How would you design Shop-ad ranking?
Company: Meta
Role: Data Scientist
Category: Machine Learning
Difficulty: hard
Interview Round: Technical Screen
Suppose the previous experiment shows that, in some contexts, users are more likely to convert when shown an ad that leads to an in-app Shop rather than an external website.
Now design a ranking algorithm for Meta ads that can **intelligently uprank Shop ads** without hurting user experience or advertiser objectives.
Discuss:
- What the ranking objective should be.
- What features and labels you would use.
- Whether you would use a heuristic boost, a learned ranking model, or a multi-objective system.
- How you would handle heterogeneous advertiser goals, such as direct online conversion versus offline foot traffic.
- How you would evaluate the algorithm offline and online.
- What risks, fairness issues, and cold-start problems you would watch for.
Quick Answer: This question evaluates a candidate's understanding of ad-ranking and multi-objective machine learning, focusing on ranking objectives, feature and label design, and the trade-offs between user experience and diverse advertiser goals.