Build a Churn Prediction Model
Company: Gusto
Role: Data Scientist
Category: Machine Learning
Difficulty: medium
Interview Round: Onsite
You are asked to build a churn prediction model for a consumer product. The business wants to identify users who are likely to churn so that the product or marketing team can intervene.
Answer the following:
1. How would you define the prediction label and prediction horizon?
2. What features would you use, and how would you avoid leakage?
3. What model families would you consider?
4. How would you evaluate the model offline and online?
5. If a tree-based model performs much better than logistic regression, how would you explain that result?
Quick Answer: This question evaluates a data scientist's applied machine learning skills, specifically supervised learning for churn prediction with competencies in defining prediction labels and horizons, feature engineering and leakage prevention, selecting model families, and designing offline and online evaluation.