Forecasting

Forecasting can be performed in two different component groups in M3: Forecasting or Sales Statistics and Performance.

The forecast can be entered manually or calculated mathematically in both areas.

If you choose to forecast in Sales Statistics, you will have a greater freedom to choose in what dimensions you would like to work (i.e. sales per region and item). It's possible to build a dataset with up to 6 keyfields of your own choice, but in Forecasting, you forecast for the fixed combination of item/warehouse.

If the choice is to use Sales Statistics, you have to transfer the values to the Forecasting to be able to trigger the material plan. This means that if your dataset applies for sales per region and if you have more than one production plant (warehouse)/facility, you will have to distribute your original dataset to a dataset per production plant before transferring the values. This means that if we have two factories and we sell on five markets, we budget our sales per item and market in the first dataset then we have to distribute on how much of each item we want to produce per factory. We distribute one dataset to two. The next step is to transfer the values in the dataset to each warehouse. The transfer is only done from Sales Statistics to Forecasting, not in the opposite direction.

The triggering of the material plan can be done either by manual or mathematical forecast or a master demand schedule. You choose from what source your demand should come; this is set on your forecast logic (see documentation on Forecasting and MRP). It is your decision if you forecast on finished items, semi-finished, or raw material.

Trend and seasonal affects can be taken care of with predefined curves or by letting the system calculate values or index. Trend and seasonal curves can be attached to each item group. For product group, it's possible to attach seasonal curve. These curves have to be defined per division if it's MUC installation. (Different divisions probably means different markets and, accordingly, different trend and season)

It is possible to use several different automatic forecast methods and letting the system select the one with least error for next period (comparing forecasts).

This documentation will focus on the manual and mathematical forecast done in Forecasting and how to transfer from Sales Statistics.

General rules for forecast selection

What type of forecast is suitable to use? How do I know when to use manual or mathematical forecast? Should I use Master Demand Schedules (MDS) or not?

This list shows the types of forecast that are available:

See Manual Forecast

See Family Item

See Competing Forecasts

See Mathematical Forecasts

See Mathematical Forecast with Seasonal Influence

See Mathematical Forecast with Trend Quantity

See Forecast with Trend Factor

If you have commodity items with many independent demands and short accumulated lead times (planning time fence), it's possible to use mathematical forecast. If the planning time fence for an item is 3 months, it is not recommendable to use a mathematical forecast, because we recalculate the forecast for the current period each period. Doing this will have the effect that we don't have time enough to buy raw material and produce semi-finished items in time to produce according to the forecast. For purchased items, it may be advisable to use mathematical forecast, depending on the lead time.

Whether to use MDS or not, is more or less a question of who is making the forecast: The sales or production department. If the company is big and communication between sales and production is complex, it's likely to be the sales department. It's also possible to use the MDS as an extra step for the production to adjust the forecast to suit production before updating the material plan.

Example: Let's say that the forecast is 1000 in every period one year, and the economical order-quantity is 3000, then the MDS might be 3000 in period XX01, XX04, XX07 and XX10 instead of 1000 per month.

The forecasts methods are defined in 'Forecast Method. Open' (FCS300) or 'Sales Forecast Method. Open' (OSS430).

Before you start

The following prerequisites must be met:

  • Define a system calendar per division in 'System Calendar. Open' (CRS900). You create it by pressing F13 twice and entering a from and to value and than F13 again.
  • Define a system period for each period type in 'Company. Connect Division' (MNS100/F). If MPM run in period-type 1 and the number of periods per year is 12, this means that production have period length of one month. COM might have period type 2 and the number of periods is 52, meaning weeks. MPM period is the same as forecast period.
  • In 'System Calendar. Open Period' (CRS910) create a period table for every year and period type used in the installation.
    Note: If you want to use weekly forecasts, you must create a period type 9 table. This table must follow the standard week definition in a regular calendar, starting day 1 in the week on a Monday.
  • Make sure that your item is connected to the right Item and Product group in 'Item. Open' (MMS001/E).
  • If you are using the forecast engine in the sales statistics, refer to the following documents:

    See

  • In 'Item. Connect Warehouse' (MMS002/E), connect the following to your item:
    • Planning method

      Select MRP or RoP.

    • Period frame

      The period frame has to be at least as long as planning horizon.

    • Planning policy
    • Planning horizon

      The planning horizon defines how far in future to forecast.

    • Demand time fence

      The number of days from today in the future where no forecast will appear in the material plan (if you choose to use it this way) 'Forecast Logic. Open' (FCS305/E).It is used with auto allocation as well as an allocation time fence overrules the value in (CRS701).

    • Forecast method
    • Forecast logic

      This will define how orders will consume the forecast, when appearing in the material plan.

Related topics