What makes a good research question?

It is specific, open-ended, decision-relevant, and researchable through user contact. 'Where in onboarding do new users feel uncertain about whether to continue?' is far stronger than 'Do users like our onboarding?' A good question does not presuppose its answer ('Do users like feature X?' invites a yes), and the answer to it changes what gets built. Bad questions produce uninterpretable answers, which is why framing earns disproportionate time.

How many participants do I need?

For in-depth interviews, 5 to 15 depending on the question, run in 60-minute sessions. The recruit makes or breaks the study, so match the target audience rather than recruiting 'anyone willing', mix behaviors (active, lapsed, never-users), exclude friends, family, and employees, and pay participants (commonly $50 to $150 for 60 minutes, more for executives or specialized professions). Plan for a 20 to 30 percent no-show rate: recruit 7 to reliably run 5.

How do I run an interview without biasing it?

Open with rapport rather than the research questions, ask open-ended questions ('tell me about the last time...'), use silence and let participants fill it, ask for specifics and examples, and probe contradictions gently. Do not sell or convince; this is listening, not pitching. Avoid leading questions, hypotheticals (a poor predictor of real behavior), asking several questions at once, interrupting, and interviewing your own hypothesis by only asking what confirms the team's existing beliefs.

How do notes become insights?

Through synthesis. Capture observations (each a single quote, behavior, or moment), affinity-map them into themes, find the patterns that appear across multiple participants, and identify insights: non-obvious findings that explain a why and imply a so-what. Test each insight against the data, because one that only fits some interviews is a hypothesis, not an insight, and distinguish signal from noise (a belief 6 of 8 participants hold is signal; one 1 of 8 holds may be noise). The strongest insights surprise the team and hold across data points.

When should I use a different skill?

UX research is generative and discovery work, answering what is true. Use usability-testing to test a specific design or prototype, journey-mapping to map the full end-to-end experience, analytics-strategy for quantitative measurement, and cro-optimization for conversion testing in production. Research and these methods compose: research surfaces the questions, and the others measure or test specific answers.

Skill · UX research

UX research.

Research that ends in decisions, not decks.

Plan and run generative and discovery research that produces decisions. Six phases carry it: frame the question, choose the method, recruit, conduct, synthesize, and communicate. The framing phase earns disproportionate time, because a bad question produces an uninterpretable answer.

This skill is for understanding what is true. Testing a specific design, mapping the full journey, and measuring quantitatively are separate jobs with their own skills.

Audience: researchers, designers, and PMs starting a product without enough user understanding, diagnosing an unclear problem, or building empathy on a team that has drifted from its users.

View the skill on GitHub Browse the full catalog

The framework

Six phases from question to communicated finding.

The method follows the question, and the synthesis turns sessions into insights. Each phase feeds the next.

01Frame the question: specific, open-ended, decision-relevant, and answerable through user contact. Spend disproportionate time here; bad questions produce bad research.
02Choose the method: generative methods (interviews, contextual inquiry, diary studies) for what is true, validation methods (concept testing, card sorts, tree tests) for whether a hypothesis holds.
03Recruit: match the target audience, mix behaviors, exclude friends and employees, and pay participants. Plan for 20 to 30 percent no-show.
04Conduct: open with rapport, use open-ended questions and silence, ask for specifics, and probe contradictions gently, without selling or leading.
05Synthesize: capture observations, affinity-map into themes, find the patterns that hold across participants, and distinguish signal from noise.
06Communicate: top-line insights, highlight reels, and in-room workshops, framed per stakeholder, so the findings change decisions rather than sitting in a deck.

What separates research from theater

If no decision changed, the research failed.

Research starts from a decision, not a curiosity. Before recruiting anyone, name the decision the findings will inform and who owns it. A study with no decision attached produces findings that have no home and effort that is wasted, however polished the deck.

The interview is listening, not pitching, and the fastest way to ruin it is to interview your own hypothesis: only asking questions that confirm what the team already believes. Leading questions ('do you not find this confusing?') and hypotheticals ('would you use a feature that...') make the findings reflect the researcher rather than the user. Suspect insights that simply confirm what the team already thought.

Notes do not become insights on their own. An insight is more than a theme: it is a non-obvious finding that explains a why, implies a so-what, holds across multiple data points, and can be said in a sentence or two. And findings die in slide decks, so plan distribution in formats people actually absorb, then track whether decisions changed. If they did not, the research failed regardless of its quality.

Reference files

The reference that goes alongside the SKILL.md.

references/interview-guide-template.md
A structured interview guide template with example openings, probes, and closes.

Browse all reference files on GitHub

Bridges to other skills

Where research hands off.

Generative research is one mode. These cover testing a design, mapping the experience, and the quantitative side.

Test a design
usability-testing
Testing a specific design or prototype for task completion is its own method. UX research answers what is true before there is a design to test.
Map the experience
journey-mapping
Mapping the full end-to-end experience across touchpoints happens there. Research feeds it the user understanding that keeps the map honest.
The quantitative side
analytics-strategy
Quantitative measurement is a different instrument. Research explains the why behind the numbers that analytics surfaces.
Conversion testing
cro-optimization
Conversion testing in production answers a narrower question. Reach for it to optimize a flow; reach for research to understand the people in it.

Open source under MIT

Read the SKILL.md on GitHub.

The skill source lives in the rampstackco/claude-skills repository alongside dozens of other skills covering the full lifecycle of brand and product work. This page is a structured overview; the SKILL.md is the source. MIT licensed.

View SKILL.md Browse the full catalog

Frequently asked questions.

What makes a good research question?: It is specific, open-ended, decision-relevant, and researchable through user contact. 'Where in onboarding do new users feel uncertain about whether to continue?' is far stronger than 'Do users like our onboarding?' A good question does not presuppose its answer ('Do users like feature X?' invites a yes), and the answer to it changes what gets built. Bad questions produce uninterpretable answers, which is why framing earns disproportionate time.
How many participants do I need?: For in-depth interviews, 5 to 15 depending on the question, run in 60-minute sessions. The recruit makes or breaks the study, so match the target audience rather than recruiting 'anyone willing', mix behaviors (active, lapsed, never-users), exclude friends, family, and employees, and pay participants (commonly $50 to $150 for 60 minutes, more for executives or specialized professions). Plan for a 20 to 30 percent no-show rate: recruit 7 to reliably run 5.
How do I run an interview without biasing it?: Open with rapport rather than the research questions, ask open-ended questions ('tell me about the last time...'), use silence and let participants fill it, ask for specifics and examples, and probe contradictions gently. Do not sell or convince; this is listening, not pitching. Avoid leading questions, hypotheticals (a poor predictor of real behavior), asking several questions at once, interrupting, and interviewing your own hypothesis by only asking what confirms the team's existing beliefs.
How do notes become insights?: Through synthesis. Capture observations (each a single quote, behavior, or moment), affinity-map them into themes, find the patterns that appear across multiple participants, and identify insights: non-obvious findings that explain a why and imply a so-what. Test each insight against the data, because one that only fits some interviews is a hypothesis, not an insight, and distinguish signal from noise (a belief 6 of 8 participants hold is signal; one 1 of 8 holds may be noise). The strongest insights surprise the team and hold across data points.
When should I use a different skill?: UX research is generative and discovery work, answering what is true. Use usability-testing to test a specific design or prototype, journey-mapping to map the full end-to-end experience, analytics-strategy for quantitative measurement, and cro-optimization for conversion testing in production. Research and these methods compose: research surfaces the questions, and the others measure or test specific answers.