# Churn: From prediction to prescription ## Utils + Instruccions sobre com editar aquest tipus de fitxers: https://hackmd.io/features?both + Blog interessant: https://motifanalytics.medium.com/bringing-more-causality-to-analytics-d378108bb15 ## 0. Problem definition ### Churn (or Conversion) ### Type of causal problem ## 1. Problem understanding and Explainability ### Model the problem (as a prediction?) ### With the help of the model (explainability) find all factors afecting the outcome #### Expert knowledge role ### Why prediction is not enough ## 2. Causal Inference ### Prescriptive analytics ### Define the causal graph #### Expert knowledge role ### Identify the causal query ### Estimate #### Experiments vs observational data ### Validate? ## 3. Optimal policy ### Exploration vs Exploitation