ORMS Today
October 2000

SmartForecasts for Windows Version 5

Product demand and sales forecasting package packed with innovative features

By Jon H. Marvel

Smart Software's release of SmartForecasts for Windows 95/98/NT/2000 provides a software package for product demand and sales forecasting that includes several innovative features that improve forecasting accuracy. Some of the capabilities include algorithms that: address seasonality, examine the effects of future and past promotional events, allow for automatic forecasting method selection, provide multiseries forecasting for product groups at the group and/or item level, and include the ability to automatically hedge the forecast through the application of trend constraints.

Product Description and System Requirements

SmartForecasts is available in three editions depending upon the quantity of items to be forecasted. A brief description of these editions is listed below:

Professional Edition: With capabilities to forecast up to 150 variables or items at a time.

Commercial Edition: Includes additional features such as Batch Processing and the ability to forecast up to 1,000 items at a time.

Enterprise Edition: Designed for very large forecasting jobs. (Processing speed of 100,000 items per hour on a Pentium PC is expected.)

Smart Software recommends the following minimum system requirement for use with this software:
  • Pentium PC or successor

  • Windows 95, Windows 98, Windows NT or Windows 2000

  • CD-ROM drive or 3.5 inch disk

  • 32 MB of RAM

  • 20 MB of available hard drive space

  • Color monitor with 800X600 or higher pixel resolution
For this review, the Professional Edition of SmartForecasts was installed and tested on a Windows 98 PC with a 133 MHz processor and 64 MB of RAM.

User Interface

The program interface allows for ease of navigation and includes many features similar to those that appear in popular Windows applications such as Microsoft Office so that users can become proficient in a short amount of time. Data can be imported into the program from a variety of sources, including Excel spreadsheets, Lotus 1-2-3 spreadsheets and a variety of ASCII file formats. Additionally, since this is a Windows application, there is an ability to "cut and paste" from other Windows applications.

An Example: Automatic and Multiseries Forecasting

Automatic. In the following example, the raw data was stored in a Microsoft Excel 97 spreadsheet and was accessed directly by the SmartForecasts software. The default method of data representation is for each variable/item to be represented by a row within the data table while each column represents individual cases (i.e., time periods) within the variables, as shown in Figure 1.

Figure 1: SmartForecasts main window.

In the Automatic forecasting feature, SmartForecasts utilizes six extrapolative forecasting methods: simple moving average, linear moving average incorporating a trend component, single exponential smoothing, Brown's double exponential smoothing method and Winter's exponential smoothing methods (additive and multiplicative models). After selecting the variable/item to forecast, the user can specify all (or a subset) of these methods to be included in the Automatic forecasting tournament, and indicate other forecast parameters such as data cycles and forecast horizons as shown in Figure 2. Additionally, the user can specify whether or not to apply any dampening factor to the trend of the forecast through the use of the Automatic Trend Hedging feature.

Figure 2: Automatic forecasting method selection.

The initial output of the forecast, as seen in Figure 3, is a graphical representation of the data. The graph shows the historical data plotted along with the predicted point and interval estimates using the optimal forecasting method, the one that minimized mean absolute forecast error in the forecasting tournament. From the menu at the bottom of the Forecast Graph, the user has several methods of interaction with the forecast data. The tournament rankings, shown in Figure 4, indicate the absolute and relative performance of each of the forecasting methods including the forecasting parameters the software selected for each method. In the Automatic forecasting solution, when the software identifies the parameters for either the multiplicative or additive Winter's method, the smoothing constants chosen for all components will be set at the same value. The user can set these components individually when generating the forecast in the manual method.

Figure 3: Forecast graph window.

Figure 4: Forecast method-ranking report.

From the Forecast Graph menu, the user can examine the forecasted data for the forecast horizon (see Figure 5), review the smoothed data which utilizes the selected forecasting method to smooth the historical data, or adjust the forecasted data on a case-by-case basis. Once the user has finalized the forecast, the forecast results can be saved in the same file and format as the originating application. For this example, the results were examined, modified or adjusted if necessary, then saved directly to the data table which maintained its Microsoft Excel 97 spreadsheet file structure and type.

Figure 5: Forecast report.

Multiseries. In the Multiseries forecasting method, after the selection of the component items that comprise the product group, the user chooses either a bottom-up or top-down forecasting approach. This choice depends on whether the user wishes to create the aggregate group forecast by summing the forecasts for the individual items (bottom-up) or by distributing the total group forecast down proportionally among the individual components (top-down). In this example, Products A, B and C were chosen as a product group and the bottom-up forecast was made for the group, as shown in Figure 6.

Figure 6: Multiseries forecast graph.

The audit report generated through the Multiseries analysis indicates the individual forecasting methods selected for each product item (see Figure 7). This table also indicates the performance characteristics of each product's forecast method in terms of average absolute forecast error and average percentage forecast error. The forecast data table generated from this Multiseries analysis, as shown in Figure 8, indicates both the historical (plain font) and forecasted data (bolded font) for the individual products, as well as the product group.

Figure 7: Multiseries audit report.

Figure 8: Multiseries forecast table.

User Manual

Smart Software provides users with a comprehensive user manual that includes tutorials, technical reference and forecasting overview information. The tutorials are well documented, easy to follow, and extend to many features that are incorporated into the software. The technical reference section documents the methods that SmartForecasts implemented in the forecasting and data analysis routines. The overview on forecasting discusses some of the basics regarding forecasting and also introduces some advanced forecasting techniques along with further explanation of their forecasting output reports.


Smart Software's recent release of SmartForecasts for Windows Version 5 provides a forecasting package that is versatile and addresses the needs of today's users. The variety of forecasting techniques available to the user, including methods that were not covered in this review such as multivariate regression analysis, promotion/event forecasting and the new, patented intermittent demand forecasting, provide users with a wide array of tools to completely analyze their data and generate solutions.

SmartForecasts also incorporates several features that allow users to include qualitative judgment into the forecasting process. These features permit user sto modify the entire forecast trend or individual forecast points on a case-by-case basis. Another strength of the software is its ability to append the forecast data to the original historical data file in the original file format, facilitating the ability to share this data across the organization in a format that is not specific to the forecasting process. Overall, SmartForecasts provides users of all ability levels the opportunity to generate accurate forecasts while still maintaining control over the forecasting process.

Jon H. Marvel is an assistant professor within the Manufacturing and Engineering sector of the Integrated Science and Technology department at James Madison University. He has more than 10 years of industrial experience and was employed as a manufacturing consultant prior to joining James Madison University.


SmartForecasts for Windows Version 5
is available from:
Smart Software, Inc.
Four Hill Road
Belmont, MA 02478
Telephone: 1-800-SMART-99 (800-762-7899)
Fax: 617-489-2748
E-mail: info@smartcorp.com
URL: www.smartcorp.com

Product License* Annual Support**
Professional Edition $995 $300
Commercial Edition $3,995 $800
Enterprise Edition $13,500 $2,700

*Call for discounted prices on multi-user licenses.
**First year support and product maintenance, which is required, includes customer technical support and FREE product updates.

Vendor's Comments
Editor's note: It is the policy of OR/MS Today to allow software developers an opportunity to clarify and/or comment on the review article. Following are comments from Charles N. Smart, president of Smart Software, Inc.

We would like to thank Professor Marvel for his careful review of SmartForecasts. Since he focused on the Professional Edition of SmartForecasts, we thought it would be valuable to mention some of the additional capabilities found in the product's other two editions, SmartForecasts Commercial and SmartForecasts Enterprise. Also, the version he reviewed, Version 5, has since been succeeded by Version 5.1, which contains several new features of potential interest and value to the reader.

SmartForecasts Commercial combines Automatic Forecasting with rapid Batch Processing to quickly process forecasting jobs containing up to 1,000 items at a time. Users of SmartForecasts Commercial and SmartForecasts Enterprise also enjoy Universal Database Connectivity, which provides direct access to sales and demand history in major client/server databases, such as Oracle, DB2 and SQL Server, and popular PC databases, such as Microsoft Access. Both ODBC and native database connectivity links are supported.

SmartForecasts Enterprise is designed for users with very large forecasting jobs, such as manufacturers and distributors who must forecast the demand for thousands or even tens of thousands of product items in inventory. With only a few clicks of the mouse, users can easily create accurate demand forecasts for every item, along with item-specific estimates of service level inventory requirements, at processing speeds exceeding 100,000 items per hour. SmartForecasts Enterprise users can import data directly from virtually any host database or planning system, analyze the data, and automatically transfer forecast results back to the database or planning system enabling companies to achieve more accurate demand forecasts at a fraction of the cost of proprietary forecasting modules.

SmartForecasts Enterprise contains an optional module which features Smart Software's new, U.S. patented technology for accurately forecasting sales and inventory requirements when product demand is intermittent (i.e., irregular, with a large proportion of zero values). Early users of the intermittent demand forecasting technology not only report nearly 100 percent accuracy in predicting service level inventory requirements but also project millions of dollars in annual savings due to inventory cost reductions.

The new Version 5.1 of SmartForecasts adds OLAP Data Filtering and "Send To" e-mail capabilities to the product's robust portfolio of forecasting and data management features. Data Filtering allows users to "slice" their multidimensional database and focus in on the specific subset of data they want to forecast, making analysis faster and easier. The new "Send To" feature supports collaborative planning within companies and between companies and their customers by letting users instantly distribute forecast reports and graphs via e-mails created within SmartForecasts.


  1. Hanke, John E., and Reitsch, Arthur G., "Business Forecasting," Third Edition, Allyn and Bacon, 1989.

  2. Smart, Charles N., and Willemain, Thomas R., "SmartForecasts for Windows: User's Guide," Smart Software, Inc., 1999.

Jon H. Marvel is an assistant professor within the Manufacturing and Engineering sector of the Integrated Science and Technology department at James Madison University. He has more than 10 years of industrial experience and was employed as a manufacturing consultant prior to joining James Madison University.

  • Table of Contents

  • OR/MS Today Home Page

    OR/MS Today copyright 2000 by the Institute for Operations Research and the Management Sciences. All rights reserved.

    Lionheart Publishing, Inc.
    506 Roswell Street, Suite 220, Marietta, GA 30060, USA
    Phone: 770-431-0867 | Fax: 770-432-6969
    E-mail: lpi@lionhrtpub.com
    URL: http://www.lionhrtpub.com

    Web Site Copyright 2000 by Lionheart Publishing, Inc. All rights reserved.