SCOR DS · Plan · caselet
Re-plan now or ride the noise: forecast error leaves its control band
Four weeks of MAPE creep on one product family — noise, or a trend worth re-planning for? The agent prices both answers and recommends pulling the re-plan forward seventeen days.
The situation. The components manufacturer behind the DR-7 drive-module family plans it on a monthly S&OP cadence. Between cycles, the plan is the plan — deviations accumulate until the next meeting.
The signal. The Demand Planning agent tracks rolling 4-week MAPE against a control band. On this family it has drifted from 18% to 31% (band: ≤22%), with a consistent under-forecast bias — actual orders are running ahead of plan, and the bias has persisted for three consecutive weeks. Systems consulted: demand-planning system (forecast, actuals), ERP (open orders, inventory position), and the customer master (which accounts are driving the lift).
The decision. Two honest options, both priced. Ride it to the next cycle: 19 days away, with a projected stockout on two SKUs that would force roughly $38,000 of expedited container moves to recover. Re-plan now: one planner-day plus supplier schedule changes, about $6,000. The agent recommends the early re-plan and shows its work — which accounts, which SKUs, and how much of the lift is repeat-order behavior rather than one-off spikes.
The write-back. Revised consensus forecast published to the planning system; updated requirements released to the two affected suppliers through the ERP.
The outcome.
| Metric | With early re-plan | Riding the cycle |
|---|---|---|
| Signal → re-plan | 2 days | 19 days |
| Expedite cost avoided | ~$32,000 net | — |
| Projected stockouts | 0 | 2 SKUs |
The takeaway. A forecast isn’t wrong when it misses — it’s wrong when it keeps missing in the same direction and nobody re-plans.
Representative scenario; figures are synthetic. SCOR is ASCM’s framework.