Causal Inference Is Different in Business

Applied causal inference in business requires pragmatism beyond academic rigor. Three practical rules help data scientists maximize impact: (1) start with the business problem, not the methodology; (2) use simpler associative analysis when causal inference isn't justified by the decision's stakes; and (3) apply 80/20 thinking across all dimensions of a decision rather than over-investing in the causal estimate alone. The key framing is distinguishing 'constructive' decisions (low-stakes, reversible, iterative) from 'final' decisions (high-stakes, hard to reverse), which determines how much rigor is warranted. Real examples illustrate both the cost of over-engineering analysis and the dangers of skipping causal inference when it truly matters.