The Analysis Problem
You've run five usability sessions, collected hours of recordings, pages of notes, and a spreadsheet of task completion data. Now what? Raw observations are not insights - they're ingredients. The analysis phase transforms scattered data into clear findings that stakeholders can act on. Most teams rush this step; the ones who don't consistently make better product decisions.
Step 1: Organise Your Raw Data
- Compile all notes from every session into a single document or Miro board.
- Use a per-participant structure: participant → task → observation.
- Timestamp video recordings for moments worth revisiting.
- Separate direct quotes (what users said) from behaviours (what users did) from interpretations (what you think it means). Analysis mixes all three; keeping them distinct during collection improves accuracy.
Step 2: Affinity Clustering
Write each distinct observation on a sticky note (digital or physical). Group related observations together - patterns emerge when you see that 4 out of 5 participants hesitated at the same step. Name each cluster with an insight statement, not a description: not "Users clicked the wrong button" but "The primary and secondary CTAs are visually indistinct, causing frequent mis-selection."
Step 3: Prioritise by Severity and Frequency
Rate each finding on Nielsen's 0–4 severity scale: 0 (not a real problem), 1 (cosmetic), 2 (minor, low priority), 3 (major, fix before release), 4 (catastrophe, blocks release). Then cross it against frequency - how many participants hit it. A severity-4 issue that one person out of five hit still outranks a severity-1 issue every participant noticed. The combination, not either axis alone, decides what goes into the next sprint.
Step 4: Form Actionable Recommendations
Every finding should pair with a recommendation: "We observed X [finding]. This suggests Y [interpretation]. We recommend Z [action]." Vague findings without recommendations produce research reports that sit unread. Specific, actionable recommendations produce design briefs.
Step 5: Present Findings Effectively
- Lead with the top 3–5 findings, not a list of every observed issue.
- Use participant quotes and short video clips - they're more persuasive than any chart.
- Include task completion rates and time-on-task data for quantitative context.
- Tailor depth to audience: executives want impact and recommendation; designers want specific observations; engineers want scope and priority.
Key Takeaways
- Affinity clustering turns scattered observations into named, actionable themes.
- Severity × frequency prioritisation focuses team effort on problems that actually matter.
- Every finding needs a paired recommendation - research without direction stalls.
- Video clips and participant quotes move stakeholders more than any analysis deck.
