Demand Forecasting

The statistical and judgemental process of predicting future SKU-level demand to drive inventory, replenishment, capacity and labour planning across the warehouse network. The discipline matters because every other warehouse metric — fill rate, days of supply, working capital, shift roster, replenishment workload — is downstream of how well demand is predicted and how cleanly that prediction flows into stock policy. A warehouse that runs Demand Forecasting well treats it as a continuous loop: clean history, statistical baseline, judgemental override, signed-off S&OP plan, executable replenishment policy, and measured forecast accuracy that closes back on the model.

• History is cleaned of stockouts, promos and one-offs before modelling.
• Statistical baseline is the starting point — not a guess from sales.
• Judgemental overrides are documented, signed off, and tracked for bias.
• Forecast accuracy (MAPE, bias) is measured every cycle and fed back to the model.
• Stock policy parameters are derived from the forecast — not set in spreadsheets and forgotten.

• Refresh safety-stock parameters on the same cadence as the forecast — not annually.
• Track forecast bias by product family — persistent bias signals a process problem, not a model problem.
• Tie S&OP plan to executable warehouse capacity — promises that exceed dock or labour capacity are not plans.
• Recalculate EOQ when input costs change materially — frozen EOQs become wrong quickly.
• Report fill rate by channel and customer tier — averages hide important misses.

• Forecasting at the wrong level — too high hides SKU mix, too low is statistically meaningless.
• Sales overrides accepted without governance — forecast becomes sales target plus a wish.
• Safety stock set once and forgotten — every demand shift erodes service level.
• S&OP signed off but never reconciled to actual — no learning loop.
• Fill rate measured at order level only — line-level customer pain is invisible.

• Grocery — short shelf life forces tight forecast accuracy; daily replenishment cycles.
• Fashion — high obsolescence risk, season-based planning, life-cycle curves.
• Industrial spares — long tail, intermittent demand, Croston or bootstrap methods.
• Pharma — regulated supply continuity drives high service-level targets.
• Subscription DTC — known customer base, deterministic forward demand.