Causal Inference for Data Scientists - Moving from Association to Intervention

Learn to distinguish between correlation and causation in data science through this comprehensive tutorial on causal inference using Microsoft's DoWhy library. Explore the fundamental differences between association (what SHAP reveals) and intervention (what causal inference provides) using a practical advertising campaign example with the Criteo dataset. Discover why traditional conversion metrics can be misleading when ad platforms target already-engaged customers, and master techniques to estimate true causal effects rather than just correlations. Work through real-world scenarios where simple difference calculations overestimate effects by 13% due to confounding variables, and learn to calculate Individual Treatment Effects to identify "persuadable" customers worth targeting. Apply these methods to optimize marketing spend dramatically - the tutorial demonstrates how to reduce ad expenditure by 87% while maintaining conversion rates. Follow along with practical, block-by-block coding examples in Google Colab using the same causal inference approaches employed by major companies like Uber, Netflix, and Amazon for marketing optimization.