Detect Bots From Comment Spam

The average number of comments for Daily Active Users (DAU) increased because bot accounts posted more comments. Perform a simple calculation to estimate how many bots there are, discuss how to detect bots, how to measure success, and potential trade-offs/impact.

Asked in the Analytical Execution stage. Investigation into sudden spike in daily active user comment counts.

Task

1. Estimate the number of bots from the observed increase in comments per DAU.
2. Explain a practical bot-detection approach using Meta signals.
3. Define how you would measure success after mitigation.
4. Call out likely trade-offs, including false positives and product impact.

Data Available

• comments_events with user_id, post_id, comment_id, timestamp, surface such as fb groups or ig reels
• user_daily_activity with user_id, date, dau_flag, comments_count, sessions_count
• account_integrity_signals with user_id, account_age_days, follower_following_ratio, device_fingerprint, ip_reputation, report_rate
• experiment_logs for any integrity rollout, including treatment assignment and pre-period activity for CUPED adjustment