Skip to main content

Use the Forecasting Workbench

The Forecasting Workbench is where planners review demand forecasts per item and apply human judgment when history alone is not enough. Viewing needs View Supply Chain Planning; overrides and like-item need Manage Forecasts.

Review an item's forecast​

  1. Go to Demand Forecasting → Forecasting Workbench. Open Planning Overview first if you want to see what needs attention.
  2. Search for an item and choose how many days ahead to show, or open one from Workbench items (items with open exceptions and configured forecasts). Use the filters and Clear filters to narrow the list.
  3. On the item page, read the chart of history, statistical forecast and consensus forecast. A Cold start badge means too little history for the statistical methods; Sufficient means enough.
  4. Check Dependent demand if the item is a component and Manufacturing is licensed. See Run BOM explosion.

Forecasts are recalculated every night, and the best-fitting method is chosen automatically for each item.

Override a forecast​

  1. Open the item in the workbench.
  2. In Add override, choose the date, enter the quantity, pick a reason and add a note.
  3. Click Save override.
  4. To withdraw one, find it under Active overrides and click Deactivate.

The consensus forecast uses your number for that day. A newer override for the same day replaces the earlier one, and the original system forecast is kept for reference. An override always wins over a promotion uplift on the same day.

Forecast a new item from a similar item​

  1. Open the new item. The Like-item matching panel appears when it has little history.
  2. Review the candidates: items in the same product group with enough history and the closest price.
  3. Optionally enter the expected daily volume for the new item.
  4. Click Use shape next to the item whose demand pattern fits.

The new item's forecast follows the chosen item's pattern, scaled to your expected volume, or to the matched item's own volume if you leave it blank.

Methods at a glance​

Forecasts use classic statistical methods: exponential smoothing, Holt-Winters and Croston's method for intermittent demand. They are tested on recent history per item. Very new items use a simple average until enough history exists. There is no AI or machine learning.

:::tip Spare parts and slow movers Intermittent items often fit Croston's method better than smooth methods, and the nightly job picks it automatically when it tests best. :::

Next step​