Business Context
CustomerFirst, a customer relationship management (CRM) platform serving 30,000 businesses, seeks to improve its predictive model for customer satisfaction scores. The goal is to enhance the accuracy of predictions to better tailor services and interventions for clients, thereby reducing churn and increasing engagement.
Dataset
| Feature Group | Count | Examples |
|---|
| Customer Demographics | 10 | age, location, industry, customer_tenure |
| Interaction Metrics | 15 | calls_made, emails_sent, tickets_created |
| Survey Responses | 5 | net_promoter_score, satisfaction_score, feedback_comments |
| Usage Data | 8 | active_days_last_month, features_used |
- Size: 50,000 records, 38 features
- Target: Continuous variable — customer satisfaction score (0-10 scale)
- Class balance: N/A (regression problem)
- Missing data: 10% missing in feedback_comments, 5% in interaction metrics
Requirements
- Propose a systematic feature prioritization strategy for the predictive model.
- Identify key features that should be included based on their potential impact on customer satisfaction.
- Discuss methods for handling missing data and feature engineering techniques.
- Provide a rationale for the chosen features and their expected contributions to model performance.
Constraints
- The model must be interpretable for stakeholders to understand feature impacts.
- Must handle missing data gracefully without dropping significant records.
- The solution should be scalable to accommodate future feature additions.
