Mobile Dev Memo Podcast Season 5, Episode 23: Meta's AI advertising playbook (with Matt Steiner)
40 snips
Jun 3, 2025 Matt Steiner, Vice President of Monetization Infrastructure at Meta, dives into the world of AI advertising. He discusses the intricate ad ranking and selection processes at Meta, revealing the evolution of AI tools like Lattice and Andromeda. Steiner explains the significance of auction theory in ad systems and how it influences user engagement. He also covers the role of sequence learning in optimizing ad performance and explores Meta's ongoing AI advancements. Lastly, he highlights the importance of talent and testing strategies for advertisers leveraging Meta's platform.
AI Snips
Chapters
Books
Transcript
Episode notes
Meta's Multi-Pass Ad Ranking
- Meta's ad ranking system first quickly filters a large set of ads with a lightweight model.
- Then it applies a more precise heavyweight model to a smaller set before running auctions for final ad selection.
Dual Goals of Ad Ranking
- Ad ranking balances advertiser goals and user interest to maximize value.
- It predicts not only clicks but also user satisfaction and relevance for better ad selection.
Ad vs Content Ranking Differences
- Ad ranking differs from content ranking because ads include bids influencing which ad is shown.
- Friends and organic content ranking focuses on interest without monetary bidding influence.