> ## Documentation Index > Fetch the complete documentation index at: https://docs.userintuition.ai/llms.txt > Use this file to discover all available pages before exploring further. > User Intuition automatically synthesizes your interview data into actionable reports. Learn how to generate, interpret, and use AI-powered insights. # Understanding Reports User Intuition automatically synthesizes your interview data into actionable reports. Reports transform hours of conversations into clear insights, recommendations, and supporting evidence. *** ## Accessing Reports From your Study Dashboard: 1. Click the **Reports** tab 2. View the AI-generated analysis A report needs usable interview evidence. Interview count alone does not establish whether the evidence is adequate for your research purpose. *** ## Report Structure Reports are organized into several sections designed to take you from high-level findings to supporting detail: ### Executive Summary A one- or two-paragraph overview of the most important findings, followed by a compact set of within-sample metrics. It communicates what the study found without mixing in the separate assessment of evidence adequacy. ### Total Participants A clear count of how many interviews contributed to this report (e.g., "Total Participants (N): 1"). This helps you and your stakeholders understand the sample size behind the insights. ### Key Insights A short list of the most significant, source-linked findings and material exceptions. Key Insights are synthesized from the same full cross-participant review used for Detailed Findings; they are not generated from a separate evidence source. ### Detailed Findings Each Learning Goal contains numbered findings. Every finding separates its main point, within-sample prevalence, verbatims, observed impact, variation and exceptions, interpretation and limits, and decision relevance. Counts always include both the count and percentage—for example, `2 of 5 participants (40%)`. These percentages describe only the report sample; they are not population estimates. **Example:** > * **Value is Price-Driven, Not Actively Tracked:** The participant defines value as getting "bargains" and low prices, trusting the overall basket total reflects savings rather than performing explicit ROI calculations. > > * **Pharmacy Lacks Credibility:** Despite using the pharmacy for convenience, the participant feels it lacks a "real health care experience" and desires more proactive counsel from staff. ### Detailed Analysis In-depth exploration of findings organized by research objective or theme. Each section includes: * **Key observations:** What the data shows * **Supporting evidence:** Quotes and examples from interviews * **Implications:** What this means for your business ### Recommendations Zero to two justified next steps, each linked to an important evidence gap in the same report. If a material research gap is surfaced, the report must provide at least one linked next step. A valid empty list means the assessment found no actionable, decision-relevant research gap. An unavailable assessment is a different state. Recommendations require your existing review and launch approvals. ### Evidence Coverage Evidence Coverage appraises the evidence behind Detailed Findings rather than repeating the findings. It describes the evidence available for each Learning Goal, coverage limitations, and the resulting decision boundary. The study-level sample-adequacy assessment explains what the included interviews can and cannot support and whether additional interviews would materially help. It considers the research purpose, participant specificity, depth, variation, alternatives and missing perspectives instead of applying a universal sample-size threshold. One prioritized gap list explains what remains unresolved and why it matters. Generic qualitative boundaries—such as not being able to estimate population prevalence or prove causal impact—remain decision boundaries unless the study purpose explicitly requires that decision. When report inputs are unchanged, regenerated reports reuse the last validated gap and linked recommendations. New interviews, plan edits, or generation-contract changes trigger a fresh assessment. Per-interview analysis remains available in Participant Responses and call evaluation, but is not aggregated into a separate study-level metric. Missing analysis does not automatically imply a need to recruit. These assessments are interpretations, not adequacy scores. You decide whether the evidence is sufficient for the intended decision. Older or revised plans explicitly identify inferred requirements or uncertain timing. *** ## Generating Reports ### Initial Report Generation Reports can be generated on demand when usable interviews are available: 1. Navigate to the **Reports** tab 2. Click **Generate Report** (if no report exists yet) 3. Wait for the AI to analyze your interviews and produce the report ### Updating Reports As more responses come in, you'll need to regenerate the report to include new data: 1. Navigate to the **Reports** tab 2. Click **Regenerate Report** 3. The updated report will incorporate all completed interviews **Coming Soon:** Automatic report regeneration after each new response. *** ## Working with Reports ### Public API structure `GET /api/public/v1/studies/{study_id}/report` returns a small `report-v2` overview by default. It includes identity, freshness, counts, a short `summary`, and `available_sections`. Pass `?view=study_findings`, `participant_profiles`, `participant_responses`, `evidence_coverage`, `recommended_next_steps`, or `references` to fetch one section; pass `?view=full` when the complete report is needed. `included_sections` identifies which sections were requested. `based_on_interview_ids_included` says whether the complete interview-ID list is present. The selectable sections contain: * `study_findings`: executive summary and metrics plus `learning_goals`, each containing stable `finding_id` values that Evidence Coverage can reference * `participant_profiles`: an array of profile questions with stable, heading-derived `question_id` values * `participant_responses`: every report-eligible call's `overall_takeaways` and `learning_goal_responses` * `recommended_next_steps` * `evidence_coverage`, containing its introductory `text`, `sample_adequacy`, per-goal `learning_goal_evaluations`, and prioritized `evidence_gaps`; recommendations link to these gaps through `gap_ids` The public response intentionally omits the rendered report, Key Insights, duplicate open-question text, internal assessment fields, empty rendering wrappers, and presentation metadata such as Markdown, heading level, and display order. `evidence_gaps.learning_goal_ids` identifies the affected Learning Goal, and its `finding_ids` resolve directly to findings nested under that goal. New reports preserve the interview count, response analysis, evidence coverage, and recommendations from their generation snapshot. Later evidence or plan changes mark a report stale; reading it does not silently refresh its assessment. Legacy reports need regeneration to obtain evidence coverage. Each finding has a stable ID, `reference_ids`, and structured `prevalence: { n, of }` when the saved report states an explicit within-sample frequency. Its `quotes[]` identifies source references, public interview IDs, transcript turn IDs, and stored start times when available. Older findings can have null prevalence or empty quotes; the API does not infer missing evidence. Fetch `?view=references` to resolve citation IDs, then `get_interview` for the source turn. Public report responses use `interview_id` as their interview join key. New reports are indexed for research search and cited answers. `POST /api/public/v1/research/search/` searches canonical study plans and report objects across authorized studies. Search ranks candidates, but its `content` values are loaded from persisted Study/Report JSON. Results and index freshness are grouped under each entry in `studies`; check `index_status` separately from that study's `results` array. Follow `next_cursor` with the same request to continue. `POST /api/public/v1/research/answer/` generates an answer from accessible indexed evidence and returns validated source citations; inspect the cited records before using an exact quote. ### Sharing with Stakeholders Reports are designed to be presentation-ready: * **Executive Summary** for leadership and quick reviews * **Top Insights** for team meetings and discussions * **Detailed Analysis** for deep dives and planning sessions ### Using Quotes and Evidence The report includes verbatim quotes from participants. Use these to: * Build credibility for your findings * Illustrate points in presentations * Create empathy for customer perspectives ### Connecting to Transcripts If you want to explore a finding more deeply: 1. Note the relevant insight from the report 2. Switch to the **Calls** tab 3. Review the source transcripts for additional context *** ## Report Quality ### Factors Affecting Report Quality | Factor | Impact | | - | - | | **Number of responses** | More responses = more robust patterns | | **Response quality** | High-quality interviews provide richer insights | | **Research plan clarity** | Well-defined objectives lead to focused analysis | | **Topic coverage** | Consistent coverage across interviews strengthens findings | ### How many interviews a report needs Reports can generate from **2 interviews**, but a report does not make a small sample broadly representative. Sample adequacy depends on the research purpose, how specific and relevant the participants are, interview depth, observed variation, unresolved alternatives, and whether the intended decision requires segment comparison or prevalence estimates. Evidence Coverage makes that boundary explicit and explains when additional interviews would materially help. *** ## Frequently Asked Questions Reports are generated by AI and are read-only. You can copy content to edit externally. Regenerate when you've collected a meaningful batch of new responses (e.g., 5-10 more interviews). Frequent regeneration with small increments may not show significant changes. Currently, one standard report format is generated. Custom report formats may be available for enterprise customers—contact support for details. Reports require substantive interview data. If your report seems thin, review interview depth and quality, participant relevance and variation, and whether the research plan has clear Learning Goals and evidence requirements. There is no universal recommended interview count. Yes. Use the export options to download the report in various formats for sharing or archiving. Not currently. You must manually regenerate reports to include new responses. Automatic updates are on our roadmap. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.