Segment 500k users into three groups

Divide 500,000 users into three adjacent bands—High, Medium, and Low—ordered by a churn-risk score in which Q1 carries the lowest risk and Q4 the highest. The Low band has to contain no less than 20% of all users. Using the quartile summary, pick cut points that yield the largest expected 90-day revenue.

Risk-quartile summary:

Take expected 90-day revenue per user to be $3 \times \text{Avg\_Monthly\_Spend} \times (1 - \text{90d\_Churn\_Prob})$.

Tasks:

1. Calculate the expected 90-day revenue produced by each quartile.
2. List every valid contiguous three-way split that meets the $Low \ge 20\%$ requirement, then choose the split that maximizes combined expected 90-day revenue across High and Medium; include brief working.
3. For the segmentation you select, state the user count and expected revenue for each band and for the whole population.
4. Suggest segment-specific actions (such as offers or an experimentation plan), and describe how you would track drift that might shift the best cut points.

Overview: The question tests a candidate's ability to segment customers using data, model expected revenue, and weigh churn risk against monetization trade-offs.

See the full Capital One Data Scientist interview experience that included this question.