Forecasting
FGS - HomeDemand ForecastingInventory PlanningDatabase ManagementReplenishment PlanningDistribution Planning
The Finished Goods Series
from E/Step Software Inc.
Demand Forecasting
Forecast Module
FGS - Home
    Key features
    Implementation
    Industries served
    Clients
    Undecided?

About E/Step Software

Services
    Consulting
    Education
        Introductory Workshop
        Intermediate Workshop
    Outsource
    Support


Articles

Contact

Forecasting
  Forecast Model Elements
   Outliers
   Tracking Signal
   Calendars
   Fourier Series


Inventory
  Compare service
   Set service
   Component SS


Database
   Modules & Interfaces

Replenishment Planning

Distribution Planning

The FGS Forecasting Approach

A forecast is a range, not a number. For example you could say that the forecast for an item for next month is 100 (most likely) but with 98% certainty the demand will not exceed 115. The extra 15 is safety stock which achieves your 98% service target and is computed from the item's forecast error.FGS uses the demand history for each SKU (or sales history if you do not have demand history) to compute a mathematical model of the demand. It then projects the model into the future. FGS contains multiple families of forecasting models, including the Fourier Seasonal Profiles, which usually work best where seasonality is a possibility. The system's automatic model-fitting logic determines the level, trend, and appropriate seasonal cycles, if significant. The user can override these decisions, but all changes are evaluated, to encourage what works and discourage what does not.

Most forecasting systems have a model fitting process, but lack a separate forecast revision process—instead refitting every SKU every month which creates instability and turmoil. A forecasting revision process is critical for identifying fundamental changes in the demand patterns as they occur.

Initial Model Fitting Process

Elements of a Forecast ModelSeasonal SKU

This SKU exhibits a growth trend and seasonality.

Although FGS supports multiple forecasting models we have found the Fourier Seasonal Profiles usually provide the most accurate forecasts. This method is implemented as an expert system designed to minimize the forecasting workload. FGS helps focus the analyst's attention where most needed to make the greatest improvement in forecasts, inventory, and customer service.

A bucketless forecasting system lets you pick the right calendar

Calendar Change
Calendar Simulation
This low volume SKU's forecast improves when changed from a monthly to a quarterly calendar.

An exclusive capability of FGS is the ability to identify and use the forecasting calendar most appropriate to each SKU, allowing items to be forecasted bimonthly, quarterly, semiannually, or annually; yet all items reside in the same database, regardless of calendar. The resulting forecast error reduction provides substantial benefits—in the form of lower inventory or higher service—for service parts and other hard-to-forecast or low-turnover businesses. These same user-defined calendars also enable users in completely different businesses (e.g., high-turnover process businesses such as food or chemicals) to forecast demand biweekly, weekly, or even semi-weekly. Contrary to the assumed aggregation phenomenon, these businesses often get a more accurate forecast when using a more frequent calendar because aggregation sometimes obscures the apparent demand patterns.

The forecasting calendar implementation in FGS also eliminates the error inherent in assuming that a period has the same length from one year to the next. For example, with calendar months, January could have as few as 18 or as many as 22 selling days, in different years. That's as much as a 22% error! Without this logic, which is exclusive to FGS, you are forced to keep extra safety stock against the possibility that a period could have more selling days this year.

Simulation facility to review and modify forecasts

Outlier Simulation
Outlier Simulation Graph

Increasing the outlier sensitivity from 4 to 3 sigma flags an outlier and reduces forecast error.

There are occasions when you need to look at a SKU's history and forecast and consider overriding the forecast model. The forecast model might no longer be appropriate. You might want the forecast to respond differently to a specific event. Or you might question why a particular forecast model was created. One of FGS' most popular features is the Simulation program. A picture is worth a thousand words.

The Simulation screen is divided into two windows: the current model and the simulated model. The current model is on the top. You edit the simulated model and view the results on the lower window. You can see the changed forecast as well as the changes in other data, including the inventory detail.

Click on the Outlier Simulation graph on the right to see an example of the simulation tool at work.

Periodic revision process updates the forecast and flags out-of-control forecasts

Tracking Signal
Simulate Tracking Singal
This example shows a SKU's demand climb over the last few months. It is caught with a tracking signal.

FGS includes many Statistical Process Control (SPC) tools to monitor the forecasting process, identifying those SKUs which are going outside control limits, and those which are doing fine on "autopilot." This lets users with many thousands of SKUs focus their attention only where it's needed.

Forecast at multiple levels

FGS supports a multi-tiered forecasting approach at any user-defined levels. It allows summarizing and/or forcing forecasts (up or down) according to a pyramidal organization, that can be created or changed at any time. The system even allows multiple (potentially overlapping) pyramids.

The Forecasting module of FGS is designed to support the management of make-to-stock products (finished goods or service parts) as well as assemble-to-order products and even pure distribution environments. Forecasts are generated for individual Stock-Keeping-Units (SKUs.) An SKU is a part stocked in a specific location. You can define a location to be anything, including a customer.

For example, an oil seal stocked in a Chicago warehouse is one SKU. The same oil seal in a Dallas warehouse is a different SKU. Forecasts are also generated for total product or part demand (the total demand for oil seals from all warehouses.)

FGS also contains a full bill of material (BOM) system. You can load the BOM and drive down the forecast to the components. If you manufacture plus sell service parts, a component part of a finished good could have separate dependent and independent forecasts in FGS.