usability-testing-and-evaluation

Validate usability cheaply and act on it: neutral facilitation, card sorts and prototype tests

SKILL.md

GitHub

Usability Testing and Evaluation

Validate usability cheaply and act on it: neutral facilitation, card sorts and prototype tests fitted to the research goal, immediate debrief, and ruthless triage of the most serious problems first.

When this applies

  • running a usability-test session (P004).
  • Studying users' mental models for grouping, sorting, or labeling content and tasks (P008).
  • Candidate organization, labeling, or hierarchy choices are ready for validation (P010).

Procedure

Apply these principles to the situation under review; for each, name the user need at stake and the trade-off the choice carries.

  1. Facilitate neutrally: choose a facilitator with the right temperament because a bad one tanks the test, avoid leading or rescuing lost users and embrace silences (the system's own designer is most tempted to hint), and when a participant blames themselves ask how they expected the system to work and why. (P004)
  2. Design card sorts to match the research goal, using open sorts for discovery, closed sorts for validation, and careful analysis for qualitative versus quantitative use. (P008)
  3. Test strategy before commitment with adapted research methods, closed card sorts, prototype task tests, hierarchy tests, and simple wireframes. (P010)
  4. Debrief immediately and triage ruthlessly: fix the most serious problems first via a collective list, top-ten selection, ranking, and an ordered fix list with owners; fix only enough to declassify each serious problem then stop; keep a separate low-hanging-fruit list; and prefer removing what obscures meaning over adding explanations. (P064)

Principles applied

  • P004 (high) — grounded in claims/evidence and chunk anchors in sources/.
  • P008 (medium) — grounded in claims/evidence and chunk anchors in sources/.
  • P010 (medium) — grounded in claims/evidence and chunk anchors in sources/.
  • P064 (high) — grounded in claims/evidence and chunk anchors in sources/.

Provenance

Grounded in principles P004, P008, P010, P064, their backing claims and evidence records, and paragraph-level source anchors under sources/anchors/. Every cited id resolves into this package's distilled spine; see provenance-ledger.md and reports/faithfulness-report.yaml.