Feedback Collection

Build feedback channels suited to your team for data-driven improvement

Stage-Based Feedback Collection

Teams at different stages need different feedback collection approaches. Start lightweight and gradually improve.

Early Exploration Stage

Story Card Tagging Method

Applicable Scenario: Teams just starting to try AI-assisted development

Advantages:

Zero additional process burden
Naturally integrates with existing workflow
Data accumulates naturally with work

Add Fields to Story Cards

Add a few simple fields in your existing task management system:

Field	Type	Options
AI Assistance Level	Single Select	High / Medium / Low / None
Estimated Time Saved	Number	Hours
Main AI Contribution	Single Select	Code Generation / Q&A / Refactoring Suggestions / Other
Main Pain Points	Text	Free form

Fill In Upon Task Completion

When completing a task, developers spend 30 seconds filling in these fields:


## Task: Implement User List Page
 
- [x] Development complete
- AI Assistance Level: High
- Estimated Time Saved: 2 hours
- Main AI Contribution: Code Generation
- Main Pain Points: Responsive layout needed multiple adjustments

Regular Summary Analysis

Summarize data weekly or bi-weekly to form simple reports:


## This Week's AI-Assisted Development Data
 
- Total tasks: 15
- Tasks using AI assistance: 12 (80%)
- Total estimated time saved: 18 hours
- Common contributions: Code generation (8), Q&A (3), Refactoring suggestions (1)
- Common pain points: Responsive layout, complex state management

Quick Start Tip: If you use Jira, Linear, or Notion, these tools all support custom fields and can be set up in 5 minutes.

Growth Phase

Regular Feedback Forms

Applicable Scenario: Team has some AI-assisted development experience, needs more systematic feedback

Frequency Recommendation: Bi-weekly or monthly

Form Design Principles

1. Keep It Short

Limit the survey to 10 questions or fewer, completion time under 5 minutes.

2. Combine Quantitative and Qualitative

Quantitative: Ratings, numerical estimates
Qualitative: Open-ended questions to collect specific cases

3. Focus on Actionable Feedback

Every question should guide subsequent improvements.

Recommended Question List

Overall Assessment


1. How satisfied are you with current AI-assisted development overall?
   [1-10 slider]
 
2. Would you recommend using Cursor for development to colleagues?
   [0-10 NPS rating]
 
3. What percentage of your time did you use Cursor this period?
   [0-30% / 30-50% / 50-70% / 70-90% / 90%+]

Efficiency Assessment


4. Compared to not using AI, how much do you feel coding efficiency improved?
   [No improvement / 10-30% / 30-50% / 50-100% / 100%+ / 200%+]
 
5. In which types of tasks was AI most helpful? (Multiple choice)
   [ ] UI component development
   [ ] API interface development
   [ ] Business logic implementation
   [ ] Bug fixes
   [ ] Code refactoring
   [ ] Test writing
   [ ] Documentation
   [ ] Other: _____

Quality Assessment


6. How would you rate the quality of AI-generated code?
 
   Business functionality correctness: [1-5]
   Code readability: [1-5]
   Code maintainability: [1-5]
   Performance: [1-5]
   Security: [1-5]

Pain Point Collection


7. What's the biggest challenge you face when using AI-assisted development?
   [Open text]
 
8. In which scenarios do you feel AI doesn't perform well enough, worse than manual coding?
   [Open text]

Improvement Suggestions


9. What do you think is most worth improving? (Select up to 3)
   [ ] Prompt quality and templates
   [ ] Cursor Rules enhancement
   [ ] Context management
   [ ] Team knowledge sharing
   [ ] Workflow optimization
   [ ] Training and learning resources
   [ ] Other: _____
 
10. Please share one memorable AI-assisted development case from this period (good or bad)
    [Open text]

Data Analysis Template

After collecting data, analyze using this template:


## [Month] AI-Assisted Development Feedback Analysis
 
### Participation
- Respondents: X / Total team size Y
- Response rate: Z%
 
### Core Metrics
- Average satisfaction: X.X / 10 (Change: +/- X.X)
- NPS score: X (Promoters X%, Passives X%, Detractors X%)
- Average usage ratio: XX-YY%
 
### Efficiency Perception
- People reporting 50%+ efficiency improvement: X%
- Most helpful task types: UI development, API interfaces
- Least helpful task types: Complex business logic
 
### Quality Ratings
| Dimension | Average Score | Change |
|-----------|---------------|--------|
| Functionality | 4.2 | +0.3 |
| Readability | 3.8 | +0.1 |
| Maintainability | 3.2 | -0.2 |
 
### Main Pain Points
1. [Pain point 1] - Mentions: X
2. [Pain point 2] - Mentions: Y
3. [Pain point 3] - Mentions: Z
 
### Improvement Priority
1. [Improvement 1] - Votes: X
2. [Improvement 2] - Votes: Y
3. [Improvement 3] - Votes: Z
 
### Action Items
- [ ] [Specific action] - Owner: @xxx - Due: Date

Mature Optimization Stage

Automated Data Collection

Applicable Scenario: Team’s AI-assisted development practice is mature, needs more precise data

Git Commit Correlation Analysis

Add AI assistance markers (e.g., AI-Assisted: true) in Git commit messages, then use scripts to track AI-assisted commit ratios and trends.

IDE Usage Statistics

Passively collect usage data through IDE extensions:

WakaTime: Automatically track coding time
Code Time: Track coding habits

Automated Quality Metrics Tracking

Integrate AI-related quality metrics into CI/CD Pipeline, automatically tracking:

AI assistance labels in PRs
Code change volumes
Subsequent bug correlation tracking

Feedback Collection Best Practices

Lower the Barrier to Entry

Common Failure Reason: Survey is too long or complex, developers don’t want to fill it out.

Keep survey under 5 minutes
Provide default options
Allow skipping non-required fields
Mobile-friendly

Show Feedback Value Promptly

Let the team see changes driven by feedback:


## Feedback-Driven Improvement Log
 
| Date | Feedback Source | Issue | Improvement | Effect |
|------|-----------------|-------|-------------|--------|
| 2024-01 | Monthly survey | Deep component nesting | Updated Rules | Nesting reduced 40% |
| 2024-02 | Story card tags | Missing API error handling | New Prompt template | Rework rate down 25% |

Build Feedback Culture

Leadership Example: Tech leads fill out and share first
Public Discussion: Discuss feedback results in team meetings
Reward Participation: Thank members who provide valuable feedback

Protect Privacy

Anonymous option (especially for sensitive questions like satisfaction)
Aggregate display rather than individual tracking
Clear data usage purpose

Tool Recommendations

Tool	Use	Features
Google Forms	Regular surveys	Free, easy to use, supports analysis
Notion	Task tagging + knowledge base	Good workflow integration
Linear	Task management + custom fields	Developer-friendly
Airtable	Data collection and analysis	Flexible data model
WakaTime	Automatic time tracking	Passive collection, no manual input

Next Steps

After starting to collect feedback data, read the Retrospective Guide to learn how to analyze and leverage this data.