The Volume Workspace

The Forecast Planner works with one week at a time. Forecast weeks always start with the day of the week defined for the category being viewed. Once you select a week here, the forecast factors display is labeled in relation to the selected week. The week you select is relevant for viewing, generating or editing a forecast.

The time period the forecast planner defaults to is configurable in Forecast Planner configuration setup

By default, the current forecast week is displayed with all defined volume drivers for the selected location or category. You can choose different ere-defined views or create your own custom views. See Forecast Factors for details.

Traffic patterns are a way to introduce a specific pattern of volume in 15-minute intervals, for a specific site or department, on a specific day. For details see Export and Import Traffic Patterns.

Special events are indicated on the Forecast Planner by three vertical dots in the date header. The indicator appears for both volume forecast data and volume actual data. If the special event icon appears on a given day, it indicates that one or more special events occur on that day for a loaded location or locations.

For details, see Special Events.

Note: Make sure to leave a reasonable time delay between the end of the volume import and the start of the volume forecast run. This delay allows time for the system to calculate the effects of special events. Leave at least 10 minutes; the recommended delay for a large organization is about 30 minutes.

1. In the Forecast Planner, select the Volume workspace.
2. Select a Forecast Week for the volume forecast.
3. Select a location for the forecast.
4. If you select a category above the site level or below the department level the Run Volume button is disabled.
4. If no volume forecast has been run for the selected category and forecast week, the volume forecast display is blank.
5. Click Run Volume . The Run Volume panel opens.
6. To generate the Volume Forecast using only volume intervals, select Generate Volume Intervals Only.
7. Select the loaded location you want to run.
8. Click Run. The new forecast overwrites any previous values for the selected location.
9. To check the progress of the volume run, click Tools > Engine Status. The Engine Status panel opens showing active engine runs and their status. Submitted at and Completed at reflect the timezone of the current user account; to show the date-times in UTC, click tap the Information icon. To update the engine status, click tap Refresh .

When you run a volume forecast, the system saves this generated forecast and displays the values in the Forecast Planner for the selected Forecast Week and category.

If you generate another volume forecast for the same Forecast Week and location, the new forecasted values replace the previous values. Only values for the most recent generated forecast are displayed.

If you edit the generated forecast and save the edits, the system saves the edited forecast. Changes appear where you make them. The system generated forecast remains unaltered.

If you are working with forecast data and you want to discard your unsaved edits, click Refresh. This discards your current edits and restores the workspace with the last saved data.

Note: Discarding forecast edits does not restore the last generated forecast. When you edit and save a generated forecast, the system saves the modified version of the Forecast, and retains the original as the System-Generated forecast.

Restore Volume

Select Restore Volume to delete the edited forecast and display the last generated forecast. All edited volume forecast data for the category, sub-categories, and time period are deleted.

You can make edits to the volume forecast in the Volume Forecast workspace; for example, when you are comparing forecast data to actual volume data from similar periods in the past. You can also make edits to the volume forecast in the Labor  Forecast workspace when you are comparing volume forecast data to labor forecast data.

Your changes affect the entire volume forecast. For example, if you make a change to a sales number on Sunday and click Calculate Totals:

• The total for the week in the Forecast Sales row updates.
• If Sales is a dependent driver, the data in the Sunday row updates for all other dependent drivers. For example, if sales and customers are dependent drivers and you make a change to sales, the customer data updates proportionately.
• The totals for the week for all dependent drivers and linked categories update.

  Note: Changes to the volume forecast do not automatically change the labor forecast. If you change the volume forecast and want to see its effect on the labor forecast, you must create a new labor forecast.

To edit a volume forecast from the Volume Forecast workspace:

1. In the Forecast Planner, select the Volume Forecast workspace.
2. Select a Forecast Week for the volume forecast, and select the category for which you are creating a volume forecast.
3. Edit the volume forecast directly in the cells.

Note: Depending on your access rights, you may be able to edit the values for only the Sales Volume Drivers. Rights access to the Items and Transactions Volume Drivers are managed by your system administrator.


4. You can enter a number or a percentage. A percentage can be entered as a positive number (for example 10%) or negative number (for example -10%)
5. After an edit the Calculate Totals button is enabled. You should click this before another edit is made.

6. Note: Clicking Calculate Totals does not save the change, it only updates related values on the screen. This gives you the opportunity to examine the overall effect of the changes you have made before saving them. If the effect is not what you expected, click Refresh. If the effect is satisfactory, click Save.

6. Do one of the following:
   • Click Save to save your edits.
   • Click Refresh to discard the edits and return to the last saved forecast.

After you create a volume forecast, the system saves the generated forecast until you create another volume forecast for the same category and time period.

When you save edits to a generated forecast, the system saves the edited forecast while retaining the system generated forecast.

You can view both the edited forecast and the generated forecast by adjusting the forecast factors view.

Compare a volume forecast with actual data from the past to analyze your current volume forecast. You can edit and save your changes to your volume forecast on this workspace.

Note:  Volume data must have been imported for the category and time frame you are comparing.

1. In the Forecast Planner, set up a forecast factors view that displays the actual values you want to compare to the forecasted values.
2. Select a Forecast Week for the volume forecast, and select the category you want to view.

3. For example, when defining the forecast factors to view you can, select any of the following:

   • Previous Forecast Week - The forecast week before the current forecast week
   • Current Forecast Week - The forecast week that contains today
   • Next Forecast Week - The forecast week after the current forecast week
   • Next Forecast Week + 1 - The week that is 2 weeks after the current forecast week.
   • Next Forecast Week + 2 - The week that is 3 weeks after the current forecast week.
   • Same Week Last Year - The week that is 52 weeks before the current forecast week

   Note: The week that is 52 weeks before the current week may appear quite different from the current week. For example, April 15 may be part of the week April 15-22 this year, but it may have been part of the week April 8-15 last year.

   • Select Forecast Week - You can select any week. Ensure that data is imported for the selected week
   • Click Select Forecast Week.The current month appears.
   • If the desired forecast week is in the current month click any day in that week
   • If the desired forecast week is not in the current month navigate to the month as follows, then click any day in the desired forecast week:
   • To navigate by month into the past or the future click the or , respectively
   • To navigate to a month in any year in the past or the future, click the month name, then click the or , respectively. When the desired year is shown, click the desired month,

You can create a Forecast Factor for any driver showing Forecast % of Actual which displays the calculated percentage of the daily Forecast Volume compared to the Historical Volume Actuals for the selected forecast week. Use the information in this column to compare the forecasts for an upcoming regular or holiday week to a corresponding historical week in terms of percentage difference between the 2 weeks.

When volume forecast intervals are available for a location, the forecasted row in the Forecast Planner appears as a hyperlink. This data is only available at the site level and below.

After you access the Volume Forecast Details page, you can view or edit the intervals provided you have the appropriate rights access. You can also toggle between different views by selecting either Rounded or Raw from the value type dropdown.

Each configured interval, whether it is 15, 30, or 60 minutes, has its own row.

To view volume forecast details:

1. In the Forecast Planner, select the Volume workspace.
2. Select the appropriate row, and click on any cell that is highlighted in blue text. The Volume Forecast Details page appears. Based on your access profile, you can edit and view the interval values.
3. To view rounded or raw values, use the value type dropdown list to make the appropriate selection.

Note: Note: Rounded values may not always sum to the daily total. To view additional information about the values, switch to the raw values view. The raw values view displays both user-edited and system-generated values.

4. To export interval values data to Excel, click Export.
5. Click Return to exit the details view and return to the Forecast Planner workspace.

You can view actual volume data for a day by 15-minute intervals. Showing the data in detail gives you a more accurate picture of when business actually gets done during the day. The data is read-only and the date and category selections are disabled.

Note: If you are storing your 15-minute interval data in a separate database, make sure the database is on-line and accessible.

Note: The grid or table view of Actual Volume always shows only 15-minute intervals. However in Forecast Home, you can show any configured interval in the Daily Volume Details.

To view actual data by 15-minute intervals per day:

1. In the Forecast Planner, select the Volume workspace.
2. In any Actuals row, click on any cell that is highlighted in blue text.

   The day’s actual volume data displays in a table with a row for each 15-minute period and a separate column for each volume driver. To view forecasted values broken down into 15-minute periods, drill down to the detailed view of a specific day.

3. Click Return to exit the daily detail workspace and return to the Forecast Planner workspace.

The Export Volume spreadsheet contains the following worksheets and fields:

Title Page (main worksheet)

• Forecast Week — Week in the volume forecast that is being analyzed.
• Anomaly Score Threshold — Selected Anomaly Score Threshold.
• Volume Driver to Analyze — Selected Volume driver. Note that a missing entry indicates that the primary driver of each category node will be used for that node's analysis.
• Sample Size for Analysis — Selected Sample Size.
• Machine Learning History Year — Selected Machine Learning History Year.
• Only Anomalous Data — If Only Anomalous Data is selected, an "X" is displayed in this cell.
• Analyzed Forecast Category Path — Path in the business structure to the site or department that is analyzed in this spreadsheet.

Summary worksheet (summary information about each labor category that is analyzed in this spreadsheet)

• Worksheet Name — Lists the names of the tabs in the analysis (one row per tab).
• Number of Anomalies — Number of anomalies found for the category.]
• Insufficient Data — If insufficient data existed for the history year that was selected for the analysis, an "X" is displayed.
• Forecast Model — The volume forecast model that was assigned to this location in the Category Property Sets setup page: Daily Trend, Exponential Smoothing, or Machine Learning).
• Days of Volume History — As defined in Category Property Sets setup page for the location.
• Machine Learning History Year — Selected Machine Learning History Year.
• Forecast Category — Path in the business structure to the site or department that is analyzed in this worksheet.
• Number Missing — Percent of days for which POS data is missing per Forecast Category Path. Values of .25 and higher will be flagged as anomalies. The formula that is used is:
• 100{Count(missing)/(Count(A)+Count(missing))
• where A is a specific combination of volume driver and forecast category.
• Assume that five days are being analyzed and one day is missing data. The metric value in this case would be 100{1/5} or .20 (20%).
• % Zero — Percentage of training data that has a value of zero per Forecast Category Path. Values of .20 and lower will be flagged as anomalies. The formula that is used is:
  
• 100{Count(zero-values)/(Count(A)}
• where A is a specific combination of volume driver and forecast category.
• Assume that five days are being analyzed with the values (100, 0, Missing, 200, and 100). The metric value in this case would be 100{1/4} or .25 (25%).
• Mean — Average value of training data. Size is used to group categories together and identify very small categories whose size may not be important for forecast accuracy. The formula that is used is:
• Sum(A)/Count(A)
• where A is a specific combination of volume driver and forecast category.
• Assume that five days are being analyzed with the values (100, 0, Missing, 200, and 100). The metric value in this case would be (100 + 0 + 200 + 100) /4 or 400/4 or 100.
• 75th Percentile — 75th percentile of training data. This value is used to identify intermittent demand. If it is less than 1, the category is so small that it may not need its own category. The formula that is used is:
• 75thPercentile(A) [n = round_up(0.75 * count(A)), then select the value at position n in the value-sorted list]
• where A is a specific combination of volume driver and forecast category.
• Assume that four days are being analyzed. The days are sorted as follows according to value:
• 
• According to the formula:
• n = round_up(3) = 3
• Position 3 is 2019-01-05, so the 75% percentile is 100.

Individual volume driver worksheets (analysis worksheets)

Each volume driver worksheet contains actual POS data from the days that were used to generate the volume forecast. The number of days of data that is listed depend on the volume forecast model that was defined for the driver:

• Daily Trend shows two years of data for each day of the week, sorted together. If the number of valid sorted pairs of days for all days of the week under consideration is less than required as defined in the Days of volume history to use parameter, the system will not export data and will generate the following error message in the Export Volume Spreadsheet:
• Cannot analyze volume forecast; category has insufficient historical data.
• For example, if a Daily Trend is 6 weeks, the Days of Volume history to use parameter is 6. To define sorted pairs the application looks back for the number of weeks calculated as [*Days of volume history to use* parameter: 6 + 3 = 9 weeks] to retrieve POS data and to build the pairs (the current year/corresponding day of the previous year).
• This issue can occur even when there is sufficient data for the Volume Engine to generate a forecast for the specific week.
• Exponential Smoothing shows the number of days of the week (as configured in the Category Property Sets setup page) included for the current year.
• Machine Learning shows the days of POS data used for the selected year (for example, 52 Sundays in the current year).

The displayed columns include:

• Day — Day of the week used for the forecast. For example, if the Daily Trend is 6 weeks, each weekday appears 9 times.
• Sample Number — Number of samples for each day. A “1” indicates the full day, a “0” indicates that there was no data for that day. Note that the Forecaster will go back as far as possible to get the right amount of POS data (i.e. 6-Weeks)..
• Date — Date of the POS data.
• Anomaly — An "X" identifies POS data that exceeds the selected Anomaly Score.
• Event — Special events flagged as Normalize or Discard; they are not flagged as anomalous. When an event is flagged as normalized, the historical data displayed is the actual historical value, not the normalized value. These days were generally were assigned a special event and the event was normalized.
• % Impact of Special Event — Percent impact of special events on a historical day that was used to create the volume forecast (for example, the historical impact (percent plus [+] or minus [-]) of an event on the non-event (normal) data).
• Score — Number assigned by the system for each day of POS data.
• Volume Driver — The column header displays the Volume Driver and the cells in the column displays the actual POS numbers imported for that date.

Special events are used to identify unusual business activity in the past or future. Use special events for:

• Major sales promotions
• Holidays that do not occur in the same week every year
• Unexpected events

The best way to determine what days should be marked as special events is to use the Volume Analyzer (Export Volume Data) tool in Forecast Planner. When doing so:

1. Export the data to be reviewed to an Excel spreadsheet.
2. Look at the anomalies and review the spreadsheet for other days that have high or low volume.
3. Compare those data points to other days with the same sample number. If the volume is exceptionally high or low relative to nearby days, these could be candidates to become special events.
4. For Daily Trend forecasting, compare these data points to the same week and day of the week in the previous year, which is provided in the spreadsheet.
   • If the same type of anomaly is present in both years, it does not need to be flagged as an event.
   • If the anomaly is present in only one of the years, determine year in which it should be the special event by comparing the data point to the others near it. The one that is normal should not be an event.
5. For Machine Learning, each year is in a different spreadsheet. You must review each year to determine if the same anomaly occurs in the same week.
   • If the anomaly is present in each week, then it does not need to be assigned as a special event.
   • If it is present in only one or two years, determine where the special event should be assigned by comparing the nearby days. The data points that are normal should not be special events.