Causal decision intelligence playbook
The goal of causal analytics is not a prettier dashboard. It is a defensible decision about what to ship, stop, scale, or investigate next.
Frame
State the decision, causal question, unit of analysis, and risk of acting on a false positive.
Estimate
Choose the simplest design that fits the data: randomized test, holdout, matching, difference-in-differences, or sensitivity review.
Decide
Convert effect size into expected value, operational risk, and a rollback or monitoring plan.
Decision record template
- Evidence summary and estimator used.
- Assumptions and robustness checks.
- Action, owner, date, and revisit trigger.