UX Researcher

Purpose

Products are built on assumptions about what people want, how they think, and why they behave — and most of those assumptions are wrong in ways the team cannot see from the inside. A UX researcher exists to replace guesses with evidence: to find out what real users actually need, do, and struggle with, and feed that truth back into decisions before the team spends months building the wrong thing. The discipline exists because builders fall in love with their own ideas, and someone has to systematically check those ideas against reality.

Core Mission

Reduce the team's uncertainty about users to the point where the next decision becomes obvious — and defend the gap between what people say and what they actually do.

Primary Responsibilities

The visible work is running studies — interviews, usability tests, surveys — but the actual work is shaping which questions get asked and making sure the answers change what gets built. A UX researcher scopes research questions with stakeholders, picks the method that fits the question and the timeline, recruits and screens the right participants, moderates without leading, synthesizes messy qualitative data into patterns the team can act on, quantifies behavior when numbers are warranted, and translates it into findings that land. They also build research operations — panels, repositories, consent and incentive workflows — so insight compounds rather than evaporating. Underneath it all is advocacy: keeping the user a real presence in rooms full of business and engineering pressure.

Guiding Principles

• What people say is not what people do. Self-report is a hypothesis, not evidence. Trust observed behavior over stated preference, and design studies to catch the difference.
• The question dictates the method, never the reverse. Pick the tool for the decision at hand; don't run a survey because surveys are easy or interviews because you like talking to people.
• Insight that doesn't change a decision is a hobby. Research earns its keep by altering what the team does next. Tie every study to a pending decision.
• Bias is the default; rigor is the intervention. Leading questions, confirmation bias, and the order of your prompts contaminate data silently. Assume you are biasing the result and design against it.
• Small n, deep truth. Five users surface most usability problems; you don't need a thousand to learn that the button is invisible. Match sample size to the claim you want to make.
• Generative before evaluative. Understand the problem space before you test a solution; testing the wrong solution flawlessly is still wrong.
• Triangulate. One source lies; three that agree are hard to dismiss. Combine qual, quant, and behavioral data.

Mental Models

• Behavioral vs. attitudinal / qualitative vs. quantitative (the Rohrer landscape). Every method sits on two axes — what people do vs. what they say, and how many vs. why. Knowing where a method falls tells you what it can and cannot answer.
• Generative vs. evaluative research. Generative work discovers needs and frames problems ("what should we build?"); evaluative work tests a specific design ("does this work?"). Confusing the two wastes everyone's time.
• Jobs To Be Done. People "hire" a product to make progress in a situation. Frame needs as jobs and outcomes, not features, to escape the trap of asking users to design.
• The funnel of distortion. Between what a user experiences and what lands in your report sit memory, social desirability, your phrasing, your notes, and your synthesis. Each stage loses and warps signal; design to minimize each.
• Saturation. New interviews stop producing new themes after a point; saturation, not a magic number, tells you when qualitative sampling is enough.
• Signal vs. noise in small samples. With n=5 you find problems, not rates. Resist the urge to put percentages on tiny qualitative samples.

First Principles

• You are not your user, and neither is anyone in the building.
• The presence of an observer changes the behavior observed.
• Absence of complaints is not evidence of usability; users blame themselves.
• A finding is only as good as the question that produced it.
• Numbers feel objective and are just as easy to mislead with as anecdotes.

Questions Experts Constantly Ask

• What decision will this research inform, and who is waiting on it?
• Is this a "why" question or a "how many" question?
• What would change our minds, and have we made it possible to be surprised?
• Am I leading the witness? Would a skeptic phrase this prompt differently?
• Are these the right participants, or just the easy ones to recruit?
• Are we hearing what people say, or watching what they do?
• What's the cheapest study that could resolve this uncertainty?

Decision Frameworks

• Method selection grid. Map the question onto behavioral/attitudinal and qual/quant axes: discovery → interviews and field studies; "is it usable?" → moderated usability test; "how many / which is better?" → survey or A/B test; "what do they actually do?" → analytics and behavioral logs.
• Confidence vs. speed. A rigorous longitudinal study and a two-day guerrilla test answer different stakes. Match rigor to the cost of being wrong.
• Sample sizing. Usability discovery: 5 per distinct user group. Statistical comparison: power-analyze for the effect you care about — usually hundreds. Don't claim significance you didn't power for.
• Build vs. buy the panel. Recruit via a vendor for speed and reach; build an in-house panel for repeat, high-trust, domain-specific access.

Workflow

1. Intake. Pin down the real question behind the request and the decision it feeds. "Stakeholders want a survey" is not a research question.
2. Frame. Write research questions and hypotheses; state what you'd expect to see if each were true or false.
3. Choose method and sample. Select the lightest method that answers the question; define screening criteria and target n.
4. Recruit and screen. Source participants who match the real users, not convenient proxies; screen out professional testers and obvious mismatches.
5. Instrument. Write the discussion guide or survey and pilot it on one person to catch leading and confusing items before they cost you the whole study.
6. Collect. Moderate neutrally; shut up and watch. Use think-aloud; resist rescuing struggling users.
7. Synthesize. Affinity-map and tag observations, find patterns, separate observation from interpretation, quote real users.
8. Report. Deliver findings tied to decisions, ranked by severity and confidence; make stakeholders watch real clips.
9. Archive. Put it in the repository, tagged and searchable, so the next team doesn't re-run your study.

Common Tradeoffs

• Speed vs. rigor. A clean experiment takes weeks; the roadmap meeting is Thursday. Decide how much confidence the decision actually requires.
• Depth vs. breadth. Rich interviews explain why but can't tell you how common; surveys size the problem but can't explain it.
• Internal validity vs. realism. Lab control kills confounds but breeds artificial behavior; field studies are messy but real.
• Democratizing research vs. quality control. Letting PMs and designers run their own studies scales insight but risks leading questions and bad samples.
• Telling stakeholders what they need vs. what they want to hear. The useful finding is often the unwelcome one.

Rules of Thumb

• If your question can be answered "yes" or "no," it's probably leading.
• Five participants per user type catches ~85% of usability issues.
• Never report a percentage on a sample under ~30.
• Pilot every guide and survey on at least one human first.
• Watch the hands and the face, not just the words.
• Silence is a tool; count to five before filling it.
• "Show me the last time you did this" beats "how often do you do this?"
• One vivid clip moves a room more than ten slides of charts.

Failure Modes

• Leading the witness. Phrasing, tone, or visible hope steering participants toward the answer you wanted.
• Recruiting the convenient. Testing with colleagues, friends, or professional panelists who don't resemble real users.
• Confirmation-driven synthesis. Cherry-picking quotes that fit the team's existing belief and burying the ones that don't.
• Research theater. Running the study after the decision is made, to launder a predetermined choice.
• Quant cosplay. Slapping percentages on n=8 to look scientific.
• Insight graveyard. Beautiful reports nobody reads; findings that never reach a decision.
• Solving the said problem. Building exactly what users asked for instead of what would solve their underlying job.

Anti-patterns

• The double-barreled question — "How easy and enjoyable was this?" measures nothing cleanly.
• The satisfaction survey as usability test — happy ratings while the task failed.
• Recruiting "users like me." Sampling for comfort instead of representation.
• The 90-minute interview with no guide — undisciplined chats that yield unanalyzable mush.
• Treating the focus group as truth — groupthink and loud voices drowning the signal.
• Over-generalizing from a single memorable participant.

Vocabulary

• Generative research — exploratory work to discover needs and frame problems before a solution exists.
• Evaluative research — testing a specific design against tasks or criteria.
• Think-aloud protocol — having participants narrate their thoughts while performing tasks.
• Saturation — the point where additional data yields no new themes.
• Triangulation — corroborating a finding across multiple methods or sources.
• Affinity mapping — clustering raw observations into emergent themes.
• Social desirability bias — participants answering to look good rather than honestly.
• System Usability Scale (SUS) — a standardized 10-item usability questionnaire scored 0–100.
• Top task — the small set of things users overwhelmingly come to do.
• Research ops — the infrastructure (panels, repos, consent, incentives) that makes research repeatable.

Tools

• Interview and usability platforms — Lookback, UserTesting, Maze, dscout for moderated and unmoderated sessions and diary studies.
• Survey tools — Qualtrics, Typeform, SurveyMonkey, for attitudinal data at scale.
• Analytics and behavioral — Amplitude, Mixpanel, Hotjar, FullStory, for what users actually do.
• Synthesis and repository — Dovetail, Aurelius, Miro/FigJam affinity boards, for tagging and searchable insight.
• Recruiting/panel — User Interviews, Respondent, in-house panels, for sourcing the right participants.
• Stats — significance testing, power analysis, and confidence intervals when claims go quantitative.

Collaboration

A UX researcher sits at a crossroads. They work with product managers (who own scope and often want the answer yesterday), UX and product designers (who own the solution and need actionable findings fast), data scientists (who own the behavioral and experimental numbers and make natural triangulation partners), engineers (who reveal what's feasible to test), and marketing (who hold their own customer signal). The recurring tension is the timeline: research is most valuable upstream, before commitment, but teams reach for it downstream, after. Good researchers embed early, teach stakeholders to run lightweight studies themselves (democratization), and stay the steward of quality so scaled research doesn't become scaled bias.

Ethics

Research puts a person's attention, data, and sometimes vulnerability in the researcher's hands. Core duties: obtain genuine informed consent, including how recordings will be used and stored; protect privacy and anonymize by default; compensate fairly for time; and never deceive beyond what a debrief can repair. Researchers must guard against extractive practices — mining users for ideas without regard for their wellbeing — and against research that exists to justify dark patterns or manipulative engagement loops. There is a duty to report inconvenient findings honestly rather than soften them for stakeholders, and to represent participants — especially marginalized or excluded ones — fairly in rooms where they aren't present. The researcher is often the only person in the room accountable to the human on the other side of the screen.

Scenarios

A PM demands a survey to settle a feature debate. Two teams disagree about whether users want a new dashboard; the PM wants a survey by Friday to "let the data decide." The expert pushes back: a survey reports what people say they'd want, famously unreliable for unbuilt features — users over-claim interest in everything. The real question is whether the current workflow has a painful gap. The researcher reframes it as a generative question, runs six 45-minute interviews watching people do the actual task, and finds the dashboard solves a problem only one segment has. The survey would have produced a tidy 68% "yes" and shipped the wrong thing. Method chosen by question, not by deadline.

A usability test where everyone "succeeds" but the product is failing. Logs show users abandoning checkout, yet a prior test reported high satisfaction. The researcher re-runs it with two changes: drop the moderator's habit of nudging stuck users ("try the top right") and add think-aloud. Now three of five users hesitate 20+ seconds at the shipping step, mutter "wait, is this final?", and one nearly closes the tab. The earlier test had a leading moderator who rescued people, masking the real friction. The fix — a clearer confirmation step — comes from neutral observation, not the satisfaction score that lied.

Synthesizing contradictory signals before launch. Interviews say users love a feature; analytics show almost nobody uses it. Rather than pick the flattering story, the researcher triangulates: the interviewees were recruited from power users (a sampling bias), while analytics covers everyone. Both are "true" for different populations. The honest finding — "beloved by a small, vocal segment; irrelevant to the mainstream" — reframes the decision from "launch or kill" to "keep as an advanced option, don't put it in onboarding." The value was in resisting the clean narrative.

Related Occupations

A UX researcher works closest to the UX designer, who turns insight into form; the two are constant partners but think differently — one converges on truth, the other diverges into solutions. Product managers share the obsession with user value but trade off against business and delivery constraints. Data scientists are the quantitative twin — same questions, statistical tools instead of observational ones, the ideal triangulation partner. Graphic designers and technical writers consume research to shape communication; marketing managers hold an adjacent, sales-angled view of the same customers.

References

• Just Enough Research — Erika Hall
• Rocket Surgery Made Easy — Steve Krug
• Interviewing Users — Steve Portigal
• Observing the User Experience — Goodman, Kuniavsky & Moed
• Quantifying the User Experience — Sauro & Lewis
• Nielsen Norman Group articles — nngroup.com