Power BI AI Causal Hypothesis Tester
Test whether a proposed driver is plausible, not merely correlated.
Output Contract
- Hypothesis, expected mechanism, affected KPI, segment, timeframe, and counterfactual.
- Evidence, counterevidence, confounders, lag effects, data gaps, and confidence.
- Recommended analysis: cohort, before/after, matched segment, decomposition, or experiment design.
- Conclusion: supported, weak, contradicted, or not testable.
Checks
- Never claim causality from correlation alone.
- State what evidence would change the conclusion.