Forecasting discussions often begin with model accuracy. Production systems force a wider view.
A forecast affects ordering, placement, pricing, labor, and inventory. Those decisions depend on product state, promotions, supply constraints, and timing. The model is important, but it sits inside a larger system of evidence and action.
That is why forecasting architecture should be judged by more than predictive performance. It should also preserve meaning, expose disagreement, and support recovery.