Statistics, experimentation, causal reasoning · insight-0033
Refuse to answer a "what if we change X" question with a forecast model
- Evidence type
- practitioner heuristic
- Difficulty
- medium
- Roles
- Data scientist, Data analyst, ML engineer
- Industries
- aviation, retail, energy
Summary: situation, action, reason, boundary
Situation
An accurate demand forecast uses fare as a feature, and a manager asks it to predict bookings after a 5% fare cut.
Action
Separate the questions: use the forecast for "what will happen if policy continues", and use an experiment, a quasi-experimental estimate or an explicit causal model for "what would happen if we intervene". Say which question each number answers.
Reason
Predictive accuracy is measured on data generated by the current policy; interventions create data the model has never seen.
Boundary
Pricing, promotions, staffing and capacity planning built on ML forecasts.
Symptom or risk
The model learned how fares and bookings co-move under the existing pricing policy, so its "what-if" output reflects confounding and may have the wrong sign.
Exceptions
Features that were randomized or set independently of the outcome's drivers can support limited what-if use.
Trade-offs
Causal estimates are slower and noisier than forecasts; they answer the question that was asked.
Pitfalls
Reading SHAP values or partial dependence of a forecast as price response.
Example
Synthetic airline simulator: fare is set higher when demand is expected to be high, so a forecast's fare dependence can be positive while true elasticity is negative.