Not screens. Surfaces.
Auditing feature by feature would never have finished. So the unit of audit became the surface — the recurring construct that every feature is assembled from. Four carry their own criteria. The fifth carries everyone else's.
Every control on the bench is a decision a designer or developer actually makes. The sheet on the right is the inspection sheet rebuilt in this site's own system — same taxonomy, same status vocabulary, same shape as the one the team works from. It re-renders as you work.
- Passed
- Tested on this surface, meets the criterion.
- Failed
- Defect raised against this feature.
- Inherited
- Resolved once at the platform. Carried, not re-tested.
The specimens and inspection sheet are reconstructed in this site's own system. Companies shown are real; identifiers and positions are illustrative.
Audited once. Carried forever.
The first version of the sheet made every feature test all ten selected criteria. It was correct and it was unusable — nobody was going to re-measure the focus ring on the forty-first form.
So the taxonomy grew a third status. Inherited means the criterion is satisfied by a governed platform component, was audited once against Web apps & web services, and is carried by every feature that consumes it. A feature audit only tests what the feature itself introduces.
Open the platform block on the sheet, then switch surfaces. The four blue rows are identical on Tables, Forms, Status indicators and Transactions — same criteria, same wording, same source. That repetition isn't laziness in the artifact. It's the mechanism: a single platform pass, then nothing to re-litigate. Which is why the sheet keeps them folded shut until you ask.
It passes the checks it demonstrates.
Every control here is reachable and operable from the keyboard alone. Each slider is duplicated by paired −/+ buttons, so no part of this page depends on a dragging movement — the same criterion the artifact marks inherited. The verdict panel is a live region, so it is announced rather than merely shown. Contrast ratios are computed with the WCAG relative-luminance formula, not estimated.
Claiming accessibility work without shipping an accessible artifact would be the tell.
The instrument is mine. The readings are not.
Defect counts, closure rates, current platform scores and remediation costs are Avaloq's data. They are not mine to publish.
The evidence I can publish is the instrument itself, rebuilt and running. You can operate every control, break the feature-level criteria exposed on each surface and judge whether the method holds. No quoted outcome could rescue a method that does not.
The same rule governs every case on this site: the artifact is the evidence, the client keeps the outcome.