How would you debug a sudden drop in a key metric?

This tests your ability to systematically debug complex data systems, not just list analytics quirks. A great answer first validates the drop against a source of truth (e.g., backend DB), segments data to find patterns (channel, device), and then distinguishes between data loss and misattribution. Based on the pattern, you can form specific hypotheses. A major red flag is jumping to niche technical causes without first doing this broad analysis to narrow the problem space.