Most analytics programmes optimise for the report. The organisation needed an action. When recommendations stop at a deck, the interval between signal and write-back stays long — and contested.
A production-minded recommendation carries three things: evidence, a stated assumption set, and a destination. Evidence without a destination is commentary. A destination without evidence is improvisation.
Closing that loop means modelling how entities decide, capturing logic as something executable, and supervising the first deployments on site. Analysis and action are one system — or they are two budgets that never meet.