Online customer targeting as a classification problem
summary
Predicting conversions from retargeting data is framed as a supervised classification problem with heavy class skew and censoring and Maxpoint supplies a concrete dataset and infrastructure story. Key claims include working at massive scale with 60 billion bid opportunities per day and 12 trillion calculations per day and using Python tooling like scikit-learn (logistic regression, ensembles, ROC evaluation). Practical issues: time windows, survival framing, feature engineering, regularization, parallel training and whether models actually reduce CPA.
AI-written summary · about
chapters (15)
- 0:00Max Point and online ads
- 7:31ML in ad targeting
- 10:21What is retargeting and why it matters
- 14:35Retargeting classification pipeline
- 25:41Max Point ML takeaways
- 26:48Q&A feature interactions
- 27:57Q&A features and data
- 29:21Q&A demand-supply balance
- 30:43Q&A model training
- 31:51Q&A descriptive models
- 32:14Q&A privacy and features
- 33:19Q&A retargeting with cookies
- 35:47Q&A conversion funnel
- 36:29Q&A online learning
- 37:13Q&A campaign optimization
related talks
38mHow to Enhance Customer Targeting in Marketing
30mPredictive Analytics in Social Media and Online Display Advertising
19mSession: Re-Engagement In 2026: How To Design & Scale Winning Retargeting Campaigns
40mNext Best Action: Maximizing Customer Value
31mLarge Scale CTR Prediction Lessons Learned
21mDemystifying Machine Learning for Programmatic UA & Retargeting
28mAutomating SEA Retargeting for Smarter Audience Engagement and Higher Conversions
32mClassifying the variety of customers online engagement for churn prediction