Correlation versus causation

Data science is full of patterns: users who do X also do Y, regions with more A have less B. The crucial question is whether changing one thing would change the other. Mixing up correlation and causation leads to bad decisions, from wasted marketing budgets to harmful policies.