Design an ad frequency capping system
Company: Netflix
Role: Software Engineer
Category: System Design
Difficulty: hard
Interview Round: Onsite
Design a real-time **ad frequency capping** system for an ads platform.
When the ad server is deciding whether to show an ad, it must enforce caps such as:
- Per user **per campaign**: at most `K` impressions in the last `T` hours (rolling window).
- Optional: per user per advertiser, per user per creative, etc.
Requirements:
- Very low latency on the serving path (single-digit ms budget).
- Very high QPS (millions/sec) and large cardinality (users × campaigns).
- Events are generated on every impression and may arrive late/out of order.
- Define what consistency you need (strict vs best-effort) and justify tradeoffs.
- Support multiple time windows (e.g., 1 hour, 24 hours, 7 days).
- Provide APIs for: (1) check-and-reserve (decision time) and (2) impression logging.
Explain data model, storage choices, write/read paths, deduplication, TTL/expiry, sharding strategy, and how you monitor/alert on correctness and system health.
Quick Answer: This question evaluates competency in distributed systems architecture, real-time stateful services, high-cardinality time-series data modeling, API design for low-latency decision paths, and operational monitoring including consistency and latency trade-offs.