Eating Humble Py: From Toy Problem to Real-World Solution in Predicting CLV
summary
Claims a pipeline that moves from toy data to real CLV estimates and shows what actually breaks: cohort averages, RFM clustering, and a simple spend×frequency×lifespan formula give intuition but fail on one‑time buyers and skewed data. Implements probabilistic BG/NBD + beta geometric / gamma-gamma models, regressions and classifiers (random forest, XGBoost, AutoML), uses the lifetimes package, extensive log transforms and clustering, struggles with returns/discounts, profit vs revenue, messy one‑year datasets, and stops short of full production.
AI-written summary · about
chapters (12)
- 0:00Predicting customer lifetime value
- 2:16What is CLV and why it matters
- 8:55Simple CLV formulas limits
- 11:22Cohort analysis for segmentation
- 13:49RFM analysis applications
- 17:19Probabilistic CLV models
- 21:33Machine learning for CLV
- 26:51Real-world CLV data challenges
- 30:01Handling returns and discounts
- 34:34Recalculating CLV insights
- 36:41Conclusion
- 37:02Q&A: CLV and churn
related talks
36mCustomer Lifetime Value: Models, Metrics and a Multitude of Uses
36mImplementing and Training Predictive Customer Lifetime Value Models in Python
34mModelling and predicting customer lifetime value
23mCustom-made games with machine learning and big data
24mUnlocking revenue: newcomer classification for boosted profits
34mMaximizing Credit Card Profitability with ML–Driven CLV
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
45mKnow Your Customer