UX Research Repository Examples: Structures, Schemas, and Reusable Evidence

The best UX research repository examples are not giant folders of reports. They are small, deliberate information systems that preserve the chain from a research question to evidence, interpretation, and a decision. The right example depends on the job: a solo researcher may need a study index, while a product organization needs linked evidence and decision records.

This article provides copyable structures and schemas. It is deliberately different from our broader UX research repository guide, which explains governance, launch, and maintenance. Here, the focus is the actual shape of records: what tables to create, what fields to include, and how those records connect.

Example 1: a simple study index for a small team

Start with one table when the repository has fewer than a few dozen studies and one person can maintain it. Each row represents a completed or active study; the full report and approved materials sit behind that row.

This is better than a folder because users can filter by product area, method, participant group, and date. It remains intentionally study-centric. Do not cram every quotation and implication into one long cell.

Example 2: linked studies, evidence, insights, and decisions

For repeated reuse, create four linked record types. A study preserves method and scope. An evidence item contains an approved observation, excerpt, or artifact reference. An insight interprets one or more evidence items. A decision records what the team chose and which insights it considered.

The relationship is:

Study → approved evidence → qualified insight → decision or experiment

One study may produce several insights. One insight may draw on several studies. A decision may consider supporting and contradictory insights. This structure prevents a polished summary from being mistaken for raw evidence and makes “research says” traceable.

Use a stable ID for every record, such as STU-024, EVD-118, INS-047, and DEC-031. IDs survive title edits and are easier to reference in planning documents.

Example 3: an atomic insight card

An atomic insight should communicate one idea without losing its boundary. Copy this schema:

Insight ID: INS-[number]
Statement: [One qualified sentence]
Observed behavior: [What happened, without interpretation]
Interpretation: [What the pattern may mean]
Population and context: [Participants, market, journey, device]
Supporting evidence: [EVD links]
Contradictory evidence: [EVD links or “not assessed”]
Confidence: [Low / medium / high, with reason]
Limitations: [Recruitment, method, gaps]
Implication: [A question or option, not an order]
Owner and review date: [Name; date]
Status: [Provisional / active / superseded / archived]

Write “Four of eight recruited participants using older Android devices could not find the edit control” rather than “Users cannot edit.” The first statement preserves sample and context. The second invents universality.

Example 4: a research question and evidence matrix

A matrix works when several studies address one strategic question. Rows represent evidence sources; columns represent propositions the team needs to assess.

The matrix makes disagreement visible. It is not a voting machine: three weak sources do not automatically outweigh one strong source. Researchers still evaluate method, relevance, recency, and independence.

For cleaner descriptive writing, use the distinction in our objective summary guide between what a source contains and what the reviewer concludes.

Example 5: a continuous-discovery repository

Continuous interviews need a lighter intake path than quarterly studies. Use one session record per conversation and publish only reviewed evidence into the durable repository.

Recommended session fields are date, interviewer, participant code, recruitment segment, topic, consent and reuse boundary, source location, review status, and linked candidate evidence. Keep participant identity separate from the broadly searchable record.

The workflow is: capture authorized notes, review the source, redact unnecessary identifiers, create evidence items, draft insights, and approve them with a researcher. Do not allow an AI summary to publish directly as organizational knowledge. Our voice recorder for interviews guide covers practical capture choices, while the interview transcription workflow explains how source audio becomes reviewable text.

Capture consented in-person research conversations without losing the room context. Kuno is designed for authorized face-to-face conversation capture and helps turn recordings into notes that researchers can review before repository publication. Explore Kuno

Example 6: a decision-linked repository

A repository creates more value when evidence is connected to product choices. Add this compact section to every decision record:

Decision: [What was chosen]
Owner and date: [Who; when]
Question: [What uncertainty was addressed]
Evidence considered: [INS and STU links]
Conflicting evidence: [Links and explanation]
Constraints: [Technical, commercial, accessibility, policy]
Review trigger: [New evidence, metric threshold, or date]
Outcome evidence: [Experiment or post-launch review]

Research informs a decision; it does not make the decision. Accessibility, engineering constraints, strategy, and customer commitments may also matter. Record them so later readers understand why the team did not simply follow one finding.

Turn agreed follow-up into owned work using a consistent meeting follow-up process rather than leaving actions inside a readout deck.

Example 7: a privacy-aware interview evidence record

Raw recordings and identifiable transcripts deserve tighter access than synthesized findings. A privacy-aware evidence record should include source type, participant code, approved excerpt, redaction status, permitted uses, access group, retention or review date, and the location of the controlled source.

Consent is not a universal reuse license. Permission to record one interview may not cover broad internal search, external quotation, model training, or indefinite retention. Laws and contractual duties vary, so obtain appropriate legal or privacy guidance for your context. Provide a genuine no-recording alternative where participation should not depend on capture.

For general recording considerations, see legally recording conversations. It is general information, not legal advice.

Choose the smallest schema that preserves trust

Use one study table if the primary job is finding prior reports. Add atomic insight records when findings need reuse across studies. Add evidence records when traceability matters. Add decision records when the organization wants to learn whether research affected outcomes.

Avoid mandatory metadata that nobody uses. A field earns its place when it improves retrieval, interpretation, access control, or maintenance. Pilot the schema against five real questions, such as “What do we know about first-time setup for administrators in France?” If the repository cannot answer without opening every report, refine the metadata.

Test retrieval with realistic searches

People use different words for the same concept: sign-up, registration, account creation, and onboarding. Maintain synonyms for high-value themes and test queries from designers, product managers, support staff, and researchers.

Every result preview should expose the insight, population, context, date, confidence, status, and owner. A result that shows only a catchy title invites decontextualized reuse. Track failed searches and requests answered by asking a colleague; both reveal gaps in vocabulary or trust.

Copyable repository launch checklist

• Choose two retrieval questions for the first release.
• Define study, evidence, insight, and decision as separate concepts.
• Create stable IDs and controlled values for core filters.
• Record participant scope, method, limitations, and freshness.
• Keep identity and raw media behind appropriate access controls.
• Link every published insight to approved evidence.
• Show contradictory evidence and superseded status.
• Assign an owner and review date.
• Test search using real team vocabulary.
• Review AI-drafted labels, summaries, and quotations against the source.

The repository is useful when another person can retrieve an insight, understand where it applies, inspect its evidence, and see what happened next. Volume alone is not success.

Keep human review between a recording and reusable evidence. Kuno can support consented in-person capture and draft notes, while researchers remain responsible for checking quotations, context, access, and interpretation. See Kuno for in-person research