Design an Ad Delivery System

Question

Design a personalized ad serving system. Cover ad selection and auction mechanics, real-time bidding, targeting based on user signals, frequency capping, and serving latency requirements under 100ms.

Codemia · Accepted Answer

- **Functional Requirements:**  
  - **Ad Selection:** The system must select relevant ads based on user profiles, interests, and behavior.  
  - **Auction Mechanics:** Implement a real-time bidding auction where advertisers can bid for ad placements based on user signals.  
  - **Targeting:** Use demographic and behavioral data to deliver personalized ads.  
  - **Frequency Capping:** Limit the number of times a user sees the same ad to prevent ad fatigue.  
  - **Reporting:** Provide advertisers with performance metrics such as CTR (Click Through Rate), conversions, and impressions.  
  
- **Non-Functional Requirements:**  
  - **Latency:** Ensure that ad serving latency is under 100ms.  
  - **Scalability:** The system should handle millions of ad requests per second during peak traffic.  
  - **Reliability:** The system should have high availability and resilience, with a target uptime of 99.99%.  
  - **Data Privacy:** Comply with data protection regulations (e.g., GDPR) in handling user data.

- **User Base:** Assume Meta has approximately 2 billion active users.  
- **Ad Requests:** If each user generates an average of 10 ad requests per day, that equals 20 billion requests daily, or approximately 230,000 requests per second.  
- **Bids and Auctions:** Assuming an average of 5 advertisers bidding per ad request, we can estimate 1.15 million bids per second.  
- **Data Processing:** For real-time processing, we may need to handle processing for up to 5 million user signals per second to personalize ad selection, considering data updates in real-time.

- **Components:**  
  - **User Signal Processor:** Kafka for managing real-time user interactions and signals.  
  - **Ad Auction Engine:** A microservice built with Node.js for handling bidding logic and selecting ads.  
  - **Ad Database:** NoSQL database (e.g., Cassandra) for storing ad data, user profiles, and bidding history.  
  - **Cache Layer:** Redis to cache frequently accessed data like user profiles and ad inventory for low-latency access.  
  - **API Gateway:** NGINX for routing requests from different clients to appropriate services.  
  - **Monitoring:** Prometheus with Grafana for monitoring system health and performance.  
  - **Load Balancer:** AWS Elastic Load Balancing to distribute traffic across servers.

- **User Table:**  
  - *user_id* (PK)  
  - *interests* (list of strings)  
  - *demographics* (JSON)  
  - *frequency_count* (integer)  
  
- **Ad Table:**  
  - *ad_id* (PK)  
  - *advertiser_id* (FK)  
  - *targeting_data* (JSON)  
  - *bid_price* (float)  
  - *impressions* (integer)  
  
- **Bids Table:**  
  - *bid_id* (PK)  
  - *ad_id* (FK)  
  - *user_id* (FK)  
  - *bid_amount* (float)  
  - *timestamp* (datetime)  
  
- **Access Patterns:**  
  - Fast lookups for user profiles in the User Table.  
  - Frequent updates to the Bids Table during auctions.

- **Real-time vs Batch Processing:**  
  - Choosing real-time processing (Kafka) allows immediate ad selection but increases complexity and costs. Batch processing could be simpler but may not meet latency requirements.  
  
- **Consistency vs Availability:**  
  - Prioritizing availability with eventual consistency in the NoSQL database can improve response times but risks showing stale ad data. Strong consistency would ensure accuracy but could increase latency.  
  
- **Cost vs Performance:**  
  - Using in-memory databases (e.g., Redis) boosts performance but adds cost. Balancing between caching strategies and data retrieval costs is crucial.  
  
- **Ad Personalization vs User Privacy:**  
  - Robust targeting mechanisms may require extensive user data, which could conflict with privacy regulations. Striking a balance between effective targeting and compliance is essential.

Design an Ad Delivery System

Meta

How to Approach This

Sharpen Your Skills on Codemia

Sample Answer

Requirements

Capacity Estimation

Submit Your Answer

Meta Software Engineer Interview Guide

Related Questions

Design a high-throughput Inventory Management System

Design a low-latency Rate Limiting System

Design a fault-tolerant Payment System

Design a fault-tolerant Messaging System

Design Walmart Product Search