Demand classification, intermittent-item forecasting (Croston), seasonality detection, and dead-stock flagging.
How this forecaster works: Looks at 180 days of paid orders. Each SKU gets classified by how it actually sells:
SMOOTHregular, predictable — use velocity forecastERRATICregular pace, variable qtyINTERMITTENTsporadic sales — uses Croston's methodLUMPYrare + unpredictable qtyDEAD0 sales in last 180d or flagged do-not-restock ⚙️ changeNEW<30d of history
Reorder tier (critical/urgent/soon/ok) is on-order-aware:
Critical = action needed now — out of stock with no PO, or on-hand+on-order is still below reorder point.
Urgent = on-order exists but is insufficient to cover demand during lead time, or 7d to ROP.
Soon = covered by an existing PO, or 30d cover.
OK = ample.
Items with 0 forecast (one-off sales, no recent activity) skip the tier system and go straight to Dead.
Dead stock is hidden from the main list — click the "Dead" KPI above the table to view/override.
YoY (year-over-year) compares this 90d window to same period 1 year ago — catches seasonal items.
Intermittent forecasting (Croston's method): forecast = avg sale size ÷ avg interval between sales. Much better than naive velocity for spare parts.
Reorder Recommendations—
Preparing forecaster…
Will start in a moment.
0selected for reorder
📤 Create Draft RMAs from Dead Stock
One RMA per supplier. Adjust quantities, vendor emails, and per-line reasons before creating. RMAs save as drafts — review them in /rmas and send to vendors when ready.