Build an evidence-based usability teardown

Why use this

1. Usability discussions can drift toward personal taste when reviewers compare unrelated screens or apply different standards to each product. A polished competitor interface may also hide interaction costs that a static screenshot cannot reveal.

2. This job could give the team a common evidence format for discussing design tradeoffs. Separating observed behaviour, rubric interpretation and unanswered questions would make the teardown useful without pretending that desk research measures actual user success.

Why an agent / simpler alternative

Useful judgment

Apply a supplied rubric consistently and connect observations to specific interface evidence.

Simpler alternative

Use a facilitated design critique when only one internal screen needs review.

Unsuitable when

The task expects a universal usability score or conclusions about users without research evidence.

Proposed inputs

1. Name

   Comparison scope

   Owner

   Design lead

   Minimum scope

   Matched user tasks, competitor screens and internal design versions

   Quality check

   Check that products are compared at equivalent stages of the task.

2. Name

   Evaluation rubric

   Owner

   Research owner

   Minimum scope

   Criteria, scoring anchors and evidence requirements

   Quality check

   Separate accessibility observations from conformance claims needing formal testing.

Access and setup — proposed or unknown

Proposed minimum

1. Read authorized screenshots and documentation; work in private draft files or an approved board copy.

Compatibility unknowns

1. Unknown: current screen versions, interaction access and design-tool compatibility.

External actions requiring approval

1. Sharing internal designs, publishing competitor claims or editing a shared workshop board.

Expected deliverable and review checks

Proposed artifact

Evidence-based teardown draft with scored criteria, screen references, uncertainty and discussion prompts.

Critical checks

1. Every score cites a visible observation or is marked unassessable.

2. Internal and competitor designs receive the same scoring standard.

Missing output evidence

No new run or output inspection in this editorial layer; see linked source-specific limits.

Failure signals and proposed responses

1. Signal

   A competitor’s completed state is compared with an internal empty state.

   Why it matters

   The score reflects different contexts rather than design quality.

   Proposed response

   Align task states or keep the comparison explicitly incomplete.

2. Signal

   A static screenshot is used to claim an interaction is inaccessible.

   Why it matters

   The conclusion may exceed the available evidence.

   Proposed response

   Describe the visible concern and propose a separate authorized interaction test.

Effort and cost — unknown

Setup work

1. Align tasks and calibrate rubric examples.

Ongoing review

1. Review scoring disagreements and update stale screen evidence.

Unknowns

1. Unknown: research effort, tool access costs and need for user testing.

First check — proposed, unexecuted

Status

proposed_unexecuted

Synthetic fixture

Synthetic screens show the same task at different stages, and one rubric criterion requires keyboard interaction.

Acceptance checks

1. The mismatched states are not directly ranked.

2. Keyboard behaviour remains unassessed from screenshots alone.

Stop conditions

1. Private design sharing or unsupported user-performance claims would be required.

Limits

1. A teardown is a structured expert assessment; it does not substitute for usability testing or establish competitive superiority.