AgentSkillsCN

Analyze Customer Churn

分析客户流失率

SKILL.md

Skill: Analyze Customer Churn

Domain

technology

Description

Analyzes customer churn risk using behavioral signals, usage patterns, and engagement metrics to identify at-risk accounts for retention intervention.

Tags

churn, retention, customer-success, SaaS, analytics, machine-learning

Use Cases

  • •Churn risk scoring
  • •Early warning detection
  • •Retention campaign targeting
  • •Customer health monitoring

Proprietary Business Rules

Rule 1: Engagement Decay Detection

Identification of declining usage patterns over time.

Rule 2: Risk Signal Weighting

Weighted scoring of behavioral churn indicators.

Rule 3: Cohort Comparison

Risk assessment relative to similar customer segments.

Rule 4: Intervention Prioritization

Ranking of at-risk customers by value and savability.

Input Parameters

  • •customer_id (string): Customer identifier
  • •usage_metrics (dict): Product usage data
  • •engagement_history (list): Interaction timeline
  • •account_info (dict): Customer profile
  • •support_tickets (list): Support history
  • •billing_history (list): Payment patterns

Output

  • •churn_risk_score (float): Risk probability 0-100
  • •risk_factors (list): Contributing factors
  • •risk_trend (string): Increasing/stable/decreasing
  • •recommended_actions (list): Retention interventions
  • •predicted_churn_date (string): Estimated churn timing

Implementation

The analysis logic is implemented in churn_analyzer.py and references data from CSV files:

  • •usage_benchmarks.csv - Reference data
  • •engagement_windows.csv - Reference data
  • •sentiment_weights.csv - Reference data
  • •billing_thresholds.csv - Reference data
  • •savability_factors.csv - Reference data
  • •risk_weights.csv - Reference data
  • •signal_definitions.csv - Reference data
  • •intervention_playbook.csv - Reference data
  • •cohort_benchmarks.csv - Reference data
  • •parameters.csv - Reference data.

Usage Example

python
from churn_analyzer import analyze_churn

result = analyze_churn(
    customer_id="CUST-001",
    usage_metrics={"logins_30d": 5, "features_used": 3, "api_calls": 100},
    engagement_history=[{"date": "2025-12-01", "type": "login"}],
    account_info={"mrr": 5000, "tenure_months": 18, "plan": "enterprise"},
    support_tickets=[{"date": "2025-12-15", "category": "bug", "sentiment": "negative"}],
    billing_history=[{"date": "2025-12-01", "status": "paid", "days_late": 0}]
)

print(f"Churn Risk: {result['churn_risk_score']}%")

Test Execution

python
from churn_analyzer import analyze_churn

result = analyze_churn(
    customer_id=input_data.get('customer_id'),
    usage_metrics=input_data.get('usage_metrics', {}),
    engagement_history=input_data.get('engagement_history', []),
    account_info=input_data.get('account_info', {}),
    support_tickets=input_data.get('support_tickets', []),
    billing_history=input_data.get('billing_history', [])
)