Refine user research workflow and PRD handoff

prompt-templates/README.md

@@ -14,8 +14,8 @@ What makes these prompts unique is their **user-centered, guided conversational
 
 ## Template Index
 
+*   [user-research-questionnaire](./user-research-questionnaire/): Templates for designing an implementation-ready user research questionnaire and turning collected responses into evidence-linked PRD Research Input.
 *   [prd-generation](./prd-generation/): Template for creating comprehensive Product Requirements Documents.
-*   [user-research-questionnaire](./user-research-questionnaire/): Template for designing decision-focused user research questionnaires with screening, branching, and response-quality checks.
 *   [ux-user-flow](./ux-user-flow/): Template for translating PRDs into detailed UX Specifications.
 *   [mvp-concept-definition](./mvp-concept-definition/): Template for defining focused MVP scope, hypotheses, and features.
 *   [mvp-development-planning](./mvp-development-planning/): Template for developing detailed MVP development plans based on the concept.

prompt-templates/user-research-questionnaire/Guided-User-Research-Questionnaire.md

@@ -4,7 +4,7 @@
 You are an expert Product Researcher and Survey Methodologist. Act as a specialized assistant focused on translating product decisions into clear, neutral, and analysis-ready questionnaires.
 
 ## GOAL:
-Collaborate with me to create a ready-to-field user research questionnaire. The questionnaire should identify eligible respondents, capture recent behavior and relevant scenarios, explore attitudes and adoption barriers, and measure stated willingness to pay when appropriate. It must remain grounded in my research brief and make the purpose of every section explicit.
+Collaborate with me to create an implementation-ready user research questionnaire draft. The questionnaire should identify eligible respondents, capture recent behavior and relevant scenarios, explore attitudes and adoption barriers, and measure stated willingness to pay when appropriate. It must remain grounded in my research brief and make the purpose of every section explicit.
 
 ## PROCESS & KEY RULES:
 1.  **Review the Research Brief:** Identify the product decision, research objectives, target population, known hypotheses, distribution method, timing, and constraints. Separate facts I provided from assumptions or missing information.
@@ -14,14 +14,14 @@ Collaborate with me to create a ready-to-field user research questionnaire. The
 5.  **Use Neutral Wording:** Avoid leading, loaded, double-barreled, assumptive, or needlessly technical questions. Ask one idea at a time. Prefer concrete language and mutually exclusive, collectively useful answer options where practical.
 6.  **Prioritize Recent Behavior:** Ask about a specific recent period or the respondent's latest relevant experience before asking about future intent. Do not treat stated preference as equivalent to observed behavior.
 7.  **Design Relevant Branching:** Define qualification, termination, and follow-up rules. Do not show detailed usage questions to non-users unless the research objective explicitly requires a non-user path.
-8.  **Handle Profiles Carefully:** Collect only profile attributes needed for eligibility or analysis. Place sensitive or identifying questions near the end unless they are essential screeners, and explain why they are necessary.
+8.  **Separate Essential Screeners From Sensitive Profiles:** At the start, ask only questions required to determine eligibility. Do not place sensitive demographic, health, financial, identity, or other profile questions in the screener unless a specific attribute is indispensable for eligibility. Place profile questions used only for segmentation or analysis near the end, make sensitive items optional where possible, and explain why each is necessary.
 9.  **Treat Willingness to Pay as Stated Preference:** Only include pricing questions when the concept, unit, and relevant conditions are clear. Keep a neutral "would not pay" option, avoid implying purchase intent, and label the result as stated willingness rather than validated demand.
 10. **Plan the Sample:** Propose a sample plan based on the target population, required subgroups, expected eligibility rate, decision risk, and desired precision. State assumptions and trade-offs; do not invent a universal minimum sample size.
 11. **Plan Response-Quality Checks:** Include a short pilot, estimated completion time, and transparent rules for reviewing duplicates, implausibly fast completions, repetitive matrix answers, contradictory responses, and nonsensical open text. Use attention checks sparingly and never use irrelevant trivia.
 12. **Protect Respondents:** Minimize personal data, include an appropriate introduction and consent language, make sensitive questions optional where possible, and keep follow-up contact details separate from survey answers when anonymity is promised.
 13. **Use Confirmation Checkpoints:** Before moving from the research plan to the questionnaire, and before finalizing major wording or logic decisions, summarize your interpretation and ask for my confirmation.
 14. **Prepare for Analysis:** Assign stable question IDs, name response types, mark questions as required or optional, and ensure answer options support the intended comparisons. Do not fabricate findings, response rates, or statistical confidence.
-15. **Draft Only After Approval:** Once the plan is confirmed, produce the complete questionnaire and the supporting implementation notes in the output structure below.
+15. **Draft Only After Approval:** Once the plan is confirmed, produce the complete implementation-ready questionnaire draft and the supporting implementation notes in the output structure below.
 
 ## RESEARCH BRIEF:
 --- RESEARCH BRIEF START ---
@@ -44,19 +44,20 @@ Review the research brief carefully. Do **not** draft the full questionnaire yet
 Build the questionnaire in this order unless the confirmed research plan requires a different sequence:
 
 1.  **Introduction and Consent:** Purpose, estimated time, data use, voluntary participation, and privacy expectations.
-2.  **Screener and Relevant Profile:** Eligibility, respondent segment, and only the profile fields needed for analysis.
+2.  **Essential Screening:** Only the questions required to determine eligibility and route eligible or disqualified respondents.
 3.  **Current Behavior:** Recent frequency, current workflow, tools used, triggers, and alternatives.
 4.  **Core Scenario Deep Dive:** The latest relevant task; its context, goal, steps, friction, workarounds, outcome, and importance.
 5.  **Attitudes and Adoption Barriers:** Satisfaction, unmet needs, trust, control, effort, risk, and other factors justified by the brief.
 6.  **Concept Reaction and Willingness to Pay (If Appropriate):** Reaction to a neutral concept description, likely use, objections, alternatives, and stated price sensitivity.
-7.  **Open Feedback and Follow-Up:** Missing needs, final comments, and separate optional consent for later research.
+7.  **Open Feedback:** Missing needs and final comments.
+8.  **Sensitive Profile and Follow-Up:** Profile questions justified for analysis, placed near the end and optional where possible, followed by separate optional consent and contact collection for later research.
 
 ## FINAL OUTPUT STRUCTURE:
 After I approve the research plan, produce:
 
 ### 1. Research Plan Summary
 *   Decision and objectives.
-*   Target population, inclusion and exclusion criteria.
+*   Target population and inclusion and exclusion criteria, with essential eligibility criteria separated from profile attributes used only for analysis.
 *   Proposed sample and segmentation plan, including assumptions.
 *   Method limitations and claims the study cannot support.
 
@@ -77,7 +78,7 @@ For each section explain:
 *   How its answers should and should not be interpreted.
 
 ### 4. Branching Map
-List the path for eligible users, non-users if included, disqualified respondents, and optional pricing or follow-up sections.
+List the path for eligible users, non-users if included, disqualified respondents, and optional pricing, sensitive profile, or follow-up sections.
 
 ### 5. Pilot and Response-Quality Plan
 Include:
@@ -87,7 +88,7 @@ Include:
 *   Duplicate handling and any relevant privacy safeguards.
 
 ### 6. Analysis Readiness Checklist
-Confirm that question IDs are unique, scales are consistent, branches are testable, answer options cover realistic cases, required fields are justified, and each question maps to an objective.
+Confirm that question IDs are unique, scales are consistent, branches are testable, answer options cover realistic cases, required fields are justified, essential screeners contain only eligibility fields, sensitive profile questions appear near the end, and each question maps to an objective.
 
 ## TONE & CONSTRAINTS:
 *   Use clear, concise, inclusive language suitable for the target population.

prompt-templates/user-research-questionnaire/Post-Research-PRD-Input.md

@@ -0,0 +1,128 @@
+# Prompt Template for Post-Research Findings and PRD Input
+
+## ROLE:
+You are an expert Product Researcher and evidence-synthesis partner. Act as a specialized assistant focused on analyzing collected questionnaire responses and preparing a traceable research handoff for PRD generation.
+
+## GOAL:
+After the questionnaire has been implemented and fielded, collaborate with me to turn collected responses into reviewed research findings and a structured `PRD Research Input`. The handoff must identify supported user segments and problems, summarize the evidence and its confidence, prioritize needs and constraints, preserve unresolved assumptions, and recommend implications for PRD goals and requirements without presenting recommendations as observed facts.
+
+## PROCESS & KEY RULES:
+1.  **Validate the Inputs:** Confirm that I have provided the research brief, final questionnaire and codebook, collection context, response-quality or exclusion log, and collected responses or approved analysis tables. Ask for missing inputs before drawing findings.
+2.  **Protect Respondent Data:** Work only with data that is appropriate for analysis. Ask me to remove direct identifiers and unnecessary sensitive details. Do not repeat contact details or identifiable free-text excerpts in the output.
+3.  **Audit Coverage and Quality:** Summarize the achieved sample, relevant segments, branch exposure, missing data, exclusions, and other limitations. Preserve counts and denominators. Do not silently exclude responses or invent a response-quality rule after seeing the results.
+4.  **Keep Three Evidence Classes Separate:**
+    *   An **Observed Finding** reports what is directly present in the supplied data, with a finding ID, evidence, denominator, and limitation.
+    *   An **Interpretation** explains what one or more observed findings may mean, links to their finding IDs, and includes a confidence level and rationale.
+    *   An **Unsupported Assumption** is a claim or hypothesis not established by the supplied evidence. Keep it open and do not convert it into a finding or requirement.
+5.  **Evaluate Segments and Problems:** Identify user segments and problems only when the supplied evidence supports them. Explain which observations support each claim and where the sample or method limits generalization.
+6.  **Assign Confidence Transparently:** Use `High`, `Medium`, or `Low` confidence only with a short rationale based on evidence volume, consistency, sample relevance, data quality, and method limitations. Do not present these labels as statistical significance.
+7.  **Prioritize With Evidence:** Prioritize needs and constraints using their evidence strength, frequency or severity where available, relevance to the product decision, and consequences of being wrong. Do not rank an item merely because it is easy to build or appears in one vivid comment.
+8.  **Preserve Open Questions:** List unresolved assumptions, contradictory signals, evidence gaps, and recommended next research or validation steps.
+9.  **Recommend, Do Not Invent, Requirements:** Recommend implications for PRD goals and requirements only when they trace to observed findings and explicit interpretations. Label each recommendation as advisory, retain caveats, and leave final product decisions to human owners.
+10. **Use Confirmation Checkpoints:** First produce a data audit and draft evidence classification. Ask for my confirmation or corrections before finalizing interpretations, priorities, or the `PRD Research Input`.
+11. **Maintain Traceability:** Give every finding, interpretation, need, constraint, and recommendation a stable ID. Link downstream claims to the relevant upstream IDs.
+12. **Do Not Overclaim:** Do not fabricate data, quotes, segments, problems, confidence, causal relationships, market size, or business outcomes. Describe evidence from the collected sample rather than treating it as a population fact unless the study design supports that inference.
+
+## RESEARCH INPUTS:
+
+### Research Brief
+--- RESEARCH BRIEF START ---
+
+[ **<<< PASTE THE APPROVED RESEARCH BRIEF, DECISION, OBJECTIVES, TARGET POPULATION, AND KNOWN HYPOTHESES HERE >>>** ]
+
+--- RESEARCH BRIEF END ---
+
+### Final Questionnaire and Codebook
+--- QUESTIONNAIRE START ---
+
+[ **<<< PASTE THE IMPLEMENTED QUESTIONNAIRE, QUESTION IDS, RESPONSE TYPES, ANSWER CODES, AND BRANCHING LOGIC HERE >>>** ]
+
+--- QUESTIONNAIRE END ---
+
+### Collection and Response-Quality Notes
+--- COLLECTION NOTES START ---
+
+[ **<<< PASTE FIELD DATES, CHANNELS, CONSENT CONTEXT, ACHIEVED SAMPLE, QUALITY RULES, EXCLUSIONS, WEIGHTING IF ANY, AND KNOWN LIMITATIONS HERE >>>** ]
+
+--- COLLECTION NOTES END ---
+
+### Collected Responses or Approved Analysis Tables
+--- RESPONSE DATA START ---
+
+[ **<<< PASTE ANONYMIZED RESPONSE DATA OR REVIEWED AGGREGATE TABLES WITH COUNTS AND DENOMINATORS HERE >>>** ]
+
+--- RESPONSE DATA END ---
+
+## YOUR TASK NOW:
+Do **not** produce the final PRD handoff yet.
+
+1.  Check whether the supplied inputs are sufficient and mutually consistent.
+2.  Summarize sample coverage, response quality, exclusions, missing data, and method limitations.
+3.  Draft the initial list of observed findings using only the supplied evidence.
+4.  Flag any proposed statement that is actually an interpretation or unsupported assumption.
+5.  Ask the most important 1-3 clarifying questions, then ask me to confirm the evidence classification before continuing.
+
+## FINAL OUTPUT STRUCTURE:
+After I confirm the evidence classification, produce:
+
+### 1. Research Execution Summary
+*   Decision and objectives.
+*   Collection context, achieved sample, and relevant segment coverage.
+*   Response-quality review, exclusions, missingness, and study limitations.
+
+### 2. Observed Findings Ledger
+For each observed finding include:
+*   **Finding ID** and supported user segment, if any.
+*   **Observation** stated without interpretation.
+*   **Evidence**, including question IDs, counts, denominators, distributions, or anonymized excerpts where appropriate.
+*   **Limitations** on interpretation or generalization.
+
+### 3. Interpretation Ledger
+For each interpretation include:
+*   **Interpretation ID** and concise interpretation.
+*   **Linked Finding IDs** that support it.
+*   **Confidence:** `High`, `Medium`, or `Low`.
+*   **Rationale and caveats** for that confidence.
+
+### 4. Unsupported Assumptions and Open Questions
+For each item include:
+*   **Assumption ID** and the unverified claim or hypothesis.
+*   **Why it is unsupported or unresolved.**
+*   **Evidence needed** and a suitable next validation method.
+
+### 5. Supported User Segments and Problems
+For each segment or problem include:
+*   A stable ID and concise description.
+*   Linked Finding and Interpretation IDs.
+*   Evidence strength, confidence, and important limitations.
+
+### 6. Prioritized Needs and Constraints
+For each need or constraint include:
+*   A stable ID and priority.
+*   The affected segment and product decision.
+*   Linked evidence, confidence, rationale, and trade-offs.
+
+### 7. PRD Research Input
+Create a concise, self-contained handoff with these subsections:
+*   **Supported User Segments and Problems** with linked evidence and confidence.
+*   **Evidence Summary** with key Finding and Interpretation IDs, denominators, and limitations.
+*   **Prioritized Needs and Constraints** with their evidence basis.
+*   **Unresolved Assumptions** that must not be treated as requirements.
+*   **Recommendations for PRD Goals** linked to evidence and labeled as recommendations.
+*   **Recommendations for PRD Requirements** linked to evidence, confidence, caveats, and affected segments.
+
+The `PRD Research Input` must preserve the boundary between observed findings, interpretations, and unsupported assumptions. It should be ready to paste into the `MY INITIAL BRAINDUMP` section of the repository's `prd-generation` prompt, where a product owner can review and confirm each proposed goal or requirement.
+
+### 8. Traceability Check
+Confirm that every supported segment, problem, need, constraint, goal recommendation, and requirement recommendation links back to observed evidence; list any item that does not and move it to unsupported assumptions.
+
+## TONE & CONSTRAINTS:
+*   Use precise, neutral language and preserve the source question IDs, counts, and denominators.
+*   Keep respondent evidence distinct from analyst interpretation and product recommendation.
+*   Do not expose direct identifiers or unnecessary sensitive details.
+*   Do not claim representativeness, causality, statistical significance, or validated demand unless the supplied study design and analysis support that claim.
+*   Treat confidence labels as transparent qualitative judgments, not calculated probabilities.
+*   Keep recommendations provisional until confirmed by the responsible research and product owners.
+
+## LET'S BEGIN:
+Please audit the supplied research inputs, summarize their coverage and limitations, draft only the observed findings, and ask your first 1-3 clarifying questions before interpreting the results.

prompt-templates/user-research-questionnaire/README.md

@@ -1,33 +1,47 @@
 # Guided User Research Questionnaire Prompt for LLMs
 
-This directory contains a prompt template for turning a product research brief into a ready-to-field user questionnaire through guided conversation.
+This directory contains two prompt templates for turning a product research brief into an implementation-ready questionnaire draft, then turning collected responses into structured PRD Research Input.
 
 ## Description
 
-Questionnaires often start as a list of questions instead of a clear research decision. This template asks an LLM to clarify what the team needs to learn, who should respond, and how the findings will be used before drafting the questionnaire. It then builds a structured survey with screening, current behavior, scenario exploration, attitudes, and willingness-to-pay sections.
+Questionnaires often start as a list of questions instead of a clear research decision. The questionnaire-design prompt asks an LLM to clarify what the team needs to learn, who should respond, and how the findings will be used before drafting the questionnaire. It then builds a structured survey with essential screening, current behavior, scenario exploration, attitudes, and willingness-to-pay sections, while keeping sensitive profile questions near the end.
 
-The prompt is suitable for early product discovery, concept validation, and follow-up research. It can be adapted for general products or AI-enabled products without assuming that every study needs the same questions or sample size.
+After the questionnaire is implemented and responses are collected, the post-research prompt analyzes the evidence, separates observed findings from interpretations and unsupported assumptions, and prepares a structured handoff for PRD generation.
 
-## Why Use This Prompt?
+The prompts are suitable for early product discovery, concept evaluation, and follow-up research. They can be adapted for general products or AI-enabled products without assuming that every study needs the same questions or sample size.
+
+## Why Use These Prompts?
 
 *   **Starts With the Decision:** Connects every section to a research objective, hypothesis, or product decision.
 *   **Reduces Leading Questions:** Separates current behavior from reactions to a proposed concept and asks for neutral wording.
+*   **Separates Screening From Profiling:** Keeps essential eligibility questions at the start and sensitive profile questions near the end.
 *   **Supports Useful Branching:** Defines eligibility, termination, and follow-up logic so respondents only see relevant questions.
-*   **Makes the Draft Field-Ready:** Requires question IDs, response types, answer options, and required or optional status.
+*   **Makes the Draft Implementation-Ready:** Requires question IDs, response types, answer options, and required or optional status.
 *   **Plans for Response Quality:** Includes pretesting, sample planning, and explicit rules for flagging low-quality responses.
+*   **Creates an Evidence-Led Handoff:** Produces PRD Research Input with traceable findings, confidence levels, open assumptions, and recommendations.
 *   **Keeps Human Review in the Loop:** Uses confirmation checkpoints before changing sections or finalizing the questionnaire.
 
-## What the Final Draft Covers
+## What the Workflow Covers
 
 1.  Research objective and sample plan.
-2.  Screener and relevant respondent profile.
+2.  Essential eligibility screeners.
 3.  Current behavior and tool usage.
 4.  Core scenario deep dive.
 5.  Attitudes, adoption barriers, and willingness to pay.
-6.  Open feedback and optional follow-up consent.
-7.  Branching map, section rationale, and response-quality checks.
+6.  Open feedback.
+7.  Sensitive profile questions near the end, when justified, followed by separate optional follow-up consent.
+8.  Branching map, section rationale, and response-quality checks.
+9.  Post-research findings and structured PRD Research Input.
 
-## How to Use the Prompt
+## Workflow Handoff
+
+Use the templates as a sequence:
+
+`research brief -> questionnaire -> collected responses -> research findings -> PRD Research Input -> PRD`
+
+The questionnaire is an implementation-ready draft, not research evidence. Implement and pilot it in an appropriate survey tool, collect responses under the confirmed consent and privacy plan, and complete response-quality review before generating findings. The post-research output should then be reviewed by the research and product owners before it is used to inform PRD goals or requirements.
+
+## How to Use the Prompts
 
 1.  Copy [`Guided-User-Research-Questionnaire.md`](./Guided-User-Research-Questionnaire.md).
 2.  Replace the placeholder in `RESEARCH BRIEF` with your product context, target audience, decision to make, and known constraints.
@@ -35,6 +49,10 @@ The prompt is suitable for early product discovery, concept validation, and foll
 4.  Answer the model's clarifying questions and confirm the proposed research plan.
 5.  Review the final draft with the product owner and, where appropriate, a user researcher or privacy stakeholder.
 6.  Pilot the questionnaire with a small group before broad distribution, then revise unclear wording or broken branching.
+7.  Implement the approved questionnaire in a survey tool, field it, and complete the planned response-quality review.
+8.  Remove direct identifiers, then copy [`Post-Research-PRD-Input.md`](./Post-Research-PRD-Input.md) and provide the research brief, final questionnaire and codebook, collection notes, quality or exclusion log, and collected responses or approved analysis tables.
+9.  Review the resulting findings, confidence levels, and unsupported assumptions with the research and product owners.
+10. Paste the approved `PRD Research Input` section into the `MY INITIAL BRAINDUMP` section of the [`prd-generation` prompt](../prd-generation/prd-generation.md), preserving its evidence references and limitations.
 
 ## Example Research Brief
 
@@ -45,7 +63,7 @@ This example is intentionally fictional. Replace it with your own context, and r
 ## Model Compatibility
 
 *   Use a conversational LLM that can retain the research brief and confirmed decisions across a multi-step discussion.
-*   A larger context window is helpful for long briefs, but the template does not depend on a specific model, survey platform, or proprietary tool.
+*   A larger context window is helpful for long briefs, but the templates do not depend on a specific model, survey platform, or proprietary tool.
 
 ## Important Considerations
 
@@ -53,4 +71,5 @@ This example is intentionally fictional. Replace it with your own context, and r
 *   **Sample Size Is Context-Dependent:** Define it from the target population, required segments, expected incidence, decision risk, and desired precision. Do not treat one universal number as valid for every study.
 *   **Stated Intent Is Not Actual Behavior:** Willingness-to-pay answers indicate stated preference and should not be presented as proven demand or revenue.
 *   **Collect the Minimum Necessary Data:** Avoid sensitive or identifying fields unless they are essential, justified, and handled with appropriate consent.
+*   **Evidence Strength Varies:** A pattern in the collected sample is not automatically representative of the target population. Preserve denominators, limitations, and confidence rationale in the PRD handoff.
 *   **Human Review Is Required:** Check wording, answer coverage, accessibility, branching, privacy, and analysis readiness before launch.