Purpose
Turn a pile of raw research and behavioral data into a defensible, prioritized set of UX action items — each linked to its supporting evidence, scored for impact and effort, and written as a problem statement a design team can act on immediately. Solves the gap between "we ran the research" and "we know what to build first."
When to Use
Invoke this skill when the user asks to:
- Turn usability test notes or interview transcripts into action items
- Prioritize a backlog of UX issues from research findings
- Convert survey results or NPS verbatims into design decisions
- Translate analytics or funnel drop-off data into UX fixes
- Synthesize support ticket themes into product improvements
- Build a research-to-roadmap handoff for stakeholders or PMs
Inputs Required
Before starting, collect:
- Raw findings: The research or data — interview quotes, test observations, survey results, analytics numbers, ticket themes (paste as-is)
- Product area: The feature, flow, or surface these findings relate to
- Research method: How the data was gathered (moderated test, survey, analytics, support logs, etc.) and rough sample size
- Primary business goal: What outcome the team is optimizing for (activation, retention, conversion, task success, etc.)
- Team capacity: Rough sense of engineering/design bandwidth (small / medium / large) to calibrate effort realism
- Known constraints: Tech limitations, brand rules, or fixed decisions that bound the solution space
If business goal or capacity is missing, default to task success as the goal and medium capacity, and note the assumption.
Instructions
Step 1: Extract and Cluster Raw Insights
Read every finding and pull discrete observations. Do not editorialize yet — capture what was actually said, done, or measured.
- Tag each observation with its evidence type: Behavioral (what users did), Attitudinal (what users said), or Quantitative (what the data measured)
- Note the frequency signal: how many participants/sessions/tickets exhibited it (e.g., "4 of 6 testers," "22% drop-off," "recurring in ~30 tickets")
- Cluster observations that point to the same underlying problem into themes — aim for 4–8 themes, not one per quote
Step 2: Separate Symptom From Root Problem
For each theme, distinguish what the user experienced from what's actually broken.
- Symptom: The observable friction ("users hesitated at the payment step")
- Root problem: The underlying design cause ("payment page introduces three new fields with no explanation of why they're needed")
- Flag any theme where the root cause is still a hypothesis vs. confirmed by evidence — these may need more research before action
Step 3: Write a Problem Statement Per Theme
Convert each theme into an actionable, solution-neutral problem statement using the format:
[User type] is trying to [goal], but [obstacle], which causes [consequence]. Evidence: [frequency + method].
Keep it solution-neutral so designers explore options rather than anchoring on the first fix.
Step 4: Score Each Item on Impact and Effort
Score every problem statement on two axes, 1–5.
Impact — weighted by:
- Frequency signal (how many users are affected)
- Severity (blocker vs. annoyance vs. polish)
- Alignment to the stated business goal
- Position in the funnel (upstream problems compound downstream)
Effort — weighted by:
- Design complexity, engineering lift, and cross-team dependencies
- Whether a fix is copy/UI-level (low) or flow/architecture-level (high)
- Calibrate against the team capacity provided
Step 5: Assign Priority Tiers
Map the impact/effort scores into four tiers:
- P0 — Do Now: High impact, low-to-medium effort. Ship this sprint.
- P1 — Plan Next: High impact, high effort. Needs scoping but clearly worth it.
- P2 — Quick Wins: Low-to-medium impact, low effort. Batch these opportunistically.
- P3 — Backlog/Watch: Low impact or unconfirmed root cause. Revisit or research further.
Within each tier, order by impact score descending.
Step 6: Recommend a First Move Per P0 and P1
For each P0 and P1 item, add one concrete next action — a specific design direction, a test to validate the fix, or a decision the team must make. Be opinionated; this is the handoff.
Output Format
- Structure: Executive summary → themed insight table → prioritized action list by tier → next-move recommendations
- Length: 800–1,200 words
- Format: Structured document with one summary table (themes, evidence type, frequency, impact, effort, tier) and grouped action lists
- Sections: Summary of findings → Insight-to-theme table → P0/P1/P2/P3 action lists (with problem statements) → Recommended first moves → Open questions / research gaps
- Tone: UX lead briefing a product team — evidence-anchored, decisive, no hedging on priority calls
Quality Checklist
Before delivering output, verify:
- [ ] Every action item traces back to a specific piece of evidence with a frequency signal — no unsupported opinions
- [ ] All 4–8 themes have a solution-neutral problem statement in the required format
- [ ] Every item has both an impact score and an effort score (1–5)
- [ ] Items are sorted into exactly four priority tiers, ordered by impact within each
- [ ] Every P0 and P1 has a concrete recommended first move
- [ ] Themes with unconfirmed root causes are flagged as research gaps, not shipped as fixes
Notes
- Behavioral evidence (what users did) should outweigh attitudinal evidence (what users said) when the two conflict — stated preference is often unreliable.
- Resist collapsing every finding into a P0; a backlog where everything is urgent is unprioritized. The tiering is only useful if it's honest.
- When quantitative and qualitative findings agree on the same theme, note it — that convergence is the strongest signal in the report and belongs at the top of the summary.
- If findings are thin or contradictory, the highest-value output may be a recommended follow-up study, not a fix. Say so.
How to use this skill
This file is a Claude Code skill in Anthropic's standard format — markdown with YAML frontmatter at the top.
Installing in Claude Code:
~/.claude/skills/<folder-name>/SKILL.md— auto-loads in every session anywhere on your machine (recommended for skills you'll use across multiple projects)<your-project>/.claude/skills/<folder-name>/SKILL.md— auto-loads only when working in that project (recommended for project-specific skills)
Restart Claude Code or open a new session to register.
Using in Claude.ai, ChatGPT, Gemini, or other AI tools:
- Drag this file into a new chat, or copy and paste its contents
- Send your task or request as your next message — the AI will apply the skill where it fits
For Claude Code skill details: https://docs.anthropic.com/en/docs/claude-code/skills