 2.5.3)
Open sample site: Physiotherapie Lindenhof
Screenshot of the first screen of Rösterei Kornblum

Online shop with subscriptions

Rösterei Kornblum

A shop of the kind the European Accessibility Act covers: product pages with taste profiles, subscriptions, a coffee finder, a checkout with a review step and its own accessibility information page.

What is built in

  • A page per coffee with taste profile, pack size, grind and subscription option
  • Coffee finder: three questions, one recommendation that updates instantly
  • Checkout with a review before the binding order button and an error summary with jump links

Measured on 12 September 2026

automatically confirmed findings
0
items for review
6

All six came from the same heuristic: a prompt like “The postcode has five digits, for example 04229” was not recognised as a correction suggestion. False alarm, on our list.

Measured with the rules, keyboard, zoom and reflow, accessibility tree and dynamic content layers. The AI layer was not run for this measurement.

Found in the first run and fixed

  • Six required fields with no indication that they are required (WCAG 3.3.2)
Open sample site: Rösterei Kornblum
Screenshot of the first screen of Klarwerk Gebäudereinigung

Service business with a price estimator

Klarwerk Gebäudereinigung

A cleaning company that wants enquiries instead of phone calls. An estimator that turns floor area, building type and frequency into a price range before anyone has to ask, and a form that carries those answers along.

What is built in

  • Price estimator: area, building type and frequency give a range, derived from output per hour and hourly rate
  • Enquiry form that carries the estimator answers over, with required fields spelled out instead of marked by an asterisk
  • Structured data as ProfessionalService so search engines can read service, area and contact

Measured on 16 September 2026

automatically confirmed findings
0
items for review
3

Two come from the same heuristic as the roastery: “Please enter an email address, for example name@firma.de” was not recognised as a correction suggestion. On the third, one of 22 tab stops could not be measured for a visible focus indicator.

Measured with the rules, keyboard, zoom and reflow, accessibility tree and dynamic content layers. The AI layer was not run for this measurement.

Found in the first run and fixed

  • The page scrolled sideways at 320 CSS pixels because the contact list had a fixed column width (WCAG 1.4.10)
  • The borders of the input fields reached only 1.38:1 instead of the required 3:1 (WCAG 1.4.11)
  • At 200 % text size the header button slid out of the viewport (WCAG 1.4.4)
  • The skip link turned dark on a dark background while hovered (WCAG 1.4.3)
Open sample site: Klarwerk Gebäudereinigung

Before, after, measured

For these three pairs we deliberately built broken versions and then fixed them, the way we do on client sites. Both versions are public, every number can be measured again.

Why we show the false alarms

No testing software is free of errors, ours included. The difference is what happens next: nothing ambiguous is decided automatically, every review item gets resolved, and every false alarm becomes a fix to the software. The four sample sites uncovered four of them.

Pilot audit of a non-profit organisation

A website on a content management system, all key page templates tested. The September 2026 report lists 154 findings that trace back to 23 causes: one colour pair in the stylesheet produces 14 findings, one template without a main region five. Fixes are made per cause, not per finding. The name follows once the organisation agrees.

Your site next?

We build new or fix what was measured on your site, at a fixed price and with evidence for every point.

Send a request