Skip to content

Parameters in Demand Forecasting

· Forecast Parameters are the specifications that have a significant effect on the Forecast Plan.

· Following are the Parameters used in the Application.

Parameters

  1. Forecast Start Date: This is the date from which the forecast has to begin. The date will be set when we move from one forecast cycle to another.

  2. Statistical Start Date: This is the date from which the Statistical forecast begins. If nothing is specified, then it will be taken as Forecast Start Date.

  3. Forecast Horizon Years: This is the number of years for which we are forecasting the demand. This is in Years.

  4. Forecast History Horizon Years: For any forecast, sales history is required and Forecast History Horizon Years says how many years of previous sales are considered for forecasting.

  5. Disaggregate History Days Limit: This parameter defines how the aggregate level like Product Group and Region Group is Disaggregated into the Stock Keeping Units (SKU). Based on the number of history buckets to be considered, it creates the contribution of each of the SKUs and uses this factor to disaggregate.

  6. Number of Decimal Places: Number of decimal places to consider rounding off Forecast output.

  7. Store Forecast in Monthly Bucket: Whether to store the forecast value in Monthly or Daily bucket. This has implications for the performance. It has True/False dropdowns. It should be set to False only if there is a business need to see the forecast value in the daily bucket.

  8. Weekly Calendar Type: This Parameter specifies if the Forecast Calendar generated is to be Week of Year or Week of the Month.

  9. Outlier Type: An Outlier is a data point in history that diverges from the other data points. An Outlier can either be overly high or overly low compared to the other data points in the time series. In this parameter, the user can specify if the Outlier should only be Detected or it should also be corrected.

  10. Outlier Standard Deviation: Sigma value that should be considered for outlier detection.

  11. Accuracy Horizon: While calculating the forecast error, how much of historical sales should be considered.

  12. Statistical Forecast Buckets: For how many buckets statistical forecast should be generated.

  13. Split Type: Possible values are Average of Proportion and Proportion of Average. If there are 4 items and 2 months, then the Average Proportion for an item is the sum of two months for that item and divided by the sum of all items for the two months. In proportion to the average, we find the proportion of each item for each month. We then take the average of the proportion for an item to arrive at the split factor for each item.

  14. Disaggregate Based on Last Year: This parameter has two dropdowns, True/False. This will decide whether disaggregation of the statistically generated forecast should happen based on historical sales from the forecast date or one year before the forecast date. If Forecast Date is 1st June and if the Flag is False, then it will consider the previous months from 1st June. If the Flag is True, it will take the forward months from 1st June but last year.

  15. Accuracy Horizon Months: This is to specify, how many months Forecast and History need to be considered for forecast error calculation

  16. Recalculate All Techniques for Accuracy: This parameter is dependent on Accuracy Horizon Months. When Accuracy Horizon Months is greater than 0, then by default, only the Best Fit forecast data measure will be recalculated from the Forecast Start Date. If this flag is True, then all the techniques will be recalculated from the Forecast Start Date.