Design an AB test system.

An A/B testing system requires several key components: a way to define experiments (variants, target audience, metrics), a mechanism for assigning users to variants (e.g., using a consistent hashing algorithm based on user ID), a way to serve the correct variant to the user, and a system for collecting and analyzing experiment results. Data models would need to represent experiments, user assignments, and metric data. The core logic involves user bucketing to ensure consistent assignment and robust data pipelines for tracking and aggregation.