Walk From Correlation to Causal Question
Identify when a data science question requires causal design rather than predictive modeling alone.
Prediction or cause? Users with onboarding calls retain more. The business wants AI to allocate calls. The observed pattern may reflect selection bias, not treatment effect. Causal graph lens Treatment -> Outcome, plus confounders Draw the treatment, the outcome, and variables that may influence both. If the action changes the outcome, predictive fit is not enough. Shortcut Predict retention and call the riskiest users. The team estimates lift instead of mistaking selection for impact. Prediction asks who is likely. Causality asks what the action changes. 01 Classify 02 Graph 03 Design Classify the question Prediction vs intervention
Sign up free — one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in