Inventory Planning > Inventory Optimization > Processing > IO Stock Levels Modeling > IO Forecast Accuracy Query

IO Forecast Accuracy Query

You use this program to asses the accuracy of previous forecasts.

The program calculates the accuracy scores (Standard deviation, Mean Absolute deviation and average hits), which are used as inputs to either the Poisson or Normal algorithm in the IO Stock Levels Modeling program.

Toolbar and menu

Field Description
Functions  
Stock Code Options Select this to use the Options Maintenance program to maintain the forecasting options to use.
Policies Select this to use the IO Policies program to maintain policy options to generate the suggested minimum/maximum levels.
Manual Forecasting Select this to use the Manual Forecasting program to make changes to the forecasts.
Demand History Select this to use the Demand History Maintenance program to adjust the sales history (i.e. the demand history) for the forecast.
View  
Quantities Select this to display quantities in the Details listview.
Sales Values Select this to display sales values in the Details listview.
Cost Values Select this to display cost values in the Details listview.
Selection  
Manual Select this to indicate individual stock items for which you want to compute the forecast accuracy. The Warehouse and Stock code fields are enabled when you select this option.
Batch Select this to use the IO Selections program to indicate a selection set or range of stock codes for which you want to compute the forecast accuracy.
Warehouse This indicates the warehouse for which the forecast accuracy is being computed. You can only enter the warehouse if you selected Manual at the Selection menu.

If you are running the program in Batch selection mode, then you use the Prev and Next options to navigate to the required warehouse.

Stock code This indicates the stock code for which the forecast accuracy is being computed. You can only enter the stock code if you selected Manual at the Selection menu.

If you are running the program in Batch selection mode, then you use the Prev and Next options to navigate to the required stock code.

Change Rev/Rel Select this to enter a different revision/release if the stock item is ECC controlled.
Calculate Select this to re-calculate the forecast accuracy after changes have been made in the Information pane. The results are displayed in the Details listview.

Information

When you make any changes to options in this pane, you use the Calculate function to display the results of those changes in the Details listview. If you are satisfied with the results displayed in the Details listview, you can manually make the same changes to the Modeling Options in the IO Stock Levels Modeling program.

Field Description
Options Apart from the Run date (which defaults to the current system date), the selections in this section default to the selections made on the Modeling tab of the Inventory Optimization Setup program.
Run date Indicate the date from which the forecast accuracy must be run.
Months to compare Select the default number of months to compare for forecast accuracy (1-12).
[Note]

The more months selected the more accurate will be the standard deviation used in the calculation using normal distribution.

Although the entry here is in months, the actual number of periods is dependent on the calendar used (e.g. if you choose 6 months to compare and you are using a weekly calendar then this equates to comparing 26 periods).
Comparison type  
Period on period Select this to calculate forecast accuracy by comparing each period's sales and forecasts to get the absolute deviation.
Moving average Select this to calculate forecast accuracy by averaging the sales and forecasts over a number of periods.
Moving average months If you set the comparison type to Moving average, then you need to enter the number of months to use for the moving average calculation of sales history.

As with Months to compare, this is converted to periods depending on the calendar in use.

Forecast to use  
Last Select this to use the last forecast taken for the period to compare against sales for each period.
Last outside lead time Select this to use the last forecast taken for the period outside of the lead time and dock-to-stock (i.e. planning time fence) to compare against sales for each period.
[Note]

Forecasts generated within the lead time should not be used as it is too late to affect changes based on them.

Average Select this to use the average calculated for snapshots taken for this period to compare against sales for each period.
Average outside lead This uses the average, but only uses snapshots taken outside of the lead time.
Snapshot months If the forecast to use is Average or Average outside lead, then the number of months entered here determines how many forecast snapshots to use when calculating the average.

For example:

You have a weekly calendar and, historically, you created a live forecast for an item for each week.

If you select 2 months here then the snapshot quantities are added for the period taken over the last 8 weeks and averaged out for comparison.

Calendar to use Indicate the IO forecast calendar to use for the forecast accuracy.

Sales vs Forecasts

This displays a graphical representation of the sales versus the selected forecast for the warehouse/stock code combination.

SKU Information

This pane displays the static stock code information from the Inventory Master table for the stock code.

Summary

This displays the forecast's accuracy scores. These are the same scores used as an input to either the Poison or Normal algorithm in modeling (see IO Stock Levels Modeling - Modeling options - Calculated information). These algorithms are used to generate the min/max levels in modeling.

These values can vary widely depending on the number of Moving average months you select and the actual accuracy of your forecasts over those months.

Details

The forecast accuracy results are displayed in this listview when you select the Calculate option. The forecast values are compared to the actual sales values.

The values displayed are either the quantities, sales values or cost values depending on your selection in the View menu.

Additional columns can be added to the listview by right clicking in any column heading and selecting the Field Chooser option (see Listviews).

Column Description
Hits This indicates the actual number of sales for the selected period.
Absolute error This is the difference between the Forecast value and the Sales value expressed as a positive value.
Status This indicates whether you over forecast (forecast is more than sales) or under forecast (forecast is less than sales).
CFE

This is the cumulative forecast error.

This measure represents the difference between the forecast and the actual sale(s) for the period under review, displayed cumulatively.

A positive number indicates a forecast that is too high for a period and a negative number indicates a forecast which is too low for a period.

MAD

This is the calculated mean absolute deviation.

The mean absolute deviation (MAD) is the average absolute deviation from the mean (i.e. the average error, ignoring the sign of the error).

MSD

This is the calculated mean square deviation.

The sum of the squared forecast errors for each observation divided by the number of observations. It is an alternative to the Mean Absolute Deviation except that, because the errors are squared, more weight is placed on larger errors.

MAPE

This is the calculated mean absolute percentage error.

This is calculated as the average of the sum of all the absolute percentage errors for the data set. The absolute values of the percentages are summed and the average is computed.

The difference between actual value and the forecast value is divided by the actual value. The absolute value of this calculation is summed for every forecast point in time and divided again by the number of fitted points.

Tracking signal

This is calculated as the algebraic sum of forecast errors divided by the Mean Absolute Deviation.

It indicates whether the forecast average is keeping pace with any genuine upward or downward changes in demand.

Error % This is calculated as: (Absolute error / Forecast) * 100
Accuracy % This is calculated as: (100 - Error %)

Period Information

This pane displays the forecast accuracy results and related information for the stock item for the period currently highlighted in the Details listview.