BACKGROUND: Deere & Company is the world's leading producer of agricultural equipment and a leading manufacturer of equipment for construction, forestry, and lawn and turf care. Additionally, Deere provides credit, special technology, and managed care products and services. John Deere Waterloo Works (JDWW) manufactures agricultural tractors ranging from 95 to 425 PTO horsepower at their facilities based in Waterloo, Iowa.

PLATFORMS: The Deere Retail Sales Forecasting System runs on an enterprise intranet exchange Web server.

PROBLEM SOLVED: Getting highly accurate forecasts is very important because the agricultural markets can be very turbulent from year to year. Retail sales can be affected by current and future farm-economic trends, crop prices and production levels, commodity futures, used equipment inventory levels, consumer confidence, etc. Information Discovery's Retail Sales Forecasting application models these complex interactions and accurately forecasts JDWW tractor sales levels successfully a year in advance.

PRODUCT FUNCTIONALITY: The Retail Sales Forecaster provides users with a gateway to knowledge that has been predistilled from data and stored in a pattern-base, providing an environment for highly accurate forecasts. Using the PatternWarehouse knowledge discovery system, users without a technical background can freely access patterns on the intranet using a point-and-click GUI and interact with the system for further drill down and analysis. The operation cycle is as follows: data mining is performed on weekly/monthly data in a SQL database, the discovered patterns are stored in a pattern-base, forecast system applies the trends and patterns for short- and long-term forecasting, the patterns are incrementally combined and trend analyzed periodically. Users query and access the pre-mined pattern-base. Users can build simulations of future sales activity by adding their knowledge to events that have not occurred using the "what-if" scenario module.

STRENGTHS: The accuracy of the system has been extraordinary, considering the recent turmoil in the agricultural community. The system also quickly adapts to the changing trends in the market to ensure consistent, accurate results. Sales patterns are automatically identified and stored in a PatternWarehouse as well as the impact of industry factors such as cash receipts by farmers, crop-future contract prices, etc. The pattern-base is incrementally updated, ensuring that Deere users always access recent sales patterns. The system is Web-browser based and allows users to immediately access both historical data and forecasts for the future months. Business users can access refined information on their intranet and obtain forecasts without needing a technical background.

WEAKNESSES: Within the agriculture sector, some variables with impact on retail sales (e.g., industry- wide incentives) are not easily quantifiable and difficult to incorporate into models. The impact of these is also dependent on similar incentives offered from competitive manufacturers as well as current governmental programs.

SELECTION CRITERIA: This product is unique because it provides an extremely high level of accuracy. The system uses refined patterns, rather than data. These patterns are then used to develop a predictive model. The model is built on the real database, not on a sample and, hence, provides higher quality grains of information through data distillation. The system is also capable of adapting quickly to market fluctuations. The Java-based Web access to patterns enables users to access patterns simply using a Web browser. No technical know- how is required. Another unique functionality is the quick response time: the patterns are available for the users on demand. The system is fully capable of modeling highly complex interactions and forecasts sales trends by considering a number of different factors.

DELIVERABLES: By using Information Discovery's pattern-based approach to data analysis, Deere can forecast tractor sales with an extremely high level of accuracy. The data mining approach used by Information Discovery produces highly accurate forecasts, because it uses many more parameters in the patterns it uses. Accurate forecasts allow Deere to better manage its inventory and make significantly better strategic marketing decisions.

VENDOR SUPPORT: Information Discovery has skilled consultants who have been very supportive. We were very impressed with Infor- mation Discovery's commitment to our pattern discovery project.

DOCUMENTATION: The help files and instruction provided made this complex system easy for both business users and IT staff to operate.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access