Unveiling the Potential: Exploring Upselling, pLTV, and ML for Operations
summary
Built a Scala naive bias module inside a Java/Scala offering tool for a 3M daily‑active app and shows a ~24% conversion uplift in a 10k experiment that scales to production, reducing acquisition cost and covering extra pricing slots. A separate linear‑regression pLTV model trained on flawed test splits hits zero uplift in 30–50k runs and reveals 98% false positives. Emphasis is on robust experiment design, operational alerting, and measuring real impact rather than ML as a novelty.
AI-written summary · about
chapters (5)
- 0:00Upselling and ML in operations
- 1:53Smart offering system with bias
- 9:00Predicting LTV with linear regression
- 12:31Automated merchandising tool
- 16:27ML implementation lessons
related talks
24mUnlocking revenue: newcomer classification for boosted profits
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
13mOptimizing Subscription Services with Uplift Modeling
38mHow to Enhance Customer Targeting in Marketing
23mCustom-made games with machine learning and big data
24mBeyond the P-Value: A Data Scientist’s Guide to Tier-1 Feature Launches
25mHow we bring Performance Marketing to linear TV by using machine learning
42mSolving 3 business problems with ML.NET in 45 minutes