Lecture 29: User Feedback and Iterative Optimization

Master user feedback collection and data analysis methods to continuously optimize Skill experience.

一、Feedback Collection Channels

1.1 Feedback Channel Matrix

Channel	Advantages	Disadvantages	Applicable Scenario
In-app feedback	Immediate	Low user initiative	Real-time issue collection
User interviews	High depth	High cost	Before major redesigns
Survey	Wide coverage	Low response rate	Satisfaction surveys
Data analytics	Objective quantification	Lacks reasons	Usage behavior analysis
Community forum	User mutual help	Scattered information	Long-term operations

1.2 Feedback Collection Implementation

# feedback.py
from datetime import datetime
from typing import Dict, List

class FeedbackCollector:
    """Feedback collector"""

    def __init__(self, storage):
        self.storage = storage

    def collect_feedback(self, user_id: str, feedback_type: str,
                        content: str, metadata: Dict = None) -> str:
        """Collect user feedback"""
        feedback_id = generate_id()

        feedback = {
            'id': feedback_id,
            'user_id': user_id,
            'type': feedback_type,  # bug, feature, improvement, other
            'content': content,
            'metadata': metadata or {},
            'status': 'new',  # new, processing, resolved, closed
            'created_at': datetime.now().isoformat(),
            'updated_at': datetime.now().isoformat()
        }

        self.storage.save(f"feedback:{feedback_id}", feedback)

        # Send notification
        self._notify_team(feedback)

        return feedback_id

    def get_feedback_stats(self, days: int = 30) -> Dict:
        """Get feedback statistics"""
        feedbacks = self._get_recent_feedbacks(days)

        stats = {
            'total': len(feedbacks),
            'by_type': {},
            'by_status': {},
            'avg_response_time': 0
        }

        for fb in feedbacks:
            # Count by type
            fb_type = fb['type']
            stats['by_type'][fb_type] = stats['by_type'].get(fb_type, 0) + 1

            # Count by status
            status = fb['status']
            stats['by_status'][status] = stats['by_status'].get(status, 0) + 1

        return stats

二、Data Analytics

2.1 Core Metrics

# analytics.py
from dataclasses import dataclass
from typing import List

@dataclass
class SkillMetrics:
    """Skill metrics"""
    # Usage metrics
    total_sessions: int
    active_users: int
    avg_session_duration: float

    # Performance metrics
    avg_response_time: float
    error_rate: float
    availability: float

    # Satisfaction metrics
    completion_rate: float
    fallback_rate: float
    user_rating: float

class AnalyticsDashboard:
    """Analytics dashboard"""

    def __init__(self, storage):
        self.storage = storage

    def get_daily_metrics(self, date: str) -> SkillMetrics:
        """Get daily metrics"""
        data = self.storage.load(f"metrics:{date}")

        return SkillMetrics(
            total_sessions=data.get('sessions', 0),
            active_users=data.get('active_users', 0),
            avg_session_duration=data.get('avg_duration', 0),
            avg_response_time=data.get('avg_response_time', 0),
            error_rate=data.get('error_rate', 0),
            availability=data.get('availability', 100),
            completion_rate=data.get('completion_rate', 0),
            fallback_rate=data.get('fallback_rate', 0),
            user_rating=data.get('user_rating', 0)
        )

    def get_intent_distribution(self, days: int = 7) -> Dict:
        """Get intent distribution"""
        intents = {}

        for i in range(days):
            date = (datetime.now() - timedelta(days=i)).strftime('%Y-%m-%d')
            daily_intents = self.storage.load(f"intents:{date}") or {}

            for intent, count in daily_intents.items():
                intents[intent] = intents.get(intent, 0) + count

        return intents

    def identify_issues(self) -> List[Dict]:
        """Identify issues"""
        issues = []

        # Check error rate
        recent_errors = self._get_recent_errors(hours=24)
        if len(recent_errors) > 10:
            issues.append({
                'type': 'high_error_rate',
                'severity': 'high',
                'description': f'{len(recent_errors)} errors occurred in 24 hours',
                'suggestion': 'Check recent updates or external dependencies'
            })

        # Check response time
        avg_response_time = self._get_avg_response_time(hours=24)
        if avg_response_time > 3.0:
            issues.append({
                'type': 'slow_response',
                'severity': 'medium',
                'description': f'Average response time {avg_response_time:.2f}s',
                'suggestion': 'Optimize performance or add resources'
            })

        # Check user satisfaction
        recent_ratings = self._get_recent_ratings(days=7)
        if recent_ratings and statistics.mean(recent_ratings) < 3.0:
            issues.append({
                'type': 'low_satisfaction',
                'severity': 'high',
                'description': 'User satisfaction below 3.0',
                'suggestion': 'Collect feedback and improve experience'
            })

        return issues

三、Iterative Optimization

3.1 A/B Testing

# ab_test.py
import random
from typing import Dict, Any

class ABTestManager:
    """A/B test manager"""

    def __init__(self, storage):
        self.storage = storage
        self.tests = {}

    def create_test(self, test_id: str, variants: List[str],
                   traffic_split: List[float]):


## 🎓 AI 编程实战课程

想系统学习 AI 编程？程序员晚枫的 **AI 编程实战课** 帮你从零上手！

- 👉 **课程报名**：[点击这里报名，前3讲免费试听](https://r7up9.xetslk.com/s/1uP5YW)
- 👉 **免费试看**：[B站免费试看前3讲，先看看适不适合自己](https://www.bilibili.com/cheese/play/ss982042944)