Causal Effect Estimation in Practice: Lessons Learned from E-commerce & Banking
summary
Claims concrete trade-offs: randomized A/B is ideal but often impossible for finance and e‑commerce, so solutions use inverse propensity weighting with ordinary least squares to retarget observational bias, plus survival analysis to handle staggered treatment and censoring. Practical steps include rule-based control construction (copy campaign filters, exclude overlap), checking confounders versus colliders, and comparing potential‑outcome and structural causal models for stability. Real wins come from simple engineering and domain checks more than fancy models; results require time windows matched to business KPIs and multiple methods for validation.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 2:15Causal estimation overview
- 4:24Business use cases
- 6:52Randomized experiments challenges
- 9:23Addressing lack of randomization in practical scenarios
- 13:46Confounding variables
- 15:52Potential vs structural models
- 18:21KPIs over time
- 21:56Rule-based control groups
- 24:27Practical challenges summary
- 27:20Key takeaways
- 28:03Q&A bias control
- 31:58Retargeting without controls
- 33:25Validity frameworks
- 34:53Funnel KPI accuracy
- 35:57Instrumental variables
- 38:04Final Q&A
related talks
43mBeyond Just Prediction: Causal Thinking in Machine Learning
41mKeynote: Personalization, Bandits, and Causal inference
31mA Gentle Introduction To Causal Inference
41mOn the Path to Causal Inference
38mHow to Enhance Customer Targeting in Marketing
31mLearning And Testing Causal Models: A Property Testing Viewpoint
32mCausal Forecasting: How to disentangle causal effects, while controlling for ...
11mCausal Inference For Socio-Economic & Engineering Systems