The Scenario
You're the analytics lead at StreamFlix Plus, a streaming service losing $2.3M/month to churn. For each customer, you'll analyze their support tickets (text sentiment) and usage chart (time series pattern), then predict whether they'll churn.
Scoring: +10 per correct ticket sentiment, +20 for correct usage pattern, +40 for correct churn prediction. But false alarms (predicting churn on happy customers) cost -15 points!
8 customers. One screen each. Text + time series combined — just like real churn models.
Final Prediction: Will this customer churn?
Analysis Complete
Skills
Customer Log
Key Takeaways
Text + time series is more powerful than either alone. Angry tickets with declining usage = high churn risk. But happy tickets with seasonal usage dips ≠ churn — context matters. False alarms waste retention budget, so confidence in your combined signal is critical.