Iron: Keyword Grouping Model for Text Ads Bidding in Paid Search
summary
Iron trains a Gradient Boosting Machine on eBay search features, paid-search performance and Google feedback to predict per-keyword conversion rates, then computes expected value using predicted CR and price and groups keywords by that value; group-level bids are set and adjusted with a market-feedback auction algorithm called Brutus. Deployed A/B test holds cost steady while revenue rises, delivering about a 27% ROI uplift versus category-only grouping.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:19How Google Paid Search works
- 3:30Keyword Bidding Process
- 4:24Algorithm Structure
- 4:59Pipeline
- 6:22Results
related talks
39mAI-Powered Search in Modern E-Commerce
39mRanking Listings with Machine Learning and ElasticSearch
44mRobust AI Search Ranking for Radical C2C Marketplace Growth
54mOn How Machine Learning and Auction Theory Power Facebook Advertising
15mAI-Driven Application Optimization
33mMachine Learning to moderate ads
29mAI and ML for Computational Advertising
21mRise of the Machines: Removing the Human-in-the-Loop