Predicting content virality to drive social traffic
summary
Conde Nast builds an AI system called Falcon that scores every article using models trained on past editor-selected posts and live site signals to predict next-hours pageviews and recommend whether and when to post on Facebook and Twitter; it treats biased log data with offline-policy techniques and keeps editorial overrides. Deployment enables 20% more posts, drives +50% social traffic year-over-year for an initial brand, lifts first-48-hour traffic +35% and older-article traffic +120%, and now runs across 13 publications.
AI-written summary · about
chapters (14)
- 0:00Intro
- 0:22Traveler
- 0:46Digital Scale at Condé Nast
- 1:19Social Traffic
- 3:2920% of Articles get (almost) 80% of Traffic
- 4:48Predicting Virality
- 5:24Incremental Traffic
- 7:01Learning a policy
- 9:03The Solution
- 10:05Falcon is an Al solution that
- 10:29Falcon Features
- 12:26YOY Results on One Brand
- 14:26Two more Metrics
- 16:34Roadmap
related talks
22mGenerating a Billion Personal News Feeds
03mServing Billions of Personalized News Feeds with AI
41mNews in a Platform World
10mSocialSpider AI: Build Your AI Social Media Team
18mTim Shokair, Condé Nast | Databricks Data + AI Summit 2024
15mAI Content Creation: How to Create Signal Instead of Noise
54mOn How Machine Learning and Auction Theory Power Facebook Advertising
09mHow AI Changed Content Creation for Technical Founders