Working Design skill
Measure design quality
Measure design quality through user outcomes, accessibility, consistency, reliability, and operational evidence instead of taste or output volume.
Measure the experience and its consequences, not how many screens shipped.
When to use
- Use for product areas, journeys, design-system adoption, or release quality.
- Do not collapse distinct harms and outcomes into one score without preserving the underlying measures.
Procedure
- Define users, critical journeys, intended outcomes, risks, and decisions the measures should inform.
- Select a balanced set covering task success, time or effort, errors, accessibility, comprehension, trust, support, reliability, consistency, and user-reported experience.
- Define each measure's source, population, denominator, segment, period, owner, target, limitations, and privacy boundary.
- Establish baseline and production-shaped test cases before changing the design.
- Combine quantitative behavior with usability observations, complaints, accessibility findings, and expert review.
- Segment by relevant journey and context without exposing small groups or using protected traits improperly.
- Distinguish leading checks, release gates, live outcomes, and lagging harm signals.
- Review tradeoffs and guardrails; an improved conversion measure must not hide coercion, errors, or inaccessible completion.
- Reconcile measures after release and change the design or measurement when evidence contradicts the hypothesis.
Failure plan
- If a metric is easy to improve without helping users, demote it and add an outcome guardrail.
- If data collection creates disproportionate privacy or accessibility harm, use a safer evidence source.
--- name: measure-design-quality category: design description: Measure design quality through user outcomes, accessibility, consistency, reliability, and operational evidence instead of taste or output volume. Use when a team needs repeatable signals for design decisions and improvement. --- # measure-design-quality Measure the experience and its consequences, not how many screens shipped. ## When to use - Use for product areas, journeys, design-system adoption, or release quality. - Do not collapse distinct harms and outcomes into one score without preserving the underlying measures. ## Procedure 1. Define users, critical journeys, intended outcomes, risks, and decisions the measures should inform. 2. Select a balanced set covering task success, time or effort, errors, accessibility, comprehension, trust, support, reliability, consistency, and user-reported experience. 3. Define each measure's source, population, denominator, segment, period, owner, target, limitations, and privacy boundary. 4. Establish baseline and production-shaped test cases before changing the design. 5. Combine quantitative behavior with usability observations, complaints, accessibility findings, and expert review. 6. Segment by relevant journey and context without exposing small groups or using protected traits improperly. 7. Distinguish leading checks, release gates, live outcomes, and lagging harm signals. 8. Review tradeoffs and guardrails; an improved conversion measure must not hide coercion, errors, or inaccessible completion. 9. Reconcile measures after release and change the design or measurement when evidence contradicts the hypothesis. ## Failure plan - If a metric is easy to improve without helping users, demote it and add an outcome guardrail. - If data collection creates disproportionate privacy or accessibility harm, use a safer evidence source. ## Done - A design-quality measurement plan records journeys, measures, definitions, baselines, sources, owners, targets, guardrails, and limits - A review report reconciles quantitative, qualitative, accessibility, reliability, and support evidence into explicit decisions