Climbing the causal ladder for fun and profit
summary
Builds a Bayesian structural causal model in PyMC that uses a censored regression to infer latent demand under stockouts, fits with the PyMC do operator, and produces calibrated forecasts that match observed sales. Quantities reported include estimated 20–35 lost sales from stockouts and an estimated 20–30 uplift from a sustained marketing campaign. Shows how interventions (raise stock, change ad spend) and counterfactual scenarios are generated and evaluated, with profit impacts computed from posterior distributions.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 1:47Latent demand estimation
- 7:56Statistical limits in forecasting
- 10:04Pearl’s causal ladder
- 12:30Structural causal model for demand
- 16:45Probabilistic programming implementation
- 19:13Model validation and lost sales
- 21:27Causal interventions for stock
- 24:44Marketing campaign impact
- 26:56Summary
related talks
27mUsing Causal thinking to make Media Mix Modeling
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
42mSolving Real-World Business Problems with Bayesian Modeling
01h 24mBayesianism and Survival Analysis
25mParadoxes in model training and evaluation under constraints
41mPyMC-Marketing: Customer and Marketing Analytics the Easy Way
21mHazards on the Causal Path
36mWhat-if- Causal reasoning meets Bayesian Inference