Customer Lifetime Value: Models, Metrics and a Multitude of Uses
summary
Claims that probabilistic CLV models recover missing future purchases from sparse histories using real credit-card transactions and open tools: modified BG/NBD for joint purchase-rate and churn, lifetimes library and lifelines for fitting, and per-region fits that find a 16% lower 24‑month spend in the South versus West. Data scales (billions of transactions), short cohorts (180 days) force heavy imputation, and practical notes show models fit ~100k customers on a laptop in seconds.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:30Customer-centric analysis matters
- 6:50Lifetime spend with real data
- 11:48Incomplete data in CLV models
- 16:28Transaction counts and churn jointly
- 20:01Customer heterogeneity in CLV
- 25:15Applying models to business cases
- 28:53Key takeaways from CLV workflows
- 30:16Q&A
related talks
44mEating Humble Py: From Toy Problem to Real-World Solution in Predicting CLV
36mImplementing and Training Predictive Customer Lifetime Value Models in Python
34mModelling and predicting customer lifetime value
34mMaximizing Credit Card Profitability with ML–Driven CLV
35mUsing Survival Analysis to understand customer retention
45mKnow Your Customer
24mUnlocking revenue: newcomer classification for boosted profits
45mLifetime Value and the Actual Costs of Acquiring Customers