Running a Usability Test
Test design, collect data, and improve UX.
Step 1: Plan the Test
Define research questions: Can users complete checkout? Where do they hesitate on onboarding? Recruit 5 participants matching your persona — more returns diminishing returns for qualitative issues.
Choose moderated (live session with think-aloud) or unmoderated (async tool). Moderated tests surface why users struggle; unmoderated scales faster but give less depth.
Step 2: Write Tasks and Script
Tasks should be scenario-based, not UI instructions. Bad: "Click the blue button." Good: "You want to change your delivery address before placing the order."
Open with intro, consent, and rules: think aloud, no wrong answers, prototype may be incomplete. Close with debrief questions about overall impression and missing features.
Step 3: Run Sessions
Record screen and audio with permission. Take notes on quotes, errors, and time on task. Avoid leading help — if stuck, wait, then offer a neutral prompt.
Watch for success, partial success, failure, and workaround behavior. Workarounds often reveal mental models your design does not match.
Step 4: Synthesize Findings
Affinity map observations into themes: navigation confusion, unclear labels, trust issues. Rate severity: critical (blocks task), major (slows task), minor (cosmetic annoyance).
Turn themes into actionable recommendations tied to evidence: "3 of 5 participants could not find settings — add label or move entry point." Avoid solution theater without linking to observed behavior.
Step 5: Prioritize Fixes
Use impact vs effort matrix. Fix critical blockers before polish. Quick wins — clearer button copy — ship in days; structural flow changes need design and eng time.
Share a one-page summary with clips or timestamp links for stakeholders. Video clips of real struggle persuade better than bullet points alone.
Step 6: Iterate and Retest
Update wireframes or prototype based on findings. Retest changed areas with 3–5 new or returning participants when flow changes are significant.
Usability testing is continuous, not one gate before launch. Lightweight monthly tests beat a single expensive study that goes stale after release.
FAQ
How many participants do I need?
Five moderated sessions usually surface most major usability issues on a single flow. Add more for diverse segments or quantitative confidence, not because five is a magic law for every study.
Prototype or live product?
Test prototypes early to fix flow cheaply. Test production for performance, real data, and edge cases prototypes fake. Match fidelity to your research question.
Should I help when users are stuck?
Let struggle happen briefly — it is data. Intervene with neutral prompts if they abandon the task entirely. Note every hint you give; heavy assistance means a critical issue.