Design ad click tracking and dashboard at scale

Problem Statement

Design a distributed system to track ad clicks for a major advertising platform. The system must process millions of click events per second and provide real-time metrics (e.g., clicks per ad, click-through rates) to advertisers via a dashboard.

Functional Requirements

1. Ingestion: Capture incoming click events with attributes: ad_id, user_id, timestamp, geo, and device_type.
2. Aggregation: Provide real-time reporting of click counts, grouped by ad_id over time windows (e.g., 1-minute, 5-minute, and hourly buckets).
3. Querying: Advertisers should be able to view current-day metrics with low latency (< 500ms).

Non-Functional Requirements

1. Scale: Handle 5 million clicks per second during peak hours.
2. Reliability: Ensure minimal data loss (at-least-once delivery is acceptable, but minimize double-counting).
3. Consistency: Dashboard counts should be accurate within a few seconds of the actual event.
4. Storage: Data must be retained for at least 90 days for historical reporting.

Deliverables

• Provide a high-level system architecture diagram (described in text).
• Define the data model for the ingestion pipeline and the storage layer.
• Explain how you would handle high-volume write traffic and partitioning strategies.
• Address how the system handles late-arriving events or out-of-order data.