Design a CPA system for ad bidding
Company: Apple
Role: Machine Learning Engineer
Category: ML System Design
Difficulty: medium
Interview Round: Technical Screen
You are designing an ads bidding/optimization system where advertisers care about **CPA (cost per acquisition)**.
Describe how you would design a system to **optimize toward a target CPA** while still participating in an auction.
Cover:
- What the objective is (platform vs advertiser) and how CPA is defined.
- What models you need (e.g., CVR, value, conversion delay).
- How to compute bids from predictions (e.g., bid shading, target CPA control).
- How you handle delayed/attributed conversions, sparse data, and fraud.
- Offline evaluation and online A/B testing metrics for CPA systems.
Quick Answer: This question evaluates a candidate's competency in ML system design for online advertising, covering modeling conversion probability and value (e.g., CVR and conversion delay), bid computation, and operational challenges like delayed/attributed conversions, sparse data, and fraud mitigation.