Demand planning is the nervous system of a physical business. Replace it clumsily and you inherit stockouts, overstocks, and a planning team that no longer trusts the tool.
Start with the decisions, not the data
Before touching a forecasting model, map the decisions the plan feeds: purchasing cadence, production runs, transfer orders, promotional lift. Every model choice should be justified by a decision it improves.
Run the new system in parallel — for real
Shadow mode isn't shadow mode if planners aren't comparing outputs weekly. Give the new system a scorecard and a review ritual from week one.
Change the roles, not just the tools
- Split forecasting (statistical) from planning (judgement).
- Give commercial teams a structured way to input signal, not spreadsheets.
- Move the S&OP conversation from 'whose number is right' to 'what are we going to do'.
Cutover is a decision, not a date
Set clear exit criteria — forecast accuracy, planner confidence, exception volume — and cut over when they're met. A calendar-driven cutover into a system nobody trusts is how these programmes fail.