Detect fake accounts and measure their impact

Scenario

You are an Analytics Engineer at a social media platform. The product team suspects a surge in bot activity and automated fake accounts that inflate advertising metrics and degrade user experience. You have access to a massive event-level dataset containing user login timestamps, IP addresses, device IDs, account creation dates, and ad-click logs.

Task

1. Detection Strategy: Propose a methodology to identify fake accounts. Define at least three behavioral or technical signals (features) that would distinguish a bot from a legitimate user.
2. Impact Measurement: How would you calculate the 'negative impact' of these fake accounts on two specific metrics: (a) Ad Revenue (e.g., invalid clicks) and (b) Daily Active Users (DAU) inflation.
3. Validation: Suppose you implement a removal policy that deletes detected fake accounts. How would you design an experiment (or quasi-experiment) to measure the true impact of this removal on the platform's ecosystem without negatively impacting legitimate user engagement metrics?

Constraints

• Assume the platform logs billions of events daily.
• You must account for false positives—avoiding the deletion of legitimate users is critical.
• The measurement of impact should distinguish between 'organic' noise and 'malicious' bot behavior.

Expected Output

• Describe the technical approach for feature engineering.
• Explain the mathematical framework for attribution (i.e., how do you subtract bot activity from total metrics?).
• Detail the experimental design, including treatment/control group definition and potential confounding variables.