Show TOC

 Demand Influencing Factors (DIF)

Purpose

To achieve an automatic replenishment process that creates optimized order quantities, you need an accurate prediction of future demand. This is normally done by using forecasting algorithms that calculate future demand volume based on historical data. The following factors impact the quality of the demand forecast

  • High quality of the historical data on which the forecast is based

  • Selection of the most appropriate forecast model that describes the demand characteristics of a specific produce

  • The effect of past and future external influences on demand (these are also known as Demand Influencing Factors, or DIFs)

The Demand Influencing Factors (DIF) module provides input data for the automatic replenishment process. It allows the demand forecast to take into account the effect of predictable external factors on demand. It also allows you to review, manage, and adjust the DIFs as necessary.

Some examples of DIFs are:

  • Promotions and advertising campaigns

  • Public holidays or vacations (Easter, Thanksgiving, Christmas)

  • Sports events (Super Bowl, Soccer World Championship)

  • Sales prices and price changes

  • Unusual weather conditions (heat waves, hurricanes)

Implementation Considerations

In addition to the examples mentioned above, you can define DIFs for any other factors that can be quantified, maintained, stored, and used to explain the sales or demand of a product. Such factors could include the aggregated store sales, total sales area, shelf facings, store traffic, and so on.

Note Note

You should choose DIFs carefully (especially the metric DIFs) during implementation so as not to overload the system or decrease forecast accuracy.

End of the note.

Features

The influence of DIFs is related to the time during which they are in effect:

  • Event DIFs are external events that are limited to a specific and usually short time interval; that is, they have discrete occurrences. Examples include promotions, advertising campaigns, or any calendar-related events such as public holidays or sports events. The occurrences can be either yes-or-no events (that is, they occur or do not occur, such as public holidays) or they can have a certain intensity (for example, hurricanes or heat waves). Those types of DIF are generally valid for a larger selection of locations and products, and may not be product specific.

  • Slow changing DIFs have a particular value that changes only occasionally; for example, a product’s sales price. Such DIFs are generally valid for only one product at a single location.

  • Fast changing DIFs change on an almost continuous basis that can be described by a time series rather than by individual occurrences or changes to a DIF; for example, aggregated store forecasts as a predictor for the distribution center forecast.

For new store openings, where no historical data exists for predicting future demand, you can define a reference store. The system can use the historical data from the reference store to forecast demand for the new store. You can also define reference products.

As part of the preparation of the forecast calculation, you can specify that exceptionally high or low values (outliers) be automatically excluded. Since the occurrence of a DIF might cause an outlier-like sales/consumption value, you can suppress the outlier correction function for periods that contain a DIF occurrence and instead explain the exceptionally high value by the occurrence of the DIF.

DIF occurrences can be active or inactive. Inactive DIF occurrences are ignored by the forecast calculation. If you are not certain whether an event will occur, you can make the DIF inactive, then reactivate it when the situation is resolved.

There are two special types of DIFs:

  • Correction factor. You can correct forecasted demand values with a manually entered factor (for example, for first-time or one-time occurrences of a regular DIF).

  • Ignore. You use an ignore DIF to exclude historical demand periods from forecasting (for example, in the case of unforeseen events that distort the consumption figures). The system uses average historical demand values instead.

DIF data that is no longer needed can be reorganized (deleted from the database),