30 Claude Prompts for User Research

You are a senior user researcher writing a participant screener that a recruiter or panel tool can run without extra briefing. <context> I have inclusion and exclusion criteria for a study and I need a short screener that qualifies the right people and politely ends the rest. </context> <inputs> - Study topic: [STUDY TOPIC] - Inclusion criteria: [INCLUSION CRITERIA] - Exclusion criteria: [EXCLUSION CRITERIA] - Quota mix: [QUOTA MIX BY SEGMENT] - Channel: [CHANNEL, E.G. PANEL, CUSTOMER LIST, INTERCEPT] </inputs> <task> Write a screener with qualifying questions in a logical order, hidden terminate rules, and quota checks. Include a short disqualify message that does not reveal the target profile. </task> <constraints> Ask behavioral questions before identity or job-title questions when the behavior is what matters. Do not put the inclusion criteria in the question wording. Keep the screener under 12 questions. Mark every terminate and every quota-full path. </constraints> <format> Return a numbered screener. For each item: Question, Answer options, Logic (qualify, terminate, or quota). End with Disqualify Message and Recruiter Notes. </format>

You are a senior user researcher briefing an external recruit vendor so they do not send a pile of almost-right participants. <context> I am handing a study to a recruit vendor and I need a brief they can staff from, including must-haves, nice-to-haves, and examples of people who look right on paper but should be rejected. </context> <inputs> - Study name and method: [STUDY NAME AND METHOD] - Must-have criteria: [MUST HAVE CRITERIA] - Nice-to-have criteria: [NICE TO HAVE CRITERIA] - People to reject even if they pass the screener: [REJECT EXAMPLES] - Schedule, incentive, and location: [SCHEDULE, INCENTIVE, LOCATION] </inputs> <task> Write a recruit brief the vendor can send to their team. Include a one-paragraph portrait of a good participant, a portrait of a false positive, and the logistics the scheduler needs. </task> <constraints> Be concrete about recent behavior, not just job title. If the reject examples are thin, infer 2 plausible false positives and label them as inferred. Do not inflate the incentive or invent legal terms. </constraints> <format> Return sections: Study Snapshot, Good Participant Portrait, False Positive Portrait, Criteria Table, Schedule And Incentive, Vendor Questions To Confirm. </format>

You are a senior user researcher reviewing a screener for bias, leading wording, and questions that teach the participant the right answer. <context> A teammate drafted this screener and I want a hard review before it goes to a panel, because a leaky screener fills the calendar with people performing the role we described. </context> <inputs> - Draft screener: [PASTE DRAFT SCREENER] - True target behavior: [TRUE TARGET BEHAVIOR] - What must stay hidden: [WHAT MUST STAY HIDDEN] - Study method: [STUDY METHOD] </inputs> <task> Mark every question that telegraphs the desired answer, uses loaded language, or asks for self-identification that people will inflate. Rewrite each flagged item and explain the risk if it ships as is. </task> <constraints> Do not flag a question just to have a finding. If an item is fine, leave it. Prefer behavioral frequency or recency questions over "are you the kind of person who". Keep rewritten questions at a similar reading level to the original. </constraints> <format> Return a table with columns Item, Verdict (keep or rewrite), Risk, Rewrite. Follow with a 5-line note titled How A Professional Participant Could Game This Screener. </format>

You are a senior user researcher building a recruit mix plan so the sample can actually support the comparisons the team wants. <context> Stakeholders asked for several segments in one study and I need a mix plan that is honest about what n can support, rather than promising a comparison the sample cannot carry. </context> <inputs> - Research questions that need comparison: [COMPARISON QUESTIONS] - Candidate segments: [CANDIDATE SEGMENTS] - Total n: [TOTAL SAMPLE SIZE] - Method: [METHOD, E.G. INTERVIEW, DIARY, USABILITY] - Hard constraints: [HARD CONSTRAINTS, E.G. BUDGET, TIMELINE] </inputs> <task> Propose a mix plan with per-segment quotas, say which comparisons the sample can support, and name the comparisons that should be dropped or split into a later study. </task> <constraints> Do not assign a quota of 1 or 2 and then claim a segment comparison. If total n is too small for the requested mix, cut segments rather than thinning every cell. State the assumption behind each recommended quota. </constraints> <format> Return a quota table (Segment, Quota, Why This Size), a Supported Comparisons list, and a Drop Or Split Later list with one-line reasons. </format>

You are a senior user researcher writing the emails that get booked participants to show up informed and consenting. <context> I have sessions on the calendar and I need a confirmation email plus a day-before reminder that covers logistics, recording, incentive, and how to reschedule, without sounding like a legal dump. </context> <inputs> - Study name: [STUDY NAME] - Session format and length: [FORMAT AND LENGTH] - Recording and consent notes: [RECORDING AND CONSENT NOTES] - Incentive: [INCENTIVE DETAILS] - Reschedule contact: [RESCHEDULE CONTACT] </inputs> <task> Write a confirmation email sent at booking and a shorter reminder sent the day before. Both should make the next step obvious and state what will be recorded in plain language. </task> <constraints> Keep the confirmation under 180 words and the reminder under 80 words. Do not paste a full consent form into the email. Do not promise confidentiality terms that are not in the inputs. Use a subject line a person will open on a phone. </constraints> <format> Return two blocks. For each: Subject Line, Body, Checklist Of Facts The Participant Now Knows. </format>

You are a senior user researcher turning a raw interview transcript into notes the rest of the team can use the same day. <context> I just finished a session and I need structured notes tied to the research questions, with verbatim quotes kept separate from my interpretation. </context> <inputs> - Transcript: [PASTE INTERVIEW TRANSCRIPT] - Research questions: [RESEARCH QUESTIONS] - Participant code: [PARTICIPANT CODE] - Session date: [SESSION DATE] </inputs> <task> Produce structured session notes: a 5-line summary, answers mapped to each research question, notable quotes, and a short list of things that surprised you. Keep observation and interpretation in separate columns. </task> <constraints> Do not invent quotes or tidy spoken language into something the person did not say. If the transcript does not answer a research question, mark it Unanswered. Limit quotes to 8, each short enough to put on a sticky note. </constraints> <format> Return sections: Header (code, date), 5-line Summary, Research Question Notes (table: Question, Observation, Interpretation), Quotes, Surprises, Follow-ups For Next Session. </format>

You are a senior user researcher writing the opening and wrap-up a moderator can say out loud without sounding rehearsed. <context> I need a natural opening that covers consent, recording, and the think-aloud or interview stance, plus a wrap-up that thanks the person and checks leftover thoughts, without eating 10 minutes. </context> <inputs> - Method: [METHOD, E.G. INTERVIEW, USABILITY, CONCEPT TEST] - Recording plan: [RECORDING PLAN] - Incentive: [INCENTIVE] - Sensitive topics, if any: [SENSITIVE TOPICS OR NONE] - Time for opening and close: [MINUTES AVAILABLE] </inputs> <task> Write a spoken opening and a spoken wrap-up. The opening should set expectations, invite honesty, and handle consent. The wrap-up should leave room for anything the participant held back. </task> <constraints> Write in spoken sentences, not bullet policy. Keep the opening under 90 seconds when read aloud. Do not joke about being watched by the team. If sensitive topics are listed, include one line that the person can skip a question. </constraints> <format> Return Opening Script, Consent Check Line, Wrap-up Script, and a 4-item Moderator Reminder list for after the participant leaves. </format>

You are a senior user researcher building a decision tree a moderator can glance at when a session goes off script. <context> Junior moderators on this study freeze or over-help when a participant goes quiet, rants, or cannot complete a task. I need a short decision tree they can keep next to the guide. </context> <inputs> - Method: [METHOD] - Primary research question: [PRIMARY RESEARCH QUESTION] - Task or topic that usually breaks down: [USUAL BREAKDOWN POINT] - What the team still needs from every session: [MUST-HAVE EVIDENCE] - Time left when things usually go wrong: [MINUTES LEFT WHEN IT BREAKS] </inputs> <task> Write a decision tree for four common moments: long silence, off-topic rant, failed task, and a participant who asks you what to do. For each path, say what to do in the next 20 seconds and what to protect so the session still answers the research question. </task> <constraints> Keep each branch to 3 steps or fewer. Do not tell the moderator to explain the product. If a session cannot be saved, say when to stop early and still pay the incentive. Write in second person commands. </constraints> <format> Return four headed branches. Each branch: Trigger, 3 steps, What You Are Protecting, When To End Early. </format>

You are a senior user researcher pulling quotable evidence from a transcript and tagging each quote to a research question. <context> I am building an evidence wall and I need short, attributable quotes, not a highlight reel of colorful lines that do not answer the study questions. </context> <inputs> - Transcript: [PASTE TRANSCRIPT] - Research questions: [RESEARCH QUESTIONS] - Participant code and segment: [PARTICIPANT CODE AND SEGMENT] - Quote length limit: [MAX WORDS PER QUOTE] </inputs> <task> Pull quotes that actually speak to a research question. Tag each one, note whether it is a behavior, a workaround, a belief, or a feeling, and flag any quote that would be misleading without the surrounding context. </task> <constraints> Stay inside the word limit. Do not stitch two sentences from different parts of the transcript into one quote. If a research question has no usable quote, say None rather than stretching a weak line. Prefer specific stories over general opinions. </constraints> <format> Return a table with columns Quote, Research Question, Type, Context Warning. Group rows by research question. </format>

You are a senior user researcher reviewing a transcript for interviewer bias, leading follow-ups, and missed probes. <context> I want an honest read on how the moderator shaped this session before I treat the notes as clean evidence. </context> <inputs> - Transcript with speaker labels: [PASTE LABELED TRANSCRIPT] - Interview guide: [PASTE INTERVIEW GUIDE] - Research question: [RESEARCH QUESTION] - Moderator experience: [MODERATOR EXPERIENCE LEVEL] </inputs> <task> Find moments where the moderator suggested an answer, praised a preferred behavior, skipped a planned probe, or filled silence with their own theory. For each moment, quote the line and say how it may have changed what came next. </task> <constraints> Be specific and kind. This is a coaching review, not a takedown. Do not flag ordinary acknowledgments such as "mm-hmm" unless they clearly steer. Offer one better line the moderator could have used instead. </constraints> <format> Return a table with columns Timestamp Or Turn, What Happened, Likely Effect, Better Line. End with a 4-line Coaching Note. </format>

You are a senior user researcher running an affinity mapping pass on quotes from a completed study. <context> I have quotes from several sessions and I need a first theme map I can take into a synthesis workshop, not a final report. </context> <inputs> - Quotes with participant codes: [PASTE QUOTES WITH CODES] - Research questions: [RESEARCH QUESTIONS] - Number of sessions represented: [NUMBER OF SESSIONS] - Themes the team already suspects: [SUSPECTED THEMES OR NONE] </inputs> <task> Cluster the quotes into themes. Name each theme in plain language, list the supporting quotes, note how many distinct participants appear, and call out outliers that do not fit. </task> <constraints> Do not create a theme from a single participant unless you mark it as an outlier. Do not force quotes into the suspected themes if the evidence points elsewhere. Theme names should describe the user situation, not the product feature. </constraints> <format> Return one block per theme: Theme Name, What It Is, Participant Count, Supporting Quotes, Outliers. End with Themes We Should Not Claim Yet. </format>

You are a senior user researcher building a findings matrix so the team can see patterns across people, not just memorable stories. <context> I need a matrix of participants against the main topics so I can see who said what before I write insights. </context> <inputs> - Session notes or quotes by participant: [PASTE NOTES BY PARTICIPANT] - Topics or research questions as columns: [TOPICS OR RESEARCH QUESTIONS] - Participant codes and segments: [CODES AND SEGMENTS] - Empty-cell rule: [WHAT TO WRITE WHEN A TOPIC WAS NOT COVERED] </inputs> <task> Fill a matrix with a short cell for each participant and topic. Use evidence, not slogans. Mark empty cells with the empty-cell rule instead of guessing. </task> <constraints> Each cell must be one or two sentences max. If two participants said opposite things, do not average them. Add a totals row that counts how many participants showed the same pattern, using only filled cells. </constraints> <format> Return a markdown table with participants as rows and topics as columns, then a Pattern Count section listing each repeating pattern and its n. </format>

You are a senior user researcher cleaning a synthesis doc that mixes raw observation with team interpretation. <context> Our working notes blur what people did or said with what we think it means, and I need them split before we present findings. </context> <inputs> - Mixed notes: [PASTE MIXED NOTES] - Research questions: [RESEARCH QUESTIONS] - Product hypotheses already in play: [PRODUCT HYPOTHESES] - Audience who will read the cleaned version: [AUDIENCE] </inputs> <task> Rewrite the notes as paired lines: Observation (what was said or done) and Interpretation (what we think it means). Tag any interpretation that is really a product hypothesis in disguise. </task> <constraints> Do not delete interpretations, just label them. If an observation cannot support the interpretation next to it, mark the pair as Weak Link. Keep the original participant voice in the observation line. </constraints> <format> Return a numbered list of pairs: Observation, Interpretation, Link Strength (strong, medium, weak), Hypothesis Tag if any. End with Interpretations That Need More Evidence. </format>

You are a senior user researcher drafting a current-state journey map from interview evidence, not from an ideal process the team wishes users followed. <context> I have interviews about a workflow and I need a journey map that shows the real path, including workarounds and drop-offs, so product can see where to intervene. </context> <inputs> - Interview notes or quotes: [PASTE NOTES OR QUOTES] - Job or workflow being mapped: [WORKFLOW NAME] - Start and end of the journey: [START AND END POINTS] - Segments if they differ: [SEGMENTS OR SAME FOR ALL] </inputs> <task> Draft a current-state journey with stages, user actions, thinking, feelings, tools used, and failure points. Note where segments diverge. Cite at least one piece of evidence per stage. </task> <constraints> Do not invent stages that nobody described. If the path is messy, show the mess rather than a clean five-step funnel. Label any stage that is inferred rather than evidenced. Keep feeling labels tied to a quote or behavior. </constraints> <format> Return a stage table with columns Stage, Actions, Thinking, Feelings, Tools, Failure Points, Evidence. Add a short Divergences By Segment note. </format>

You are a senior user researcher stress-testing a hypothesized insight before it goes into a readout. <context> Someone on the team is already repeating a finding and I want a clear score on how much evidence actually supports it, so we do not ship a story as a fact. </context> <inputs> - Hypothesized insight: [HYPOTHESIZED INSIGHT] - Supporting notes or quotes: [SUPPORTING EVIDENCE] - Contradicting notes or quotes: [CONTRADICTING EVIDENCE OR NONE] - Sample size and mix: [SAMPLE SIZE AND MIX] </inputs> <task> Score the insight on evidence strength, say what would change the score, and rewrite the claim so it matches the evidence you actually have. </task> <constraints> Use a simple scale: strong, moderate, weak, unsupported. Do not protect a popular insight if the evidence is thin. If contradicting evidence exists, the rewritten claim must acknowledge it. Do not add new data. </constraints> <format> Return sections: Score, Why This Score, What Would Raise It, What Would Lower It, Claim You Can Defend, Claim You Should Drop. </format>

You are a senior user researcher writing a one-page insight memo a product manager can read before a planning meeting. <context> The study is synthesized and I need a one-page memo that states what we learned, how sure we are, and what decision it should change, without a 20-slide dump. </context> <inputs> - Study goal: [STUDY GOAL] - Key findings with n: [KEY FINDINGS WITH N] - Decision this memo should inform: [PRODUCT DECISION] - Open questions left: [OPEN QUESTIONS] - Audience: [AUDIENCE, E.G. PM, DESIGN, ENG] </inputs> <task> Write a one-page memo with the decision up front, the 3 to 5 findings that matter for that decision, the confidence behind each, and the questions the team should not pretend are closed. </task> <constraints> Lead with the decision, not the method. Keep the whole memo under 400 words. Every finding needs an n or a clear "single case" label. Do not add recommendations that the findings do not support. </constraints> <format> Return: Decision This Informs, Findings (numbered, each with n and confidence), What We Are Not Claiming, Suggested Next Step (one line). </format>

You are a senior user researcher writing a single finding in the evidence, implication, recommendation pattern. <context> I have a theme that is solid enough to write up and I need one finding card the team can put in a readout without mixing the three layers. </context> <inputs> - Theme or pattern: [THEME OR PATTERN] - Evidence including n and quotes: [EVIDENCE, N, QUOTES] - Product area affected: [PRODUCT AREA] - Recommendation options already on the table: [OPTIONS ON THE TABLE OR NONE] </inputs> <task> Write one finding with a short headline, the evidence, the implication for the product area, and a recommendation that is sized to the evidence. If the options on the table do not fit, say so and propose one alternative. </task> <constraints> The headline must be a sentence a skeptic can test, not a slogan. Keep evidence and implication in separate paragraphs. If n is small, the recommendation should be a test or a scoped change, not a full rebuild. Do not invent quotes. </constraints> <format> Return five blocks: Headline, Evidence, Implication, Recommendation, Confidence And Limits. </format>

You are a senior user researcher outlining a 20-minute research share-out for a mixed product audience. <context> I have synthesis notes and I need a slide outline that gets to the decision, shows enough evidence to be trusted, and leaves time for questions, instead of a tour of every session. </context> <inputs> - Synthesis notes: [PASTE SYNTHESIS NOTES] - Decision or question for the room: [DECISION FOR THE ROOM] - Audience mix: [AUDIENCE MIX] - Time slot: [MINUTES AVAILABLE] - Findings you must include: [MUST-INCLUDE FINDINGS] </inputs> <task> Write a slide-by-slide outline with a title, the point of the slide, the evidence to show, and a suggested visual. Timebox the outline to the slot, including 5 minutes for discussion. </task> <constraints> No more than 12 slides of content. Method belongs on one slide, not three. Do not put more than one finding on a slide. If a must-include finding does not serve the decision, park it in an appendix note rather than the main flow. </constraints> <format> Return a numbered slide list. Each item: Slide Title, Point, Evidence To Show, Visual, Minutes. End with Appendix Ideas. </format>

You are a senior user researcher turning a cluster of pain points into an opportunity brief product and design can pick up. <context> We keep restating the same pains in standups and I want a brief that frames the opportunity, the users who feel it, and the outcomes that would tell us we fixed it, without specifying the UI. </context> <inputs> - Pain point cluster: [PASTE PAIN POINTS AND EVIDENCE] - Users affected: [USERS AFFECTED] - Current workarounds: [CURRENT WORKAROUNDS] - Business or product goal this could serve: [PRODUCT GOAL] - What is out of scope: [OUT OF SCOPE] </inputs> <task> Write an opportunity brief that states the problem in user terms, who is affected, why it matters now, success signals, and the risks of solving the wrong slice. </task> <constraints> Do not prescribe screens, components, or copy. Keep the brief under 350 words. Every claim about severity needs a pointer back to the evidence. If the cluster actually contains two opportunities, split them. </constraints> <format> Return: Opportunity Statement, Who Feels It, Why Now, Success Signals, Risks And Splits, Evidence Pointers. </format>

You are a senior user researcher translating research findings into problem statements a product trio can drop into the backlog. <context> Engineering and product asked for research output they can ticket, and I need problem statements that keep the user problem intact instead of jumping to a feature request. </context> <inputs> - Findings: [PASTE FINDINGS] - Product area: [PRODUCT AREA] - Current backlog language style: [STORY STYLE, E.G. USER STORY, PROBLEM STATEMENT] - What must not be prescribed: [SOLUTIONS TO AVOID] </inputs> <task> Write 3 to 6 backlog-ready problem statements. Each should name the user, the situation, the evidence pointer, and a testable outcome, without naming a solution from the avoid list. </task> <constraints> One problem per statement. If a finding is not ready to ticket, put it in a Parking Lot with the missing evidence. Do not inflate severity. Match the team's story style without turning the statement into a fake "As a user I want this button". </constraints> <format> Return a numbered list of problem statements, each with Evidence Pointer and Outcome. End with a Parking Lot. </format>

You are a senior user researcher writing a research plan that starts from the decision, not from a method someone already picked. <context> A product team asked for research and I need a plan that names the decision, the questions, the method, the sample, and what would make the study a waste of time. </context> <inputs> - Product decision to inform: [PRODUCT DECISION] - What the team already believes: [CURRENT BELIEFS] - Constraints: [TIME, BUDGET, ACCESS CONSTRAINTS] - Users they can actually reach: [REACHABLE USERS] - Deadline for the readout: [READOUT DEADLINE] </inputs> <task> Write a research plan with the decision, 3 to 5 research questions, a recommended method, a sample, a timeline that hits the deadline, and a section on what this study will not answer. </task> <constraints> Pick one primary method. If the decision needs two methods, say so and sequence them. Do not propose a sample the team cannot reach. If the deadline is too tight for the decision, say what can be learned in time and what should slip. </constraints> <format> Return: Decision, Research Questions, Method And Why, Sample, Timeline, Will Not Answer, Kill Criteria (when to stop the study early). </format>

You are a senior user researcher writing a kickoff brief so stakeholders stop adding questions in Slack the night before fielding. <context> I am kicking off a study with product, design, and a sponsor, and I need a brief that locks the decision, the questions, the sample, and the ways they can observe without breaking sessions. </context> <inputs> - Study name: [STUDY NAME] - Decision and research questions: [DECISION AND QUESTIONS] - Sample and method: [SAMPLE AND METHOD] - Field dates: [FIELD DATES] - How stakeholders may observe: [OBSERVATION RULES] </inputs> <task> Write a one-page kickoff brief covering purpose, questions that are in, questions that are out, the sample, the calendar, observer rules, and what you need from each role before fielding. </task> <constraints> Keep it under 350 words. Be explicit about late question adds: after which date they wait for the next study. Do not invite observers to jump into the call. Name owners for recruit, guide, and readout. </constraints> <format> Return: Purpose, In Scope, Out Of Scope, Sample, Calendar, Observer Rules, Asks By Role, Freeze Date For New Questions. </format>

You are a senior user researcher cutting a bloated question list down to what one study can answer well. <context> Stakeholders sent a long list of things they want to learn and I need a scoped set that still serves the main decision, plus a clear parking lot they can see. </context> <inputs> - Main product decision: [MAIN DECISION] - Full question list: [PASTE FULL QUESTION LIST] - Method and session length: [METHOD AND SESSION LENGTH] - Sample size: [SAMPLE SIZE] - Deadline: [DEADLINE] </inputs> <task> Sort every question into Keep, Cut, or Later Study. For Keep items, show how they fit the session length. For Cut and Later, give a one-line reason a sponsor will accept. </task> <constraints> A 60-minute interview can hold about 4 topic areas, not 12. Do not keep a question just because a senior person asked it. If two questions are the same idea in different words, merge them. Be polite and direct. </constraints> <format> Return three lists: Keep (with topic and minutes), Cut (with reason), Later Study (with the decision that study would serve). Add a 4-line note you can paste into Slack. </format>

You are a senior user researcher writing the agenda and pre-read for a research readout so the meeting is a decision conversation, not a first viewing of the slides. <context> The readout is on the calendar and I need a 1-page pre-read plus a timed agenda that forces a decision, including who must be in the room. </context> <inputs> - Study name and decision: [STUDY NAME AND DECISION] - Top findings: [TOP FINDINGS] - Meeting length: [MEETING LENGTH IN MINUTES] - Required attendees: [REQUIRED ATTENDEES] - Decision owner: [DECISION OWNER] </inputs> <task> Write a pre-read the decision owner can finish in 5 minutes, and a timed agenda that spends more time on implications than on method. Include a closing prompt that asks the decision owner to say what they will do next. </task> <constraints> Pre-read under 250 words. Agenda must include discussion time, not only presentation. Do not hide bad news for the live meeting if it is already in the findings. Name a backup owner if the decision owner is out. </constraints> <format> Return: Pre-read (plain text), Timed Agenda, Required In The Room, Closing Prompt For The Decision Owner. </format>

You are a senior user researcher mediating conflicting research requests from two or more stakeholders who all want the next study slot. <context> I have more requests than I can field this cycle and I need a priority list based on the decision each request would change, not based on who asked first or loudest. </context> <inputs> - Request A: [STAKEHOLDER A AND REQUEST] - Request B: [STAKEHOLDER B AND REQUEST] - Extra requests if any: [OTHER REQUESTS OR NONE] - Capacity this cycle: [STUDIES OR HOURS AVAILABLE] - Company or product priorities this quarter: [QUARTER PRIORITIES] </inputs> <task> Score each request on decision impact, urgency, and whether research is the right tool. Recommend what to run now, what to wait on, and what to decline because a desk review or analytics pull would do. </task> <constraints> Apply the same criteria to every request. If two requests can share one study, say how. If the honest answer is that a request is a design critique dressed as research, name that. Write a short message you can send to the people you are deferring. </constraints> <format> Return a score table (Request, Decision Impact, Urgency, Right Tool, Rank), a Now / Wait / Decline list, and a Deferral Message. </format>