Global Demand Forecast Settings – StockIQ Technologies

The Global Demand Forecast settings screen controls various system-level ("global") settings in your StockIQ forecasting configuration.

It is accessible by clicking on Admin --> System Configuration --> Global Demand Forecast Settings

General Settings

Top-Down Disaggregation Mode - This sets how forecasts are distributed when you save them from a top-down level.
- Skip Overridden Child Forecasts - This is StockIQ's default. Any children with their own overrides are skipped, reducing your initial total forecast.
- Prorate For Overridden Child Forecasts - Your top-down value is preserved, as are child forecast quantities. The total to each child node is adjusted to make sure the final total is the same as your initial quantity.
  - WARNING: This can lead to warnings/errors if your new total is less than the sum of child forecast totals, and StockIQ will warn you of this.
Operational Forecast Series - StockIQ supports multiple forecast series. Generally, you will have just one, so for most folks this is always going to be the first, default, Independent Forecast series.
Sales Forecast Series - Specifically determining which forecast series, if any, is particular to salesperson input. This is a placeholder for future features.
Budget Forecast Series - Specifically determining which forecast series, if any, is particular to accounting/finance's input. This is a placeholder for future features.
# Years Forecast History - How many years of history to consider when generating forecasts. 3 years is a good guideline here
# Years of Future Forecast - How far into the future should we generate forecast. 2 years is usually more than enough here.

Forecast Error

Error Calculation - which error calculation to use, MAE% (Mean Absolute Error Percent) or RMSE (Root Mean Squared Error). For more discussion on which one to use, see the Forecast Error KPIs topic.
Error Sample Length - What duration to use when calculation the standard error numbers shown in the application, such as on the Forecast Manager Forecast Errors tab. One Quarter, 6 months, and 1 year are the available options.
Low Error Max - Sets the threshold for what you consider "Low" forecast error.
Medium Error Max - Sets the threshold for what you consider "Medium" forecast error.
High Error Max - Sets the threshold for what you consider "High" forecast error. These three settings affect coloring/highlighting in the Forecast Error tab.

Forecast Snapshots

Monthly Forecast Snapshot Target Day- Sets which day of the month Demand Forecast Snapshots are captured
Do One-Time Recalculate Historical Error Values- When enabled, StockIQ recalculates and re-saves the forecast error values stored with each forecast snapshot. You normally don't need this. It's useful if your demand history has changed (for example, after a history correction or re-import) or after you change the Error Sample Length. The setting turns itself back off after the error values refresh successfully.

Fair Share Weights

Fair Share Weight # Months History - Allows you to configure how many months are used when evaluating the fair-share-weights to be used in top-down forecasting and the resulting distribution of top-down forecast.
Include Current & Future Periods In FSW Calculations- Sets whether sales in the current month and future
months (open or closed orders) count toward fair share weights. When disabled (the default), fair share
weights use only completed periods, from the start of the Fair Share Weight # Months History window through last month. When enabled, the window has no end date, so current and future-dated sales also shape how top-down forecasts are split among child nodes. That makes the split more responsive to orders already on the books, but it can move around more as new orders come in.
Re-Apply New Fair Share Weights Each Period - When you have a manual forecast, when you save a top-down manual forecast, you can select that this forecast is locked in time, or, when enabled, this will re-apply your top-down forecast according to changing fair-share weights over time, so that the distribution of your forecast to the children of your top-down forecast stays updated as buying mixes change.

Cost & Price Calculations

Calculate Customer-Specific Average Costs & Prices - When enabled, StockIQ will look at historical demand data and calculate expected per-customer average cost and price information for use in Revenue and COGS forecasting. It starts by looking at the last 3 months of data. If no data is found, we will go to 6 months, and finally to 12 months of data to try and find item-site-customer history. If no data is found, or when this option is disabled, item-site level values from the history table are used only, and if no history is available, current-day StandardCost and Standard Price values are used.
Allow Zero As Valid Cost Or Price? - Allows zero cost or revenue values to be marked as valid data points instead of being marked as missing data points

Stat Model Adjustments

Adjust Model for Future Sales - StockIQ has the option of upwardly adjusting the statistical model in response to future sales, if those future sales are higher than what the model called for. This is useful if you are doing forecast consumption in your Replenishment Forecast Series settings, but is generally not used.
Apply Known-Demand Events To Future Forecast- When enabled, known-demand events (lost customer, customer move, sales-order-linked, and exceptional sale events) apply their effects to current and future forecast periods, not just to history. The matching sales orders then consume the replenishment forecast so the demand isn't counted twice. When disabled (the default), these events only adjust history and never raise or lowerthe future forecast. Auto-detected unusual sale events never affect the future forecast, whatever this is set to. Takes effect on the next Calculate.
Known-Demand Sales Orders Consume Forecast- When enabled (the default), sales orders that match a lost customer record or are linked to a known-demand event consume the replenishment forecast like any other order. Disable it if those orders should be left out of consumption entirely. They are already kept from raising the forecast. You can't disable this while Apply Known-Demand Events To Future Forecast is on, because raising the forecast without consuming it would count the same demand twice. Takes effect on the next Calculate.
Calculate Depletion Forecasts - When an item is detected as discontinued, this option enables calculation of a depletion forecast, so you can predict when on-hand and demand will drop to zero for the product. See the Discontinued Items screen for more.

Seasonality Detection

These settings control how items are marked as no, low, or medium seasonality in StockIQ:

Min Avg Annual Quantity - Very low volume items can sometimes appear, coincidentally, to be seasonal, when in fact they just have sporadic demand. This filter sets a lower limit on how many sales per year the item must have on average to be considered for seasonality. Lowering this value makes an item more likely to be considered for seasonality.
Min Months of Demand History - Specifies the minimum length under which an item is not evaluated for seasonality. This is in place because StockIQ must have some amount of year-over-year overlap to be able to determine the difference between yearly seasonality and simple variation in demand.
Seasonality Min. Correlation Threshold - Specifies the minimum value of year-over-year correlation under which an item is considered to be even mildly seasonal. Items with 1.0 correlation are 100% seasonally correlated, items with 0% or less have no evident correlation for year-over-year demand. Raising this limit makes items less likely to be considered for mild seasonality.
High Seasonality Trend Calculation- Sets the method StockIQ uses to calculate the year-over-year trend
(slope) for highly seasonal items:
- Seasonality Year (the default) - Draws a trend line between each pair of years and weights recent years more heavily. It also has safeguards against sudden jumps in the trend from one year to the next
- Linear Regression - Fits a standard straight best-fit line through all of the item's history, with every period weighted equally.
High Seasonality Min Correlation Threshold - Specifies the minimum value of year-over-year correlation under which an item is considered to be highly seasonal. Items with 1.0 correlation are 100% seasonally correlated, items with 0% or less have no evident correlation for year-over-year demand. Raising this limit makes items less likely to be considered for the special highly seasonal method of the StockIQ forecast algorithm.
High Seasonality - Min Std Dev of Demand - Specifies a minimum standard deviation above which variation of actuals must be to be considered for seasonality. Lowering this makes seasonality
more likely to be allowed, whereas raising it means more items will be set to no/flat seasonality. This test primarily affects very low volume items, to prevent very low volume items from being falsely detected as seasonal among slow or sporadic demand

Under & Over Forecasted Reports

These settings control how much data is generated on the under and over forecasted reports, such as what is considered an under- or over- forecast, whether to alert on current period settings, and so on.

Tolerance Factor - This factor, expressed in terms of a number of standard deviations, specifies how tight the limits are on what StockIQ considers out-of-range for the under or over forecast comparison. A higher tolerance factor allows a wider variance before an alert is generated. Standard Deviation here is used since it is more useful than a flat percentage, since different items have different amounts of observed variability in their demand history.
Over/Under History Length - How far back to be checking for over- and under-forecast records. This determines how much data you see in the report. One quarter is pretty typical. By limiting this threshold, you can avoid getting alerts for periods from 6-12 months ago if you don't want to make adjustments for those.
Alert on Low-Volume Items - Allows you to disable the alert for items with less than 1 unit per period of forecast. Generally recommended to avoid getting alerts that are not helpful for items that don't matter much.
Alert On New Items - Allows you to suppress alerts for new items, since they will almost surely have periods outside limits, as there is limited data. Once the item is not new, alerts will appear.
Alert On Sporadic Items - Allows you to suppress alerts for items with sporadic/slow/dead usage types, since most activity for those items will be outside of forecast expectations. However, notifications for activity on such items can be useful.
Alert on Non-Forecast Policies - Allows you to set whether you want to see alerts for over/under forecast on items that are set to order policies that do not use forecast, such as Buy To Order, Min/Max, Build To Order, etc.
Show For Current Period - Sets if you want to see under and over forecasted alerts for current period or not
Show for Future Periods - Sets if you want to see under and over forecasted alerts for future periods or not.
Min % Thru Period - Specifies how far the period you must be before current period forecast vs actuals messages begin generating. This is intended to prevent getting messages that are not helpful just a few days into a month or week.

Forecast Export Settings

StockIQ provides standard forecast export tables in easy-to-read formats. These settings control what data is published into these tables. By default, no data is outputted, so you must turn these options on if you will make sure of the outputted data. Item-Site, and Item-Site-Forecast-Group level detail is available for both independent and dependent forecast, published to tables in the "dato" schema for data output in the StockIQ database. Please contact StockIQ for more information on consuming these tables.

Forecast Summary Update Frequency - how often you would like these tables updated.
# Years to Export - How many years of historical forecast to publish to the output tables.
# Years of Future To Export - How many years of future-dated forecast to publish to the tables.
Generate Monthly Export - enables or disables monthly forecast summary table creation. Disabling saves some startup/calculate/refresh time.
Generate Weekly Export - enables or disables weekly forecast summary table creation. Disabling saves some startup/calculate/refresh time.

Auto/Manual Forecast Tracking Mode

Forecast Tracking Mode - Sets how StockIQ tells apart the forecast periods you've set manually from the ones it forecasts automatically. Changing this converts every item's existing forecast settings to the new mode, and a full statistical forecast recalculation runs on the next refresh. New installs default to Per-Period
Locks. Upgraded systems stay on Auto-Resume Date until you change it.
- Auto-Resume Date (classic) - Each item gets one cutoff date. Every period before the date keeps your manual forecast, and StockIQ auto-forecasts every period from that date on. Editing a period further out moves the cutoff, so every period before it becomes manual too.
- Per-Period Locks- Each period you edit is locked on its own, and StockIQ keeps re-forecasting every period you haven't locked, before or after the locked ones. Saving a manual edit doesn't switch the whole item to manual. You can lock or unlock periods by right-clicking them in the Forecast Manager grid. Locked periods show a lock icon in the grid and a pinned marker on the chart, and the Settings tab lists them with Unlock and Unlock All options. When you unlock a period, its forecast goes back to the statistical forecast as soon as you save.

NOTE: Switching from Per-Period Locks back to Auto-Resume Date loses information if an item has gaps between
its locked periods. The resume date is set to the period after the last lock, so any unlocked periods in
between become manual. Switching from Auto-Resume Date to Per-Period Locks loses nothing.

Related to