Conduct user research

Conduct user research

Turn a product question into credible evidence and an explicit decision. Treat research as inquiry, not as validation for a preferred solution.

Establish the decision

1. Read project instructions and existing research, analytics, support evidence, product constraints, and prior decisions.
2. State the decision the research will inform, what is uncertain, what is already known, and what would change the team's direction.
3. Separate behavioral questions from attitudinal questions. Prefer observed behavior for workflow and usability claims.
4. Identify affected groups, including infrequent users, assisted users, administrators, people using assistive technology, and people in constrained environments when relevant.
5. Record assumptions and conflicts in the available evidence. Do not present stakeholder belief as user evidence.

Choose the smallest credible study

Read references/methods-and-synthesis.md when selecting a method, writing a discussion guide, or choosing a synthesis approach.

Define:

• research objective and questions;
• participant characteristics, exclusions, and recruitment rationale;
• method, sample rationale, session format, and stimulus fidelity;
• consent, recording, privacy, compensation, and data-retention plan;
• evidence to capture and the decision rule for stopping or following up;
• known limitations and likely sources of bias.

Use assets/research-plan-template.md for work that spans multiple sessions or contributors.

Do not collect sensitive data merely because it may be useful. Minimize personal data, obtain meaningful consent, and follow the organization's approved research and privacy practices.

Prepare neutral materials

1. Start the guide with context and recent real behavior before opinions about a solution.
2. Write one question at a time. Remove leading language, implied praise, and forced agreement.
3. For evaluative research, give participants a goal and realistic context rather than instructions that reveal the interface.
4. Define prompts for follow-up without scripting the participant's answer.
5. Pilot the session. Fix confusing tasks, broken stimuli, unrealistic data, and timing problems.

Conduct and capture

• Establish consent and the participant's right to pause or stop.
• Ask for concrete examples: what happened, in what order, with what information, and with what consequence.
• Observe workarounds, hesitation, recovery, environmental constraints, and discrepancies between words and actions.
• Avoid teaching, defending, or selling the design during the task.
• Distinguish observation, participant statement, moderator interpretation, and product implication in notes.
• Remove unnecessary identifying information from retained evidence.

If direct participant contact is unavailable, synthesize existing evidence and label it secondary research; do not simulate interviews or invent quotations.

Synthesize without erasing disagreement

1. Consolidate evidence by participant and research question.
2. Code recurring behaviors, needs, obstacles, triggers, workarounds, and outcomes.
3. Preserve meaningful outliers and differences between segments instead of forcing consensus.
4. Triangulate findings against product data, support evidence, and prior research where available.
5. Rate confidence from evidence quality, consistency, recency, and relevance—not from the number of sticky notes.
6. Convert patterns into findings that state the evidence, affected context, user consequence, and product implication.
7. Keep recommendations separate from findings so alternative responses remain visible.

Make the research actionable

Deliver a decision brief containing:

• question, method, participants, and limitations;
• key findings with evidence and confidence;
• important segment differences and unresolved contradictions;
• implications, risks, and opportunity areas;
• recommendation or decision supported by the evidence;
• open questions and the next smallest useful study.

Use shape-product-opportunity when research must be converted into opportunity areas, assumptions, and experiments. Let design-product-interface or design-and-build-website own design and implementation.

Apply quality rules

• Report who was studied and who was not.
• Do not turn small qualitative samples into population percentages.
• Do not claim causality from descriptive or correlational evidence.
• Do not use demographic traits as a substitute for behavior, need, or context.
• Preserve uncertainty and negative evidence.
• Never fabricate participants, sessions, quotes, observations, or research results.