Demand Forecasting
Forecasts feed anticipated future demand into the MRP engine so planning looks beyond firm orders. A forecast behaves like demand of type Forecast in the netting calculation, letting SDI warn you about shortages for demand that hasn't been booked yet.
Where to Access
- Forecasts —
/mfi/forecasts(also from the SDI Dashboard → Demand Sources → Forecasts).
The page has three tabs: Forecasts, Batches, and Accuracy.
Forecasts tab
A list of individual forecast records — Item, Location, Qty Required, UOM, Date Required, Status, Created. Filter by Item, Location, or Status. You can edit or delete a record, or select rows for bulk delete.
Creating forecasts
Click Create Forecast and pick a method:
From Historical Sales
Generate forecasts by projecting a source period of past sales onto a target period. Options:
- Period type — Daily, Weekly, or Monthly buckets.
- Multiplier — scale the projection up or down (e.g., 1.1 for expected 10% growth).
- Use seasonality — shape the forecast using historical seasonal patterns.
- Use trend — extend an observed growth/decline trend.
- Items / Location — limit to specific items and a location.
Quick-pick helpers set common source→target pairs (last month, last year's same month, last year's matching quarter) targeting next month or next quarter. You can preview the generated quantities before committing.
Period Forecast
Enter a known total demand for a period and distribute it across buckets using a distribution:
- Even — spread equally.
- Front Loaded (60/40) — heavier early.
- Back Loaded (40/60) — heavier late.
- Seasonal — follow historical seasonal patterns.
Manual Entry
Create a single forecast record: item, location (optional), quantity, date required, and notes.
Bulk Import
Add many forecast rows at once in a table — item, location, quantity, and date per row — and create them together.
Batches tab
Every generation run is recorded as a batch so it can be reviewed or undone as a unit. Each batch shows Name, Type (Historical / Period), Target Period, Period Type, forecast count, Total Qty, and Created date. You can:
- Delete batch only — keep the generated forecasts, remove the batch record.
- Delete batch and forecasts — remove both together.
Accuracy tab
Once actual demand has materialized, click Calculate Accuracy to score forecast quality.
Summary metrics
- Overall MAPE (Mean Absolute Percentage Error), graded:
- ≤ 10% — Excellent
- ≤ 20% — Good
- ≤ 30% — Fair
- > 30% — Poor
- Forecast Bias — positive means you tend to over-forecast, negative means you under-forecast.
- Items Measured — how many items have accuracy data.
Per-item table
For each item: MAPE, Bias, periods measured, trend direction (Growing / Declining / Stable), and an accuracy grade.
Typical workflow
- Generate a baseline with From Historical Sales for the upcoming quarter, enabling seasonality and trend.
- Adjust with Period Forecasts or Manual Entry where you have specific knowledge (a promotion, a known large order).
- Let MRP net the forecasts so the SDI Dashboard plans ahead of firm orders.
- After the period closes, run Calculate Accuracy and review MAPE/bias to tune the next round.
Tips
- Lean on seasonality and trend for items with clear patterns; use flat multipliers for steady movers.
- Keep batches tidy — delete superseded generation runs so the forecast set stays clean.
- Treat a persistent bias as a signal to adjust your multiplier (consistently over-forecasting? lower it).