retention-analysis

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Retention Analysis Skill

留存分析Skill

Analyze user retention patterns, predict customer churn, and optimize retention strategies using advanced statistical methods and machine learning techniques.

使用高级统计方法和机器学习技术分析用户留存模式、预测客户流失并优化留存策略。

Quick Start

快速开始

This skill helps you:

Calculate retention rates and churn metrics
Build survival curves using Kaplan-Meier analysis
Perform cohort analysis to understand behavior patterns
Predict churn risk with machine learning models
Identify retention drivers using Cox regression
Generate actionable insights for retention improvement

本Skill可帮助你：

计算留存率和流失指标
使用Kaplan-Meier分析构建生存曲线
执行同期群分析以了解行为模式
用机器学习模型预测流失风险
使用Cox回归识别留存驱动因素
生成可落地的洞察以提升留存

When to Use

适用场景

SaaS Product Analysis: User subscription renewal and cancellation patterns
Membership Programs: Member engagement and loyalty analysis
E-commerce: Customer repeat purchase behavior and subscription boxes
Gaming Apps: Player retention and engagement metrics
Service Industries: Customer satisfaction and long-term relationships
Subscription Businesses: Monthly/yearly subscription analysis

SaaS产品分析：用户订阅续订与取消模式
会员计划：会员参与度与忠诚度分析
电子商务：客户复购行为与订阅盒分析
游戏应用：玩家留存与参与度指标
服务行业：客户满意度与长期关系分析
订阅业务：月度/年度订阅分析

Key Requirements

核心依赖

Install required packages:

bash

pip install pandas numpy matplotlib seaborn scikit-learn lifelines

安装所需依赖包：

bash

pip install pandas numpy matplotlib seaborn scikit-learn lifelines

Core Workflow

核心工作流

1. Data Preparation

1. 数据准备

Your data should include:

User identifiers: Unique user/customer IDs
Time variables: Registration date, activity dates, subscription period
Event indicators: Churn status (1=churned, 0=active)
User attributes: Demographics, behavior, subscription details
Optional: Usage metrics, payment history, engagement data

你的数据应包含：

用户标识符：唯一的用户/客户ID
时间变量：注册日期、活动日期、订阅周期
事件指标：流失状态（1=已流失，0=活跃）
用户属性：人口统计信息、行为数据、订阅详情
可选：使用指标、支付历史、参与度数据

2. Analysis Process

2. 分析流程

Data preprocessing: Clean and prepare retention data
Survival analysis: Build Kaplan-Meier curves
Cohort analysis: Group users by acquisition time
Risk modeling: Identify churn drivers with Cox regression
Churn prediction: Build machine learning prediction models
Insight generation: Create actionable recommendations

数据预处理：清洗并准备留存数据
生存分析：构建Kaplan-Meier曲线
同期群分析：按获取时间分组用户
风险建模：使用Cox回归识别流失驱动因素
流失预测：构建机器学习预测模型
洞察生成：制定可落地的建议

3. Output Deliverables

3. 输出成果

Retention rate tables and charts
Survival curves with confidence intervals
Cohort heatmaps and behavior patterns
Churn risk scores and feature importance
Retention optimization strategies

留存率表格与图表
带置信区间的生存曲线
同期群热力图与行为模式
流失风险评分与特征重要性
留存优化策略

Example Usage Scenarios

示例使用场景

SaaS Subscription Analysis

SaaS订阅分析

python

undefined

python

undefined

Analyze monthly subscription renewal patterns

Predict which users are likely to churn

Identify features that drive long-term retention

undefined

undefined

Membership Program Analysis

会员计划分析

python

undefined

python

undefined

Track member engagement over time

Compare retention across membership tiers

Analyze payment method impact on retention

undefined

undefined

E-commerce Customer Retention

电子商务客户留存

python

undefined

python

undefined

Analyze repeat purchase patterns

Calculate customer lifetime value

Identify high-value customer segments

undefined

undefined

Key Analysis Methods

核心分析方法

Survival Analysis

生存分析

Kaplan-Meier Estimator: Non-parametric survival curve
Log-rank Test: Compare survival between groups
Cox Proportional Hazards: Multi-variable risk modeling
Median Survival Time: Time when 50% of users have churned

Kaplan-Meier Estimator：非参数生存曲线
Log-rank Test：组间生存情况比较
Cox Proportional Hazards：多变量风险建模
Median Survival Time：50%用户流失所需时间

Cohort Analysis

同期群分析

Time-based Cohorts: Group by acquisition month/quarter
Behavior-based Cohorts: Group by usage patterns
Retention Matrix: Visualize retention over time periods
Cohort Comparison: Compare different cohort behaviors

Time-based Cohorts：按获取月份/季度分组
Behavior-based Cohorts：按使用模式分组
Retention Matrix：可视化不同时间段的留存情况
Cohort Comparison：比较不同同期群的行为

Machine Learning Prediction

机器学习预测

Logistic Regression: Binary churn classification
Random Forest: Non-linear pattern detection
Gradient Boosting: High accuracy prediction
Feature Importance: Identify key churn drivers

Logistic Regression：二元流失分类
Random Forest：非线性模式检测
Gradient Boosting：高精度预测
Feature Importance：识别关键流失驱动因素

Common Business Questions Answered

常见业务问题解答

What is our overall retention rate?
How does retention vary by user segment?
What factors most influence customer churn?
Which users are at highest risk of leaving?
How can we improve long-term retention?
What is the typical customer lifetime?

我们的整体留存率是多少？
不同用户群体的留存率有何差异？
哪些因素对客户流失影响最大？
哪些用户流失风险最高？
我们如何提升长期留存率？
典型的客户生命周期是多久？

Integration Examples

集成示例

See examples/ directory for:

```
basic_retention.py
```
- Survival analysis basics
```
cohort_analysis.py
```
- Cohort-based retention analysis
```
churn_prediction.py
```
- ML-based churn prediction
Sample datasets for testing

查看examples/目录获取：

```
basic_retention.py
```
- 生存分析基础
```
cohort_analysis.py
```
- 基于同期群的留存分析
```
churn_prediction.py
```
- 基于机器学习的流失预测
用于测试的示例数据集

Best Practices

最佳实践

Data Quality: Ensure accurate churn definitions and time measurements
Event Definition: Clearly define what constitutes "churn"
Time Windows: Choose appropriate analysis periods
Segmentation: Analyze different user groups separately
Validation: Always validate models with test data
Business Context: Consider operational constraints and costs

数据质量：确保流失定义和时间测量的准确性
事件定义：明确界定“流失”的标准
时间窗口：选择合适的分析周期
细分分析：单独分析不同用户群体
验证：始终使用测试数据验证模型
业务场景：考虑运营约束与成本

Advanced Features

高级功能

Competing Risks Analysis: Different types of churn
Time-varying Covariates: Dynamic feature analysis
Customer Lifetime Value: Integrate retention with revenue
Retention Forecasting: Predict future retention trends
A/B Testing: Measure retention improvement impact

竞争风险分析：不同类型的流失
时变协变量：动态特征分析
客户生命周期价值：将留存与收入结合分析
留存预测：预测未来留存趋势
A/B测试：衡量留存提升措施的效果