Reference & Attachments: Research Continuity

One-line summary: Link historical research + upload documents, AI continues deep diving based on existing insights.

---

Why Do You Need Reference & Attachments?

The Fragmentation Problem of Traditional Research

Problem 1: Research Lacks Continuity

Scenario:

• Week 1: Completed market trend analysis
• Week 4: Want to do product validation
• But AI doesn't remember Week 1's analysis results
• Need to re-explain background information

Result:

• Waste time repeating context
• Unable to build on previous findings
• Research becomes fragmented, lacks continuity

---

Problem 2: External Documents Are Hard to Utilize

Scenario:

• You have competitor analysis reports (PDF)
• You have product prototypes
• You have user feedback in Excel
• Want AI to conduct research based on these materials
• But requires manual organization and input

Result:

• Spend a lot of time organizing materials
• AI may not understand completely
• Low efficiency

---

The Solution: Reference & Attachments

Reference Studies:

• Link previous research projects
• AI automatically reads complete research logs
• Build upon existing findings
• Form research chains, not isolated projects

File Attachments:

• Upload relevant documents (PDF, images, Excel...)
• AI automatically extracts content
• Conduct research and analysis based on documents
• No manual organization needed

Core Value:

• ✅ Research can accumulate, forming knowledge networks
• ✅ External materials easily integrated
• ✅ From fragmented → systematic research
• ✅ 50%+ efficiency boost

---

Reference Study: Research Continuity

What is Reference Study?

Definition:

• In a new research session, link one or more historical research projects
• AI automatically reads the complete content of historical research (questions, findings, conclusions)
• New research can build upon historical findings for deeper investigation

vs. Memory System:

• Memory System: Remembers your background, preferences, research direction
• Reference Study: Specifically links complete content of a particular research session

Applicable Scenarios:

• Progressive research: First do market analysis, then product validation
• Deep investigation: Based on initial findings, research a specific niche direction
• Comparative validation: Compare research results from different time periods

---

Real Case Study: Bubble Coffee from 0 to 1

Research 1: Market Opportunity Analysis (Week 1)

Task: "Does bubble coffee have market opportunity?"

Research Content:

• Scout Agent observes social media discussions
• Analyzes coffee market trends
• Identifies potential user groups

Core Findings:

1. Market has opportunity; young people interested in "refreshing coffee"
2. Target users: Urban women aged 25-35
3. Main motivation: Not caffeine, but "relaxed experience"
4. Price acceptance: $25-30/cup

Output: Research report + research log

---

Research 2: Product Positioning Validation (Week 3)

Task: "Based on market opportunity, validate product positioning"

Key Action: Link Research 1

AI automatically reads Research 1 content:

• Target users: Urban women aged 25-35
• Core need: "Relaxed experience" not caffeine
• Price acceptance: $25-30

Research Content:

• Discussion test of 3 product positioning directions
• Interview deep validation of most popular positioning

Core Findings:

• Most popular positioning: "Relaxed atmosphere afternoon tea substitute"
• Users like "refreshing", "not too sweet", "suitable for social occasions"
• Emotional value > functionality (caffeine boost)

Value:

• No need to re-explain "who is the target user"
• AI automatically designs Discussion topics based on Research 1 findings
• Research is coherent, not starting from zero

---

Research 3: Packaging Design Testing (Week 5)

Task: "Test 3 packaging designs; see which attracts target users most"

Key Actions:

• Link Research 1 and Research 2
• Upload 3 packaging design drafts (images)

AI automatically reads:

• Research 1: Target users urban women 25-35, price $25-30
• Research 2: Positioning "relaxed atmosphere afternoon tea substitute", users value "refreshing" and "social occasions"

Research Content:

• Discussion discusses 3 packaging designs
• AI focuses on based on existing insights:
  • Which design better fits "relaxed atmosphere"
  • Which design makes users feel "suitable for social occasions"
  • Which design is worth $25-30

Core Findings:

• Design B most popular
• Reason: Simple and refreshing, suitable for sharing on social media (social occasions)
• Design A too corporate, Design C too flashy

Value:

• AI understands complete context (target users + positioning + price)
• Test questions more precise
• Research results directly applicable

---

Research 4: Pricing Strategy Optimization (Week 7)

Task: "Validate optimal pricing"

Key Action: Link Research 1, 2, 3

AI automatically reads:

• Research 1: Initial price acceptance $25-30
• Research 2: Positioning "relaxed atmosphere", high emotional value
• Research 3: Sophisticated packaging design

Research Content:

• Interview deep-dive on pricing sensitivity
• AI focuses on: "Based on packaging design B and 'relaxed atmosphere' positioning, how much would you pay?"

Core Findings:

• Optimal pricing: $28
• Users think $25 "too cheap, doesn't feel sophisticated enough"
• $30 "a bit pricey, exceeds mental budget"
• $28 "reasonable, matches product positioning"

Value:

• Pricing research based on complete context (product positioning + packaging design)
• Not asking in isolation "how much would you pay"
• Research results more accurate

---

4 Research Sessions Form Complete Chain:

vs. Fragmented Research:

• If 4 research sessions were independent, AI would need to repeatedly ask about background
• Each time would need to re-explain "who is the target user", "what's the product positioning"
• Low efficiency, poor research quality

vs. Reference Study:

• 4 research sessions form knowledge chain
• Each research builds on previous findings
• Research depth progressively increases
• Final output is complete, systematic product solution

---

File Attachments: Document Integration

Supported File Types

Document Types:

• PDF: Competitor analysis reports, industry reports, user feedback...
• TXT/Markdown: Text materials
• CSV/Excel: Data tables
• Images: Product prototypes, design drafts, competitor screenshots...
• Maximum 10MB/file

File Quantity:

• Multiple files can be uploaded per research session
• Recommended not to exceed 10 (too many affects AI understanding efficiency)

---

Use Case 1: Research Based on Competitor Reports

Need:

• Have a 50-page competitor analysis PDF
• Want AI to find differentiation opportunities based on this report

Traditional Method:

1. Read 50-page report (1-2 hours)
2. Manually organize competitor features
3. Input to AI
4. Ask AI to analyze differentiation opportunities

Using File Attachments:

1. Upload PDF
2. Ask AI: "Based on this competitor report, how should our product differentiate?"
3. AI automatically:
   • Reads PDF
   • Extracts competitor features
   • Discussion/Interview tests differentiation directions
   • Outputs recommendations

Case Study:

Upload competitor analysis PDF for coffee brand

You: "Based on this report, how should bubble coffee differentiate?"

AI Auto-Extracts Competitor Features:

• Starbucks: Convenient, stable, but "unremarkable"
• Specialty coffee: Quality, professional, but "too expensive", "not relaxing"
• Luckin: Value for money, but "too commercial"

AI Initiates Discussion: "Discussion topic: Bubble coffee differentiation

• Direction A: More distinctive than Starbucks
• Direction B: More relaxing than specialty coffee
• Direction C: More emotional value than Luckin"

Output: Based on Discussion results, recommend differentiation positioning: "Relaxing × Distinctive × Emotional Value"

• vs Starbucks: More distinctive (bubble + emotional label)
• vs Specialty coffee: More relaxing (not pursuing professionalism, pursuing experience)
• vs Luckin: More emotional value (not work scenario, relaxation scenario)

Value:

• Save 1-2 hours reading time
• AI automatically extracts key information
• Conduct research directly from report, seamless integration

---

Use Case 2: Prototype Testing

Need:

• Designer created 3 product prototypes
• Want to quickly test user reactions

Using File Attachments:

1. Upload 3 prototype images
2. Ask AI: "Test these 3 prototypes; which do users prefer?"
3. AI automatically:
   • Show 3 images to AI personas
   • Discussion discusses which is more usable, better looking
   • Collect feedback

Case Study:

Upload 3 registration flow prototype images

AI Initiates Discussion: "Show 3 registration flow prototypes (Image A, B, C); discuss:

1. Which is easiest to understand?
2. Which flow is smoothest?
3. Any confusion or obstacles?"

Discussion Results:

• Image A: Simple but lacks guidance; users don't know next step
• Image B: Clear guidance, but too many steps; feels tedious
• Image C: Balanced; clear guidance and simplified steps

Output: Recommend using Image C, with optimizations:

1. Add "Complete registration in just 2 minutes" prompt at first step
2. Add progress bar at second step
3. Put social login button at top

Value:

• Quickly test prototypes without waiting for real users
• AI personas provide detailed feedback
• Can test multiple versions for rapid iteration

---

Use Case 3: User Feedback Analysis

Need:

• Customer service collected 200 user feedback items (Excel)
• Want to know what users care about most

Using File Attachments:

1. Upload Excel file
2. Ask AI: "Analyze user feedback; identify most common issues and needs"
3. AI automatically:
   • Reads Excel
   • Categorizes and counts
   • Extracts high-frequency keywords
   • Summarizes core issues

Case Study:

Upload user feedback Excel (200 items)

AI Analysis: High-Frequency Issues (50+ occurrences):

1. "Can't find a feature" (78 times)
   • Specifically: Settings page, export data, history
2. "Slow loading speed" (65 times)
   • Mainly on mobile
3. "Don't know how to use" (52 times)
   • Mostly new users; lack of guidance

Positive Feedback (30+ occurrences):

1. "Feature very useful" (45 times)
2. "Interface simple" (38 times)

Recommendations:

1. Optimize feature entry points (solve "can't find" issue)
2. Optimize mobile performance (solve "slow loading")
3. Add new user onboarding (solve "don't know how to use")

Value:

• Save manual analysis time (traditional requires 3-5 hours)
• AI auto-categorizes and counts
• Quickly discover core issues

---

Use Case 4: Research Material Integration

Need:

• Have multiple materials:
  • Industry report PDF
  • Competitor analysis Excel
  • User interview notes TXT
  • Product roadmap PPT screenshots
• Want AI to integrate these materials for comprehensive analysis

Using File Attachments:

1. Upload all files at once
2. Ask AI: "Based on these materials, where is our product opportunity?"
3. AI automatically:
   • Reads all files
   • Integrates information
   • Discovers connections and opportunities
   • Outputs comprehensive analysis

Case Study:

Upload 4 files:

• Health Food Industry Report.pdf
• Competitor Analysis.xlsx
• User Interview Notes.txt
• Product Roadmap.png

AI Integrated Analysis: Market Opportunity (Industry Report):

• Health food market growing 15% annually
• Emotional value becoming new selling point

Competitor Gap (Competitor Analysis):

• Competitor A: Healthy but boring
• Competitor B: Tasty but not healthy enough
• Gap: Healthy + Fun (emotional value)

User Needs (Interview Notes):

• Want to "enjoy eating without guilt"
• Care about "emotional healing"

Product Positioning Recommendation: Based on above analysis, recommend positioning: "Tasty Healthy Snack + Emotional Healing"

• Fills competitor gap
• Matches market trend
• Meets user needs

Value:

• Auto-integrate multiple materials
• Discover cross-file connections
• Quickly form comprehensive insights

---

Combined Use: Reference & Attachments

Best Practice: Continuous Research + External Materials

Scenario: Brand wants to reposition product

Research 1: Market Analysis

• Upload industry report PDF
• Scout Agent observes social media
• Integrate analysis of market trends

Core Findings:

• Emotional value becomes new trend
• Young people pursue "healing feeling"

---

Research 2: Competitor Analysis

• Link Research 1 (market trends)
• Upload competitor analysis Excel
• Discussion discusses differentiation directions

Core Findings:

• Competitors all playing "health" card
• "Emotional healing" is gap

---

Research 3: User Validation

• Link Research 1 + Research 2
• Interview validates "emotional healing" positioning
• Deep understanding of user motivation

Core Findings:

• Users willing to pay premium for "healing feeling"
• Want packaging design "sophisticated + warm"

---

Research 4: Packaging Testing

• Link Research 3 (user expectations for packaging)
• Upload 3 packaging design drafts
• Discussion collects feedback

Core Findings:

• Design B best fits "sophisticated + warm"
• Recommend adding emotional label

---

Form Complete Research Chain:

Value:

• Research is coherent, not fragmented
• External materials seamlessly integrated
• Form systematic insights
• 50%+ efficiency boost

---

Reference & Attachments vs. Other Methods

vs. Manual Material Organization

---

vs. Starting Each Research Fresh

---

Common Questions

Q1: How many historical research can I link?

Recommended 1-3:

• Too many affects AI understanding efficiency
• Choose most relevant research to link

If need more:

• Can first summarize multiple research into one
• Or use Memory System (automatically remembers all research)

---

Q2: Are uploaded files permanently saved?

Saved but can delete:

• Files saved in your account
• Can delete anytime
• Deletion doesn't affect completed research results

Privacy Protection:

• Files encrypted and stored
• Only you can access
• Not visible to other users

---

Q3: Can AI fully understand uploaded files?

Mostly yes, but with limitations:

Can Understand:

• Pure text PDF (AI can read text)
• Excel tables (AI understands data structure)
• Text in images (OCR recognition)
• Markdown/TXT text

Limited Understanding:

• Complex charts (AI may not fully understand)
• Handwritten notes (low recognition accuracy)
• Very specialized terminology (may need supplementary explanation)

Recommendation:

• After uploading, ask AI: "Did you understand the core content of this file?"
• AI will summarize its understanding
• If discrepancy, provide additional explanation

---

Q4: What's the difference between Reference Study and Memory System?

Reference Study:

• Link complete content of specific research session
• Best for: Progressive research, deep investigation
• Requires manual linking

Memory System:

• Automatically remembers your background, preferences, research direction
• Best for: Long-term use, auto-accumulation
• No manual operation needed

Combined Use:

• Memory System: Remember "who you are, who are your target users"
• Reference Study: Specific research findings and conclusions
• Both together make research more efficient

---

Q5: Can I edit uploaded files?

Cannot edit directly, but can re-upload:

1. Delete old file
2. Upload new file
3. AI will re-analyze based on new file

Recommendation:

• If file updated, re-upload
• Keep file current

---

Q6: For uploaded images, can AI understand design?

Yes, but with limits:

AI Can Understand:

• Layout structure (whether clear)
• Color style (minimal, flashy, corporate, etc.)
• Text content (button copy, titles, etc.)
• Overall feeling (professional, cute, premium, etc.)

AI Understanding Limited:

• Fine design details (font choice, spacing micro-adjustments)
• Brand consistency (whether matches brand tone)
• Emotional resonance (whether truly "healing")

Recommendation:

• Can let AI do initial screening
• But final decision still needs human judgment

---

Q7: File too large, upload fails?

Single file limit 10MB:

Solutions:

1. PDF Compression: Use PDF compression tool
2. Image Compression: Lower image resolution
3. Split Files: If Excel, split into multiple files to upload
4. Extract Core Content: Only upload most relevant parts

Recommendation:

• Prioritize uploading core materials
• Don't upload redundant content

---

Q8: Can I add files mid-research?

Yes:

• Upload new files anytime during research
• AI will auto-read and integrate into research

Scenario:

• Research halfway through, realize need more materials
• Upload directly; AI uses immediately

---

Practical Recommendations

1. Plan Research Path, Gradually Deepen

Recommended Order:

1. 1st Research: Initial market/user understanding
2. 2nd Research: Link 1st, deepen specific direction
3. 3rd Research: Link both previous, validate specific solution

Avoid:

• Research all problems at once
• Jumping research, lacking continuity

---

2. Selective File Upload

Upload Core Materials:

• Competitor reports, industry reports
• Product prototypes, design drafts
• User feedback, interview notes

Avoid Uploading:

• Redundant materials
• Unrelated content
• Outdated materials

Reason:

• Too many files affect AI understanding efficiency
• Quality over quantity

---

3. Verify AI Understanding

After Each File Upload:

• Ask AI: "Did you understand this file's core content?"
• AI will summarize its understanding
• If discrepancy, provide supplement

Value:

• Ensure AI understands correctly
• Avoid basing research on wrong understanding

---

4. Periodically Clean Up Outdated Research

Recommendation: Every 3-6 months

Clean Up Content:

• Research no longer relevant
• Based on outdated assumptions
• Replaced by new research

Value:

• Keep research library clean
• Avoid AI linking to outdated research

---

Summary

Reference & Attachments Core Value:

1. Research Continuity: Build on historical findings for deeper investigation; form knowledge chains
2. Document Integration: External materials seamlessly integrated; AI auto-extracts content
3. Systematic Research: From fragmented → systematic; research deeper
4. Efficiency Boost: No repeated context; save 50%+ time

Applicable Scenarios:

• Progressive research: Market analysis → product validation → packaging testing...
• Deep investigation: Based on initial findings, research specific niche direction
• Material Analysis: Conduct research based on competitor reports, user feedback etc.
• Prototype Testing: Upload design drafts, quickly collect feedback

vs. Other Methods:

• vs Manual organization: 10x efficiency boost; no missed key information
• vs Starting fresh: Progressive research depth; knowledge accumulation not fragmentation

Best Practices:

• Plan research path, gradually deepen
• Selectively upload core materials
• Verify AI understanding
• Periodically clean outdated research

Combined Use:

• Reference Study + Memory System: Research continuity + auto-memory
• File Attachments + Reference Study: External materials + historical research → comprehensive insights
• Maximize effectiveness

---

Document Version: v2.0 | 2026-01-15 | Pure User Perspective