Context

StreamBox uses a binary classification model to predict whether a subscriber will churn in the next 30 days so the retention team can send discount offers. The team says the model accuracy looks strong, but churn campaigns are still missing too many at-risk users.

Current Performance

The model was evaluated on a holdout set of 10,000 subscribers.

Confusion matrix counts:

The Problem

Leadership is focused on the 90% accuracy, but the retention manager argues the model is underperforming because many churners are not being flagged. You need to explain what the confusion matrix shows and whether the model is good enough for deployment.

Requirements

1. Define what a confusion matrix is and explain each cell using the numbers above.
2. Interpret accuracy, precision, recall, and F1 in this specific churn setting.
3. Explain why accuracy alone may be misleading here.
4. Identify the main business risk from the current error pattern.
5. Recommend 3-4 actions to improve model evaluation or model performance.

Constraints

• The retention team can contact at most 1,200 users per week.
• A false positive costs $8 in unnecessary retention incentives.
• A false negative costs an estimated $120 in lost subscriber value.