5.1 Lessons Learned
High-Level Insights
- Iteration is Everything: The most significant improvements came from iterative feedback sessions. The design evolved significantly from### Publications and Presentations:
- Publication 1: [Title, venue, status]
- Presentation 1: [Title, venue, date]
Community Contributions:
- Contribution 1: [What was shared with the community]
- Contribution 2: [What was shared with the community]
Internal Documentation:
- Document 1: [Internal knowledge sharing artifact]
- Document 2: [Internal knowledge sharing artifact]sketch to the final product.
- The Power of Low-Fidelity: Hand-drawn sketches were crucial. They allowed for a fast and non-committal way to explore ideas. Domain experts felt more comfortable critiquing a rough sketch than a polished visualization.
- Documentation is Key: The logs for data cleaning, sketches, and build decisions were invaluable. They created a clear and traceable record of the entire process, which was essential for our final report and for understanding the design rationale.
- Tacit Knowledge is Gold: The “think-aloud” sessions and observations were more informative than the formal interviews. They helped us uncover pain points that the domain expert took for granted.
Detailed Stage Reflections
Stage 1: Abstract Phase
What Worked Well:
- Semi-structured Interviews: Open-ended questions revealed unexpected pain points about manual data correlation
- Task Observation: Watching the domain expert work revealed the “Wednesday problem” pattern
- Iterative Abstraction: Multiple rounds of task refinement led to the three core tasks (compare, find anomaly, locate hotspot)
What Could Be Improved:
- Initial Scope: Started too broad - should have focused on specific use cases earlier
- Technical Feasibility: Didn’t consider data limitations early enough, leading to some redesign
Key Insights:
- Task Abstraction: Real tasks emerged from observation rather than interviews - people don’t always know what they need
- User Interviews: Follow-up questions were crucial - initial answers were often surface-level
- Domain Understanding: Transportation domain has hidden complexities (weather, time-of-day effects) that only emerged through deep engagement
Time Allocation:
- Planned: 1 week
- Actual: 1.5 weeks
- Variance: Extra time needed for task refinement was valuable investment
Stage 2: Design Phase
What Worked Well:
- Hand-drawn Sketching: Rapid iteration without tool constraints - generated 8 different layout concepts
- Multiple Views Early: Decided on coordinated views approach early, which guided all subsequent design decisions
- Color Strategy: Delay-based color coding (red-yellow-green) was intuitive and tested well
What Could Be Improved:
- Data Volume Planning: Didn’t consider how design would scale with larger datasets
- Responsive Design: Late addition - should have been considered from the start
Key Insights:
- Sketching Process: Physical sketching encouraged broader exploration than jumping to digital tools
- Digital Tools: Figma excellent for layout refinement, but VegaLite prototyping caught interaction issues early
- Design Validation: Quick digital prototypes prevented major implementation problems
Iteration Effectiveness:
- Number of Iterations: 5 major design iterations
- Most Valuable Iteration: Iteration 3 - added geographic view after realizing spatial patterns were crucial
- Iteration Drivers: User feedback and technical constraints drove most changes
Stage 3: Build Phase
What Worked Well:
- Incremental Development: Building one chart at a time allowed for early testing and refinement
- D3.js Choice: Provided the flexibility needed for custom interactions and coordinated views
- Responsive Framework: CSS Grid made layout adaptation straightforward
What Could Be Improved:
- Error Handling: Added too late in the process - should be built in from the start
- Performance Testing: Didn’t test with larger datasets until late - found some bottlenecks
Key Insights:
- Technology Choices: D3.js learning curve was worth it for the custom interaction requirements
- Performance Considerations: Animation and transitions significantly improved user experience
- Design-to-Code Translation: Having detailed Figma mockups made implementation much smoother
Technical Debt:
- Accumulated Debt: [Technical shortcuts that created future problems]
- Time Pressure Impact: [How time constraints affected code quality]
- Lessons for Future: [How to better manage technical debt]
Stage 4: Evaluate Phase
What Worked Well:
- Structured Tasks: The three-task evaluation protocol effectively tested core functionality
- Think-Aloud Protocol: Captured reasoning process and revealed usability insights
- Time-boxed Tasks: 3-minute limit encouraged focused interaction and revealed efficiency issues
What Could Be Improved:
- Single User: Only testing with one domain expert limited feedback diversity
- Task Realism: Could have included more complex, real-world scenarios
Key Insights:
- Coordinated Views: Proved to be the most valuable feature - users immediately understood the connections
- Learning Curve: Minimal training needed - intuitive interface was achieved
- Performance Expectations: Users expect instant responsiveness in modern web tools
Process and Team Insights
Time Management:
- Stage 1 (Abstract): 1.5 weeks - Longer than expected but crucial for foundation
- Stage 2 (Design): 2 weeks - Hand sketching saved significant time later
- Stage 3 (Build): 3 weeks - Incremental development prevented major setbacks
- Stage 4 (Evaluate): 1 week - Could have been longer for multiple users
- Total: 7.5 weeks vs. 6 week target
Role Clarity:
- Researcher/Designer: Dual role worked well for small project, clear division of abstract vs. design thinking
- Developer Role: Having design background made implementation smoother
- Domain Expert: Regular check-ins prevented development of unusable features
Stakeholder Engagement:
- Engagement Strategies: Weekly demos kept domain expert engaged and provided regular course correction
- Feedback Integration: Prioritizing feedback based on core task support worked well
- Expectation Management: Early prototypes helped set realistic expectations for final product
Technical Learnings
Tool Effectiveness:
Tool/Technology | Effectiveness (1-5) | Notes |
---|---|---|
D3.js | 5 | Learning curve steep but flexibility was essential |
Figma | 4 | Excellent for layout design, good collaboration features |
VegaLite | 3 | Good for rapid prototyping, limited for complex interactions |
HTML/CSS Grid | 5 | Modern responsive design approach, very effective |
VS Code | 5 | Excellent development environment with good extensions |
Data Challenges:
- Data Quality Issues: Transport data had inconsistent formatting - required significant cleaning
- Data Access: Mock data worked well for prototype, but real API integration would add complexity
- Data Processing: D3’s data manipulation functions were powerful but required JavaScript expertise
User-Centered Design Insights
User Understanding:
- User Model Evolution: Initial assumption of “data analyst” user evolved to “operations manager” with different needs
- Persona Validation: Domain expert matched our target persona well, but broader user base would have different priorities
- Task Understanding: Real tasks were more exploratory than we initially assumed - needed support for discovery
Design Validation:
- Early Validation Benefits: Hand-drawn sketches prevented weeks of development in wrong direction
- Iteration Impact: User feedback in Stage 2 completely changed our chart selection approach
- Design Blindspots: Assumed users would want complex filtering - simple route selection was sufficient
Methodological Contributions
Process Improvements:
- Integrated Sketching-Prototyping: Combining hand sketches with rapid digital prototyping
- Problem Addressed: Gap between conceptual design and technical implementation
- Implementation: Use sketches for broad exploration, VegaLite for interaction testing, then build
- Continuous Domain Expert Engagement: Weekly check-ins throughout development
- Problem Addressed: Building features that don’t match real-world needs
- Implementation: Schedule brief weekly demos rather than waiting for major milestones
Tool Recommendations:
- Essential Tools: Hand sketching materials, Figma (or similar), D3.js (for flexibility), basic web hosting
- Helpful Additions: VegaLite for prototyping, color picker tools, browser dev tools for debugging
- Tools to Avoid: Overly complex frameworks early on - start simple and add complexity only when needed
Template Improvements:
- Documentation Templates: Add specific time tracking templates - actual time vs. estimates revealed valuable insights
- Process Checklists: Need checklist for data quality assessment early in the process
- Evaluation Frameworks: Single-user evaluation sufficient for initial validation, but multi-user framework needed for robustness
Domain-Specific Insights
Transportation Domain Characteristics:
- Unique Challenges: Time-sensitive decision making, geographic constraints, weather dependencies, public accountability
- Opportunities: Rich spatial and temporal data, clear performance metrics, strong user motivation for efficiency
- User Behaviors: Domain experts think in terms of routes and schedules, prefer actionable insights over exploratory analysis
Design Implications:
- Spatial Context Critical: Any transportation visualization needs geographic reference
- Time Patterns Essential: Day-of-week and time-of-day patterns are fundamental to domain understanding
- Performance Focus: Users primarily interested in identifying and fixing problems, not general exploration
Recommendations for Future Projects
For Similar Projects:
- Start with Real Data: Even if limited, real data reveals domain complexities that mock data doesn’t
- Prioritize Geographic Views: Spatial context is critical in transportation domain
- Focus on Actionability: Users want to identify problems and understand causes, not just see patterns
For Other Domains:
- Invest in Task Abstraction: Understanding real tasks (not stated tasks) is critical foundation
- Design Study Process Works: The 5-stage process provided good structure and prevented major missteps
- Document Everything: The documentation proved invaluable for understanding design decisions and writing this reflection
For Process Improvement:
- Expand Evaluation: Single domain expert was sufficient for initial validation but more users needed for robust findings
- Consider Long-term Use: Our study focused on initial usability - long-term adoption would require different considerations
- Technical Sustainability: Consider maintenance and data pipeline requirements for deployed tools
Transferable Insights:
- Insight 1: [Learning that could apply to other domains]
- Insight 2: [Learning that could apply to other domains]
Domain Expertise Development:
- Learning Curve: [How long it took to develop domain understanding]
- Key Learning Resources: [Most valuable sources of domain knowledge]
- Knowledge Gaps: [Areas where deeper expertise would have helped]
Impact Assessment
Immediate Impact:
- User Adoption: [How the final tool is being used]
- Workflow Changes: [How work processes have changed]
- Efficiency Gains: [Measurable improvements in user efficiency]
Long-term Potential:
- Scalability: [Potential for broader deployment]
- Evolution: [How the tool might evolve over time]
- Replication: [Potential for similar tools in related domains]
Research Contributions:
- Novel Techniques: [New visualization or interaction techniques developed]
- Validation Results: [Insights about visualization effectiveness]
- Methodology Advances: [Contributions to design study methodology]
Recommendations for Future Projects
Project Setup:
- Team Composition: [Ideal team structure and skills]
- Timeline Planning: [More realistic time allocation recommendations]
- Resource Requirements: [Essential resources and tools]
Process Adaptations:
- Stage Modifications: [How to adapt stages for different contexts]
- Checkpoint Additions: [Additional validation points to include]
- Documentation Practices: [Better ways to capture process learning]
Risk Mitigation:
- Common Pitfalls: [Mistakes to avoid in future projects]
- Contingency Planning: [How to handle common setbacks]
- Quality Assurance: [Additional quality checks to implement]
Knowledge Sharing
Publications and Presentations:
- Publication 1: [Title, venue, status]
- Presentation 1: [Title, venue, date]
Open Source Contributions:
- Contribution 1: [What was shared with the community]
- Contribution 2: [What was shared with the community]
Internal Documentation:
- Document 1: [Internal knowledge sharing artifact]
- Document 2: [Internal knowledge sharing artifact]
Personal and Professional Growth
Skills Developed:
- Technical Skills: [New technical capabilities gained]
- Design Skills: [Design competencies developed]
- Research Skills: [Research methodologies learned]
- Communication Skills: [Presentation and writing improvements]
Career Impact:
- Role Clarification: [How this project affected career direction]
- Network Building: [Professional relationships developed]
- Portfolio Development: [How this contributes to professional portfolio]
Areas for Continued Learning:
Skills for Future Development:
- Skill 1: [Area for future development]
- Skill 2: [Area for future development]
Final Reflections
Most Valuable Learning:
“[Single most important insight from the entire project]”
Biggest Surprise:
“[Most unexpected discovery or outcome]”
Would Do Differently:
“[Major change you would make if starting over]”
Advice for Others:
“[Key advice for someone starting a similar project]”
Project Satisfaction:
- Overall Rating: [1-5 scale]
- Most Rewarding Aspect: [What was most fulfilling]
- Most Challenging Aspect: [What was most difficult]
Navigation
- ← Previous Stage: Stage 4: Evaluate Phase
- 🏠 Stage 5: Post-Design Study Overview
- → Next: Key Elements of Collaboration
- 📚 Case Study: Transport Example
Appendices
A. Timeline Comparison
[Actual vs. planned timeline visualization]
B. Resource Utilization
[Analysis of time, budget, and resource usage]
C. Stakeholder Feedback Summary
[Key quotes and feedback from project stakeholders]
D. Future Research Questions
[Questions that emerged during the project that could drive future research]