Ad Aggregation Pipelines
Difficulty: Hard | Topic #34
What to Learn
High-throughput event ingestion, server-side click deduplication, windowed counting (hourly/daily rollups), fraud signal injection, lambda architecture for real-time + batch accuracy.
Resources
- Hello Interview: Ad Click Aggregator Breakdown ↗
- Gaurav Sen: Realtime Advertisement Clicks Aggregator ↗
- System Design Interview: Ad Click Aggregator (Kafka + Idempotency) ↗
Covered by Problems
| Problem | Difficulty | Link |
|---|---|---|
| Price Tracking Service | Medium | → |
| Ad Click Aggregator | Hard | → |
Key Concepts to Master
- High-throughput ingestion via Kafka with partitioning by ad_id
- Server-side deduplication using a seen-set (Redis bitmap or bloom filter)
- Windowed aggregation (5-min, hourly, daily) with Flink
- Late data handling with watermarks and allowed lateness
- Batch reconciliation layer for billing accuracy correction