Exponential Smoothing Forecasting

The Exponential Smoothing Forecasting report is the demand-forecasting tool production planners and S&OP leads use to project future demand from historical sales patterns. It uses the exponential smoothing method to extend recent demand history into a forward-looking forecast, so you can plan procurement and production ahead of demand instead of chasing it.

Where to find it

Manufacturing > Reports > Exponential Smoothing Forecasting.

Filters

Filter	What it does
Company	Mandatory. Scopes the historical demand data to one company.
From Date / To Date	The forecast window: the period the report will project demand for. Defaults to today through today + 12 months.
Based On Document	The source document the forecast learns from: Sales Order, Sales Invoice, Delivery Note, or Quotation.
Based On	The numeric field used as the demand signal: Qty (units shipped or ordered) or Amount (monetary value).
Based On Data ( in years )	How many years of history feed the forecast: 3, 6, or 9. Longer history dampens noise; shorter reacts faster to recent shifts.
Periodicity	Monthly, Quarterly, Half-Yearly, Yearly. The granularity at which both history and forecast are bucketed.
Smoothing Constant	The exponential smoothing parameter, between 0.0 and 1.0. Lower values weight history more (smoother forecast); higher values weight recent periods more (responsive forecast). 0.3 is a sensible default.
Item Code	Scope the forecast to a single item.
Warehouse	Scope to a specific stock location.

Exponential Smoothing Forecasting filter panel

What the report shows

Column	What it shows
Item Code	One row per item in scope.
Period columns (history)	One column per past period showing actual demand from the source document.
Period columns (forecast)	One column per future period in the From Date - To Date range, showing the projected demand.
Total Quantity	A bold subtotal row showing aggregate forecast across items.

A chart at the top of the report plots actual demand against forecast over time, so the eye can spot under-forecast or over-forecast bias quickly.

Exponential Smoothing Forecasting output with demand vs forecast chart

Reading the numbers

The forecast for the next period is largely driven by the most recent few periods of actuals plus the smoothed trend.
A flat forecast with volatile actuals = the Smoothing Constant is too low. Increase it (e.g. 0.5) so the model reacts more.
A forecast that shadows the actuals exactly = the Smoothing Constant is too high (close to 1.0). Decrease it to dampen noise.
A persistent gap between forecast and actuals in one direction (always under, always over) is bias. The model is missing seasonality or a trend break; consider re-running with a different Based On Data ( in years ) value to capture the new pattern.

How to use the forecast

The Exponential Smoothing Forecasting report is read-only: it does not auto-create Production Plans or Material Requests. The intended workflow is:

Run the report at the start of each planning cycle (typically monthly).
Use the forecast quantities as input when creating a Production Plan: open the plan, populate items manually with the forecast quantities, then click Get Material Request and Get Work Order to scaffold the downstream documents.
Re-run the forecast as new actuals come in; the forecast for the next period will refine.

Common pitfalls

Sparse history produces unstable forecasts. Items with fewer than 6-12 historical periods of demand should be planned manually, not from this report.
Cancelled documents are excluded from the historical baseline. If you frequently cancel and re-issue, consider switching Based On Document to Sales Invoice or Delivery Note for a more stable signal.
Item-level scoping over-fits when volumes are small. For low-volume items, aggregate to Item Group or use a longer history window.
Periodicity mismatch. Forecasting Quarterly when planning Monthly hides intra-quarter peaks. Match the report's periodicity to your planning cycle.
Forecasts in years with no history are zero. The model needs prior actuals in the same Periodicity bucket to extrapolate; for new items, start with a manual estimate.